Behavioral Solutions to the Knowing-Doing Gap (Personal Finance)
Research Task — Behavioral Solutions to the Knowing-Doing Gap
Section titled “Research Task — Behavioral Solutions to the Knowing-Doing Gap”Round 1 of 2. Round 2 — Behavioral-Barriers-Playbook — mines other industries for what actually dissolves behavioural barriers in healthy, willing adults, and revises this round’s headline recommendation (friction removal alone is necessary but not sufficient — see the UK CASS finding).
What actually makes average adults act on effective financial strategies — the behavioral solution set for the knowing-doing gap in personal finance — and what that implies for myBetterRates’ (MBR) unique selling proposition.
Catalog a floor of 40 distinct, evidence-backed intervention strategies drawn from:
- behavioral / household finance field experiments
- government nudge units (BIT UK, US OES/SBST, Impact Canada / BIU)
- practitioner labs (ideas42, Common Cents Lab, Irrational Labs, Morningstar behavioral team, Duke CAS)
- transfer domains: medication adherence, diet, exercise, smoking cessation, vaccination uptake — each with its explicit finance analogue
For each strategy: mechanism · key citation(s) + year · measured effect size and the outcome measured · evidence tier · who must deploy it · MBR applicability.
Parameters
Section titled “Parameters”Geography
- Global (with Canada-specific context where it matters: TFSA/RRSP, CDIC, Canadian bank switching friction, open banking status)
Depth
- Deep
Goal
- Understand the landscape
- Identify opportunities
- Decision support
Search Strategy
- Scout first
Sources
- Cited in report
AI Sources
- Claude
- Gemini
Output
- Append to task file
Additional Context
Section titled “Additional Context”Hard methodology constraints
Section titled “Hard methodology constraints”- Evidence tiering is mandatory. Tag every strategy
[Field-replicated at scale]/[Single RCT]/[Lab-only or correlational]. Verify and apply DellaVigna & Linos (2022, Econometrica) — the nudge-unit mega-study reporting real-world effects roughly an order of magnitude below the published academic literature (~1.4pp vs ~8pp), attributed to publication bias. Apply a replication discount in ranking; never rank on headline paper numbers. - Three-axis scoring, not two: (1) effect size after replication discount; (2) MBR build cost + user friction; (3) deployability — can MBR, a consumer-side third party with no custody of user accounts and no employer relationship, ship this unilaterally, or does it require an institution (employer, bank, PBM, tax-filing software)?
- Two ranked lists at the end: A. MBR can ship this alone (the lowest-hanging-fruit × biggest-impact ranking) and B. Requires a partner/institution (pitch material for the David Chilton partnership and investor Jeff — do not discard).
Premises to verify, not repeat
Section titled “Premises to verify, not repeat”- “Save More Tomorrow earned a Nobel prize.” Verify attribution precisely (Thaler & Benartzi SMarT, JPE ~2004; Thaler’s 2017 Nobel was for behavioral economics broadly). This may feed marketing copy — a wrong Nobel claim is publicly falsifiable.
- “Behavioral finance is mostly academic, not practical.” Test it. The practitioner layer exists. Investigate the sharper hypothesis: the practitioner layer is concentrated inside institutions acting on their own captive customers/employees; the unoccupied slot is consumer-side, cross-institution, agent-of-the-user — exactly MBR’s position.
Specific items to chase
Section titled “Specific items to chase”- Fernandes, Lynch & Netemeyer (2014, Management Science) — verify actual numbers (variance in financial behavior explained by financial literacy interventions; decay over time; just-in-time education). If it says what it appears to, it is the strongest academic justification for MBR’s model: don’t educate — engineer, and intervene at the moment of decision.
- BIT’s EAST (Easy, Attractive, Social, Timely) and MINDSPACE cross-walked against MBR’s F.A.S.T. (Fast, Adaptive, Simple, Tailored). Where they map = external validation. Where they don’t (Attractive/Social have no F.A.S.T. counterpart; Adaptive/Tailored have no EAST counterpart) = named blind spot or differentiator.
- Medication adherence: regimen simplification (fixed-dose combination pills, 90-day supply, default-to-refill) vs reminder intensity — check whether simplification reliably beats reminders, and whether reminder-only interventions shrink most under the replication discount. Direct evidence for “Fast” and 1-Click Transfer.
- Also cover: implementation intentions (Gollwitzer); commitment devices (SEED, Ashraf-Karlan-Yin); fresh-start effect (Dai/Milkman); temptation bundling; defaults/auto-enrollment (Madrian & Shea); auto-escalation; mental accounting and labeled accounts; prize-linked savings; tax-refund-moment prompts (Refund to Savings); round-up saving; present bias / hyperbolic discounting; choice overload; loss-aversion framing; social proof and peer comparison (Opower analogue); planning prompts; deadlines and expiry; hassle-factor reduction (Bettinger FAFSA); just-in-time education; dollar-denominated vs %-denominated salience; feedback and progress visualization; gamification limits; financial coaching and accountability; self-nudges; inattention and status-quo bias in savings-account switching specifically.
MBR’s named mechanisms — required cross-walk
Section titled “MBR’s named mechanisms — required cross-walk”Map each to the behavioral literature it already instantiates, whether the evidence supports it, and where MBR is under-using a proven mechanism:
| Mechanism | What it does |
|---|---|
| Hassle Threshold Engine | Only alerts when personalized net benefit exceeds that user’s friction threshold; most users < 3 alerts/yr |
| 1-Click Transfer | MBR does the paperwork, DocuSign, reimburses closing fees |
| Dollar-denominated net-benefit math | ”Your $80,000 earns $1,880 more/yr” — not “EQ Bank is 2.35%“ |
| Guaranteed Benefit | 100% money-back if no alert triggers |
| Anti-Pitch / Honesty as Marketing | MBR cancels its own commissioned savings transfer, tells the user to pay the 20% Visa instead, then asks for 2 referrals |
| Net Worth Thermometer | Cumulative dollar value MBR has generated for you |
| Trojan Horse | Sell instant gratification (“better rates”), expose users to deeper wealth mechanics |
| Alert-fatigue discipline | Fewer, higher-quality signals |
| Target user | Busy Canadian professional who knows they should optimize but hasn’t, because friction is too high |
Deliverable constraints
Section titled “Deliverable constraints”- Do NOT edit
MBR/Strategy/*.md. End with a proposed USP revision inside this research doc for Talbot’s decision. - Include a short Research design section stating the plan actually executed.
- Tone: skeptical, quantitative, no hype. Flag weak evidence as weak. Strategic-partner voice.
Research Report
Section titled “Research Report”Generated: 2026-08-13 · Claude Opus 5 · 6 parallel deep-dive specialists + Gemini 3.5 Flash cross-check
Research Design (as executed)
Section titled “Research Design (as executed)”| Stage | What ran |
|---|---|
| Scout | 5 parallel searches → 6 research angles |
| Deep dive | 6 specialists (Opus), parallel, ~260 tool calls total, ~90 sources assessed |
| External AI | Gemini 3.5 Flash, independent pass on the same brief (ChatGPT/Grok skipped — no API keys) |
| Synthesis | Written in-session against MBR’s strategy SSOTs (Core-Philosophy.md, Core Value Propositions.md, Key Success Factors.md, Unique Selling Proposition.md) |
Angles: (1) evidence-graded intervention catalog + replication discount; (2) does financial education change behaviour; (3) health-domain transfer; (4) alert economics; (5) is the consumer-side slot vacant; (6) Canadian switching inertia + EAST/MINDSPACE cross-walk.
Source discipline carried through from the specialists. Every figure below is labelled [Primary] (peer-reviewed/official full text fetched this session), [Secondary], or [Partial] (abstract/snippet only). Partial figures are not laundered into clean assertions — several load-bearing numbers are Partial and are marked inline. Nothing is cited from model memory.
Executive Summary
Section titled “Executive Summary”Confidence: High (86)
The research supports MBR’s diagnosis and inverts the order of its three differentiators. The knowing-doing gap in personal finance is real, large, and Canadian-specific — but across every field experiment that separated telling people the dollar number from doing the work for them, the information arm lost, usually badly. The single most decision-relevant finding: Adams, Hunt, Palmer & Zaliauskas (JFE 2021), 124,000 UK savings-account holders across five depositories, run with the FCA, tested disclosures that personalised the annual dollar gain from switching and flagged an impending rate decrease — and could not reject zero effect, irrespective of design. That is MBR’s exact message, in MBR’s exact product category, at scale. In the same experiments, a pre-filled “just sign here” return switching form moved switching from 3% to 12%, and well-timed reminders around a rate event moved it 4–5pp. [Primary]
The ordering that falls out is: friction removal >> event timing >> information content. MBR’s USP currently leads with Personalized Net-Benefit Math (differentiator #1) and closes with 1-Click Transfer (#3). The evidence says reverse them. 1-Click Transfer is the only intervention in this entire catalog that combines a large measured effect with no custody requirement — it is the Bettinger H&R Block mechanism (FAFSA filing +15.7pp on a 39.9% base; college enrolment 34% → 42%), and in that same study the information-only arm had no effect and was dropped from the analysis. [Primary]
Three further findings change the plan. (a) Published behavioural effect sizes are ~4× inflated: DellaVigna & Linos (2022, Econometrica; 126 trials, 23M participants) measure 8.7pp in academic journals vs 1.4pp in nudge units, with statistical power explaining the entire gap; plan MBR’s economics on 1–3pp absolute / 5–15% relative for anything communication-only. [Primary] (b) The sharpest untapped Canadian opportunity is not the static rate spread but teaser-rate decay — three of the five top HISA rates today are 3–5 month promos that revert silently, and nobody in the market reminds anyone. (c) MBR’s structural twin already existed and died: Flipper (UK, no custody, £30/yr, fee charged only when it found ≥£50 of savings, zero supplier commission) closed September 2021 — of spread collapse, not funding model. The commission-funded competitor died at the same time of the same cause. [Partial]
Recommendation: reposition the USP around execution, keep the Hassle Threshold Engine as quality discipline but drop the untested “<3 alerts/year” claim, gate on P(acts) × dollars rather than dollars alone, and run the spread-durability test before anything else.
AI Source Analysis
Section titled “AI Source Analysis”- Agreement on the Nobel correction. Gemini independently reached the same verdict as the catalog specialist: the Prize goes to people, not programs. Both flag “Nobel-Prize-winning Save More Tomorrow” as unusable copy.
- Agreement on the pre-filled-paperwork conclusion. Gemini’s closing rule — “never show a user a rate comparison without providing an immediate, pre-filled option to act on it” — is the same conclusion the Claude specialists reached from Bettinger, Adams et al., and the FCA trials. Convergence from independent evidence paths.
- Agreement on the F.A.S.T. blind spots, with one addition Claude’s specialists under-weighted: Gemini names Messenger (MINDSPACE “M”) as the top blind spot for a non-custodial player specifically — who delivers the recommendation, not just how. Given the FCA finding that savings customers were “suspicious or confused about why their provider would send them information about better rates elsewhere,” this is a real and cheap fix. Adopted into the cross-walk below.
- Divergence — Gemini maps Fast → Defaults. Incorrect for MBR: it has no custody and cannot set a default on anyone’s account, and Phase 1 open banking is read-only. Claude’s Canada specialist established Defaults as a structural unavailability, not a design gap. Claude’s web-retrieved finding is favoured.
- Divergence — Gemini endorses the Hassle Threshold Engine uncritically, reproducing the suppression logic as a blueprint without encountering Adams et al.’s null on personalised dollar disclosure. Claude’s alert-economics specialist retrieved that paper in full text. Claude’s finding is favoured; Gemini’s blueprint would have MBR optimise the weakest lever.
Landscape Overview — the replication discount
Section titled “Landscape Overview — the replication discount”Confidence: High (95)
Everything downstream depends on not ranking by headline numbers.
DellaVigna & Linos (2022, Econometrica) [1] assembled every trial from two of the largest US nudge units — 126 trials, 243 nudge treatments, >23 million participants — against a matched sample of 26 published academic RCTs (74 treatments, 505,337 participants):
| Academic journals | Nudge units | |
|---|---|---|
| Average impact | 8.7pp (s.e. 2.5) | 1.4pp (s.e. 0.3) |
| Relative to control | +33.5% (base 26.0pp) | +8.1% (base 17.2pp) |
| Median treatment-arm n | 484 | 10,006 |
| Median MDE (80% power) | 6.30pp | 0.78pp |
Controlling for statistical power explains the entire gap — well-powered academic nudges land at ~1pp. Publication-bias correction estimates non-significant papers are published with probability γ̂ = 0.10, shrinking the academic average from 8.7pp to 3.16pp. Notably, practitioners forecast the true effect accurately (median 1.95pp) while academics forecast 6pp — the over-optimism is specific to the research community. [Primary]
Corroboration is harsher. Maier et al. (2022, PNAS) [2] re-analysed the largest nudge meta-analysis (pooled d = 0.43) with robust Bayesian meta-analysis: corrected estimate d = 0.04 [0.00, 0.14], with evidence against an effect specifically in the “information” and “assistance” messaging categories. Bakdash & Marusich (2022, PNAS) [3] show the effect distribution is sharply left-truncated at zero and name finance as one of four domains showing that pattern most clearly. Maier et al. further report finance is the one domain showing no evidence of heterogeneity — no “but some finance nudges still work” escape hatch. [Primary] This is the strongest single caution in this report for a consumer-finance company.
Megastudies give the realistic at-scale distribution. Milkman et al. (2021, Nature), n=61,293 gym members, 53 conditions from 30 scientists: 45% beat placebo, by 9–27% relative; best condition +0.40 weekly visits (+27%). Only 8% produced significant effects beyond the four-week window, and the authors warn top-performer estimates are inflated by selection-on-the-maximum. [Primary] [4]
And the adoption layer: DellaVigna, Kim & Linos (2022) [6] followed 67 US city departments running 73 RCTs. Average effect 1.9pp (+13% on a 15% base) — but departments adopted the winning treatment in only 27% of cases, best predicted not by evidence strength but by whether the trial used pre-existing communications (organisational inertia). A winning A/B test does not become a running product by itself. [Primary]
The discount rule this report applies
Section titled “The discount rule this report applies”| Rule | Statement |
|---|---|
| 0 | Discount in relative terms, then re-express in points. Convert headline → % lift over control → haircut → multiply by MBR’s own baseline. Never import someone else’s pp into a revenue model. |
| 1 | 4× haircut on any academic-journal headline (33.5% → 8.1% is the empirical ratio). |
| 2 | No haircut if already well-powered (treatment arm n ≳ 10,000, MDE ≲ 1pp). |
| 3 | No haircut on nudge-unit or megastudy estimates — already at-scale. |
| 4 | Halve again for anything claimed beyond the intervention window (only 8% of gym interventions persisted). |
| 5 | Treat as zero when the only evidence is one underpowered academic RCT — especially in finance, especially for information/assistance messaging. |
| 6 | Discount for the adoption gap (27%) — unless you own the deploying stack, which MBR does. This one is MBR’s advantage. |
Planning number for MBR: 1–3pp absolute, 5–15% relative, for any communication-only intervention on an existing contactable base. Not the 25–30% figures that populate behavioural-economics keynotes.
How It Works — the axis that actually discriminates
Section titled “How It Works — the axis that actually discriminates”Confidence: High (92)
The brief hypothesised that regimen simplification beats reminder intensity. The evidence does not support that framing — and what replaced it is stronger.
The contradiction that forced the reframe: reminder devices were flatly null at scale — the REMIND trial (Choudhry et al., n=53,480 non-adherent patients) found standard pillbox OR 1.03, digital timer cap OR 1.00, toggle strip OR 0.94, concluding “low-cost reminder devices did not improve adherence” [Primary] [7]. Yet the largest single effect in Kini & Ho’s JAMA review of adherence RCTs was telephone calls prompting refills, 33% absolute — a reminder [Partial] [8]. Those are only reconcilable on a different axis:
Does the intervention complete or remove a step in the action chain, or does it only raise awareness that the step exists?
A toggle cap raises awareness → null. A call that triggers the refill completes a step → largest effect. A polypill deletes a step → RR 1.38. A text saying a shot is “waiting for you” converts search-and-decide into collect → +2.9pp.
Six independent settings converge on it:
| Setting | Awareness-raising arm | Step-completing arm |
|---|---|---|
FAFSA / H&R Block (QJE) [9] [Primary] | Personalised aid estimates vs local college costs: no effect, arm dropped | Pre-filled + submitted on the spot: filing +15.7pp (base 39.9%); enrolment 34%→42%; persistence +8pp at 3 yrs |
FCA cash savings, 124k depositors [10] [Primary] | Switching box +3% internal, 0% external; naming competitors’ rates went negative; personalised £-gain + urgency arms null | Return switching form 3% → 12%; timed reminder +4–5pp |
Keys, Pope & Pope refi [11] [Primary] | Letter arm stating lifetime and monthly savings: no difference across arms | Pre-approved, zero-up-front-cost offer: 13.0–24.3% take-up |
EITC take-up (Bhargava & Manoli, AER) [12] [Primary] | Stigma-reduction language: 0.23 → 0.22 (null) | Simplified notice: 0.14 → 0.23; benefit salience 0.23 → 0.28 |
CARD Act 36-month disclosure [13] [Primary] | +0.5pp on a 5.7% base, 160M accounts; upper-bound $71M/yr on a $744B market | Fee rules in the same Act: ~$12.6bn/yr |
| Medication adherence [7][8] | Reminder devices: null at n=53,480 | Refill-triggering call: 33% [Partial]; polypill RR 1.38 [Primary] |
This is the report’s spine, and it is a sharper version of F.A.S.T.’s “Fast” than the original. It says a well-timed message that lands the user one tap from a completed transfer is not “reminder intensity” — it is friction removal delivered by message. A rate-alert badge is on the wrong side of the line; 1-Click Transfer is on the right side.
One hard boundary. In a 3.66-million-person CVS megastudy, text reminders raised booster uptake +1.05pp (+21% relative), but offering free round-trip Lyft rides added nothing over texts alone — the authors conclude “attentional barriers rather than transportation access” bind in already-vaccinated populations [Primary] [14]. Removing friction only works when friction is the binding constraint. MBR must diagnose which constraint binds per user — attention or steps — because the two call for opposite spends.
The Master Catalog — 50 strategies
Section titled “The Master Catalog — 50 strategies”Evidence tier: [FR] field-replicated at scale · [1RCT] single RCT · [LAB] lab-only/correlational · [NULL] measured null or counter-example · [UNEV] no causal estimate retrieved. “MBR alone?” assumes no custody of accounts, no employer relationship, Phase 1 read-only open banking.
A. Custody- or institution-required (structurally unavailable to MBR alone)
Section titled “A. Custody- or institution-required (structurally unavailable to MBR alone)”| # | Strategy | Mechanism | Citation (yr) | Measured effect (outcome) | Tier | Deployer | MBR alone? |
|---|---|---|---|---|---|---|---|
| 1 | Automatic enrolment | Inertia works for you | Madrian & Shea (2001) [15] | Participation 37.4% → 85.9% (401k) | [FR] | Employer + recordkeeper | No |
| 2 | Default level anchoring | People stay at the number you set | Madrian & Shea (2001) [15] | 76% contributed exactly the 3% default | [FR] | Employer | No |
| 3 | Save More Tomorrow | Pre-commit future raises; present bias sidestepped | Thaler & Benartzi (2004, JPE) [16] | 78% joined; saving 3.5% → 11.6% over 28 months; 80% stayed through 3 raises | [1RCT] non-randomised field | Employer + payroll | No |
| 4 | Auto-escalation | Increases need no repeat decision | Component of [16]; Beshears et al. [17] | Embedded in the 3.5→11.6% path; 12%-default variant 25% opt-out at 12mo | [1RCT] [Partial] | Employer | No |
| 5 | Quick/simplified enrolment | Collapse multi-dimensional choice to one bit | Beshears et al. [17] | Enrolment +10 to +20pp | [1RCT] [Partial] | Employer | No |
| 6 | Hard commitment device (locked account) | Illiquidity as self-binding | Ashraf, Karlan & Yin (2006) [18] | Savings +81% at 12mo; 28% take-up | [1RCT] | Bank with custody | No — can refer only |
| 7 | Mental accounting / earmark + partition | Labelled buckets create self-control | Soman & Cheema (2011) [19] | Two envelopes vs one: +72% saved (₹414 vs ₹241) | [1RCT] [Partial] | Bank/consumer | No — can prompt |
| 8 | Tax-refund-moment savings prompt | Windfall + pre-filled anchor on the filing screen | Refund to Savings [20] | Deposit rate +42% to +64% relative; ~900k filers | [FR] [Partial] | Tax software | No |
| 9 | Prize-linked savings | Lottery utility substitutes for gambling | Filiz-Ozbay et al.; Kenya RCT [21] | Lab: beats equal-EV interest. Kenya RCT: 37.3% vs 27.2%, not significant | [LAB] + one null | Bank/regulator | No |
| 10 | Round-up / spare change | Micro-amounts below the pain threshold | BoA “Keep the Change”; Acorns [22] | No RCT located. Adoption only: 2M customers yr 1 | [UNEV] | Custodian | No |
| 11 | Opt-out appointment default | Assign rather than invite | Italian RCT, n=2,000 [23] | Uptake +3.2pp (32% relative) | [1RCT] [Partial] | Whoever owns the flow | Partly — own flow only |
| 12 | Fixed-dose combination (bundle N actions into one) | Deletes steps rather than reminding about them | Polypill meta-analysis, 6 RCTs n=13,139 [24] | Adherent 76.5% vs 58.6%, RR 1.38 (1.22–1.56); CV events RR 0.72 | [FR] (self-report caveat) | Product issuer | Partly — see #13 |
B. Third-party shippable — MBR’s actual surface
Section titled “B. Third-party shippable — MBR’s actual surface”| # | Strategy | Mechanism | Citation (yr) | Measured effect (outcome) | Tier | MBR alone? |
|---|---|---|---|---|---|---|
| 13 | Pre-filled paperwork / application assistance | Eliminates the physical act, not the knowledge gap | Bettinger et al. (2012, QJE) [9] | FAFSA filing +15.7pp on 39.9% base; enrolment 34%→42%; +8pp persistence at 3 yrs | [FR] | Yes — the standout |
| 14 | ”Just sign here” return switching form | Same, in MBR’s exact category | Adams et al. (2021, JFE) [10] | Switching 3% → 12% (4×), n=124,000 | [FR] | Yes |
| 15 | Fee reimbursement / removing the exit toll | Removes a real cost, not a perceived one | Analogue of [9][11] pre-approved zero-cost offers | Take-up 13.0–24.3% on zero-up-front-cost offers | [1RCT] | Yes |
| 16 | Reminder at a rate event | Restores attention at the moment it pays | Byrne et al. (2023) [25] | Refinancing 8.9% → 15.7% (+76%), n=12,000 — but sent by the incumbent lender | [1RCT] at scale | Partly — mechanism yes, standing no |
| 17 | Recurring savings reminder | Top-of-mind | Karlan et al. (2016, Mgmt Sci) [26] | +3.2pp meeting commitment (+5.4%); +5.9–6.1% amount saved (p=0.10–0.11); n=14,167, 3 countries | [FR] | Yes |
| 18 | Goal-specific reminder content | Names the goal, not just the product | Karlan et al. (2016) [26] | Goal-mentioning reminders ~2× more effective; gain/loss framing n.s. | [FR] | Yes |
| 19 | Implementation intention (date and time) | Converts intention into a cued plan | Milkman et al. (2011, PNAS) [27] | Date+time +4.2pp on 33.1% base; date only +1.5pp, n.s. | [1RCT] | Yes |
| 20 | Default-plan message (names time & place) | Message doing structural work | CVS megastudy, n=3.66M [14] | +1.20pp, top of 8 conditions | [FR] | Yes |
| 21 | Pre-arrangement framing (“reserved for you”) | Converts search-and-decide into collect | Milkman megastudies [28][29] | “waiting for you” +2.9pp (n=689,693); “reserved for your appointment” +4.6pp (n=47,306) | [FR] | Yes |
| 22 | Short deadline (1 week) | Dates an open-ended task | Screening RCT, n=7,711 [30] | 1-week +2.5pp (4.7%→9.7%); beat 3-week deadline (p=0.01); statistically indistinguishable from paying people | [1RCT] | Yes |
| 23 | Simplification of the notice | Fewer words, one side, clear font | Bhargava & Manoli (2015, AER) [12] | Response 0.14 → 0.23, n=35,050 | [1RCT] at scale | Yes |
| 24 | Benefit salience (show the amount) | Concrete beats abstract — within a simplified notice | Bhargava & Manoli (2015) [12] | 0.23 → 0.28 | [1RCT] at scale | Yes |
| 25 | Fresh start / temporal landmark timing | New mental accounting period buries past failure | Dai, Riis & Milkman (2014) [31] | Gym visits +33.4% new week, +14.4% new month, +7.5% post-birthday | [LAB]/archival | Yes |
| 26 | Prominence / placement of the ask | Position determines whether it is read at all | Adams et al. (2021) [10] | Front page 3% → 6%; identical content on the back page: no effect | [FR] | Yes |
| 27 | Loss framing | Losses loom larger | Field-framing studies [32] | Loss-framed incentive +4.7pp (17% relative on 27% base) | [1RCT] [Partial] | Yes |
| 28 | Dollar- vs percentage-denomination | Which unit makes a price salient | Larrick & Soll (2008); energy-label studies [33] | Direction favours dollars; magnitude at MBR’s decision does not follow — see #36 | [LAB] [Partial] | Yes — but test |
| 29 | Target on movability, not on dollars | Treatment effect ≠ benefit size | Financial-aid targeting RCT, n>53,000 [34] | Untargeted +6.4pp on 37% base; treating half the population retained ~75% of benefit. Targeting low-baseline is detrimental; intermediate-baseline is best | [1RCT] [Partial] | Yes |
| 30 | Financial coaching / accountability | Repeated human accountability | Theodos et al. (2015) [35] | Site-dependent; credit score up to +20 pts, inconsistent across two sites | [1RCT] inconsistent [Partial] | Yes — labour-intensive |
| 31 | Choice simplification / fewer options | Avoids choice-overload paralysis | Iyengar & Kamenica (2010) [36] | Per 10 extra funds: participation −2%, equity allocation −3.28% | [LAB]/correlational [Partial] | Yes (own UI) |
| 32 | Run a megastudy, not a campaign | Expert forecasters cannot pick winners | Milkman et al. (2021) [4] | 45% of 53 conditions beat placebo; forecasters (professors, practitioners, laypeople) failed to predict which | [FR] | Yes |
| 33 | Non-interruptive presence layer | Keeps value visible without spending interrupt budget | Phansalkar et al. (clinical CDS) [37] | 33 low-priority alert classes = ~36% of displayed alerts, >90% override; recommendation is to make them non-interruptive, not delete them | [FR] | Yes |
| 34 | Messenger identity | Who says it, not just how | MINDSPACE [38]; FCA suspicion finding [39] | No effect size retrieved; FCA found savers “suspicious or confused” about why a provider would flag better rates elsewhere | [UNEV] | Yes |
| 35 | Descriptive social norms (in-cohort only) | Most people like you did this | EAST revised [40] | Direction established; national norm would backfire — only 12–18% of Canadians switch | [FR] w/ caveat | Yes, carefully |
C. Nulls and counter-examples — what NOT to build
Section titled “C. Nulls and counter-examples — what NOT to build”| # | Strategy | What was measured | Tier |
|---|---|---|---|
| 36 | Personalised dollar gain + urgency disclosure, at scale | Adams et al. Trial 2, n=124,000: cannot reject zero, irrespective of design [10] | [NULL] [Primary] |
| 37 | Naming competitors’ rates in a communication | Small reduction in internal switching (FCA) [39] | [NULL]/negative |
| 38 | Information only (personalised estimates, no assistance) | FAFSA arm: no effect, dropped from analysis [9] | [NULL] |
| 39 | Stigma-reduction / reassurance messaging | EITC: 0.23 → 0.22 [12] | [NULL] |
| 40 | Reminder devices with no action attached | REMIND, n=53,480: OR 1.03 / 1.00 / 0.94 [7] | [NULL] |
| 41 | Removing a non-binding friction | Free Lyft rides added nothing over texts, n=3.66M [14] | [NULL] |
| 42 | Financial education delivered in advance | Fernandes et al.: r=.032, r²=.0011 — 0.1% of variance; no significant effect of even 24 hrs instruction at ≥18.5 months’ delay [41] — contested, see below | [FR] meta-analysis |
| 43 | Gamification | One trial: no significant difference in weekly saving over 4 weeks; gamified investing platforms increased risk-taking [42] | [LAB]/mixed [Partial] |
| 44 | Progress visualisation / goal gradient | No field effect size on a financial outcome retrievable this session — loyalty-card evidence only [43] | [UNEV] |
| 45 | Money-back guarantee in financial services | No measured evidence retrieved [44] | [UNEV] |
D. Health-domain transfers (finance analogue stated)
Section titled “D. Health-domain transfers (finance analogue stated)”| # | Health intervention | Effect (outcome) | Tier | Finance analogue | MBR alone? |
|---|---|---|---|---|---|
| 46 | Regimen simplification (dose frequency) | ARV meta-analysis, 19 RCTs n=6,312: +2.55pp adherence (CI 1.23–3.87); no gain in virological suppression. Broader lit OR 3.07 [Partial] [45] | [FR] | Fewer screens/forms/fields per switch | Yes — expect single-digit pp |
| 47 | 90-day supply / auto-refill | MPR +3.0 pts (30-day fills), +1.4 pts (90-day) [46] | [Partial] | Auto-renewal, longer review cycles | Partly — can auto-recheck, not auto-execute |
| 48 | Financial incentives for adherence | Effective only while paid; post-intervention convergence (71% vs 77%, n.s.) [47] | [1RCT] [Partial] | Switching bonus / cash-back | Yes but — buys the act, not the habit |
| 49 | Deposit contract vs reward | Halpern (NEJM 2015, n=2,538): deposits 13.2pp more effective among acceptors — but acceptance 13.7% vs 90.0%; ~half of 6-month quitters relapsed by 12 months [48] | [1RCT] | Self-binding savings commitment | Yes — plan for ~10% take-up, never the main funnel |
| 50 | Commitment savings account for a health goal (CARES) | Giné, Karlan & Zinman: 11% take-up; +3pp passing 6-month test, persisting at 12 months [49] | [1RCT] [Partial] | Literally a financial product used behaviourally | No — needs custody |
What does NOT transfer from health
Section titled “What does NOT transfer from health”- Frequency, and therefore habit. A pill is daily; a savings switch is once every few years. Every mechanism depending on repetition — streaks, per-occurrence microincentives, the “return after a missed workout” bonus that won StepUp — has no substrate in MBR’s domain. MBR should not build a habit product.
- In-period lift ≠ durable behaviour. Only 8% of 53 gym interventions had measurable effects after four weeks [4]; half of Halpern’s quitters relapsed [48]. Less damaging for MBR — a completed transfer is a one-time state change that persists — but it kills any strategy premised on sustained engagement.
- The proximate metric can move while the ultimate outcome doesn’t. Once-daily dosing raised adherence with no virological improvement [45]; a statin incentive trial raised adherence and left LDL-C unchanged. Finance analogue is sharp: instrument realised dollars, not switch counts. A switch into a teaser rate that reverts is a moved metric and an unmoved outcome.
- Who holds the default. Every default result here was deployable because the intervener controlled the choice architecture. MBR does not.
- Consequence salience. Missing a pill has a named clinical outcome and a prescriber who asks. An unswitched savings account has an invisible opportunity cost — and the labelling evidence says merely showing the number probably won’t manufacture the stake [12][10].
- Scale shrinks everything. The three vaccination megastudies moved 1.05–2.1pp on average. Small-trial ORs come from small, selected, self-reported studies.
Premise Verification
Section titled “Premise Verification”Confidence: High (93)
1. “Save More Tomorrow earned a Nobel prize” — FALSE as stated
Section titled “1. “Save More Tomorrow earned a Nobel prize” — FALSE as stated”From the official Royal Swedish Academy press release [50] [Primary]: the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2017 was awarded to Richard H. Thaler alone, “for his contributions to behavioural economics” — citing mental accounting, limited rationality, social preferences and lack of self-control as a body of work. The Prize is awarded to people, not to programs or papers. Shlomo Benartzi, SMarT’s co-author, is not a laureate.
Usable copy: “…developed by Richard Thaler, who received the 2017 Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel for his contributions to behavioural economics.” Do not write: “Nobel-Prize-winning Save More Tomorrow” · “Nobel Prize in Economics” · anything implying Benartzi is a laureate.
(Gemini independently reached the same verdict and adds that the Academy’s scientific-background document cites SMarT as a flagship application — not verified this session; verify before using.)
2. “Behavioral finance is mostly academic, not practical” — FALSE in the plain version, TRUE in the sharpened one
Section titled “2. “Behavioral finance is mostly academic, not practical” — FALSE in the plain version, TRUE in the sharpened one”There is a large, well-funded, RCT-running practitioner industry. But every organisation whose own primary text was retrievable states it delivers through institutions to those institutions’ populations:
| Org | Who pays | Whose behaviour changes | Consumer’s agent? |
|---|---|---|---|
| ideas42 | Mission-driven orgs, govt, health, education [51] | The client’s beneficiaries | No |
| Irrational Labs | Financial institutions and fintechs — Credit Karma, Steady, B3 [52] | The client’s own users | No, explicitly |
| Common Cents Lab (Duke) | Fintechs, credit unions, banks, nonprofits [53] [Partial] | Each partner’s own members | No |
| BEworks (Toronto) | CIBC, TD Insurance, Canadian Tire Financial, RBC [54] [Partial] | Those banks’ own customers | No |
| BIT / Nesta | Governments, local authorities, businesses [55] [Partial] | Citizens of the commissioning body | No |
| FCAC | Self-funded regulator [56] [Partial] | Canadians directly — but non-commercial | Closest, not commercial |
This is structural, not accidental: the intervention surface (the product flow, the statement, the default) is owned by the institution. The sharpened hypothesis holds. The genuine white space is an intersection, not a slot — behavioural practitioners are always institution-side; consumer-side rate products (Ratehub, Raisin, Snoop, highinterestsavings.ca) are mostly deal-listing without deliberate behavioural design. Nothing retrieved occupies the overlap.
3. “Don’t educate, engineer” — HALF SUPPORTED, and the half MBR leans on is the contested half
Section titled “3. “Don’t educate, engineer” — HALF SUPPORTED, and the half MBR leans on is the contested half”Fernandes, Lynch & Netemeyer (2014) [41] [Primary] is real and precise: 168 papers / 201 studies; across 90 manipulated effect sizes, mean weighted r = .032, r² = .0011 — 0.1% of variance in downstream financial behaviour. Decay: “at delays of 18.5 months or greater, there is no significant effect of even 24 hours of instruction”; “brief interventions at short delays have effects equal to more intensive interventions at long delays.” Low-income samples fare worse (r = .025 vs .035, p = .02).
But the counterweight is real and is not a footnote. Kaiser, Lusardi, Menkhoff & Urban (2022, JFE) [57] [Primary], 76 RCTs, N>160,000: under the same common-true-effect assumption Fernandes used, the pooled behaviour effect is 0.065 SD (CI 0.043–0.089) vs 0.018 SD (CI −0.004 to 0.022) — ~3.6× larger and excluding zero; under their preferred RVE model 0.1003 SD (~5.5×). They state plainly that “one of the main findings of Fernandes et al. (2014) is not confirmed,” document four data-coding errors in the original’s RCT classification, and argue the “0.1% of variance” framing “creates the illusion of miniscule effects.”
They do not overturn everything: behaviour effects stay ~half the size of knowledge effects, credit/debt behaviours are least robust, and on durability they conclude one can “neither rule out sustained and relatively large effects nor close to zero effects” — i.e. long-run persistence is unproven in both camps.
Just-in-time delivery is quantified only once: Kaiser & Menkhoff’s teachable-moment dummy, +0.079 SD on behaviour (predicted g = 0.124, ~48% above the unconditional average), against mandatory formats at −0.074 SD (predicted g = 0.030, n.s.) [58] [Primary]. Caveat stated by the authors: in the RCT-only subsample the teachable-moment coefficient “loses explanatory power.” No randomized just-in-time-vs-advance comparison was located this session.
And the just-in-time idea originates with the sceptics. Fernandes et al. themselves wrote that their findings “militate toward ‘just in time’ financial education,” that knowledge “may be better conveyed via ‘just-in-time’ financial education tied to a particular decision,” and — in language that maps directly onto MBR — that just-in-time education “may have promise, alone and embedded in decision support systems that help identify a tailored consideration set of safe options,” naming choosing a mortgage as a decision where it “may find a receptive audience.” Their operational framing of the problem: “how to reach consumers at a point in time close to their decision when they are impatient for closure” [41].
Pitch-deck-safe sentence: “Financial education delivered in advance produces small and weakly durable changes in behaviour — a meta-analysis of 76 RCTs puts the average effect at about 0.10 SD, roughly half its effect on knowledge, with no established long-run persistence. Across field experiments, what moved money was intervention at the decision point that removed work, not information that improved understanding.”
Sentence to avoid: any claim that consumers fail to act “because they don’t know the numbers.” A well-read reader answers with Bettinger’s information-only arm and the FCA switching-box trials.
The Canadian Opportunity
Section titled “The Canadian Opportunity”Confidence: High (88)
Mortgage inertia — attack the 37%, not the 20%. FCAC’s Monthly Financial Well-Being Monitor data story (published 2 March 2026) [59] [Primary]: 48% personally compared lenders; 36% had someone compare for them; 20% never compared; 37% chose their lender primarily because they already bank there; 13% did not know negotiating was an option. Meanwhile 80% say comparing is important. (Note: the three comparison figures sum to 104% — not a mutually exclusive partition; do not present as a segmentation. This is origination, not renewal.)
The largest named cause of inertia is not ignorance of rates — it is a relationship default, which rate information does not move. The Competition Bureau (21 Mar 2024) [60] [Primary] puts a price on it: only 12.1% of mortgage renewers switch, and incumbent-renewers pay 6.1 basis points more than new borrowers. Big Six hold ~73% of outstanding mortgages. Show that 6.1bp as dollars over the term, not as basis points.
Consideration-to-action gap, measured by a regulator. Competition Bureau’s Your Data, Your Control (15 Jan 2026) [61] [Primary], with a nationally representative survey of 3,046 Canadians plus experiments: only 18% switched providers within three years while 25% considered switching annually — a ~7:1 gap. Stated reasons read as a spec for MBR: brand loyalty, “not knowing how to switch,” “unsure about the benefits,” “believing there are hassles related to the process.” (Insurance, not deposits — but it is the best-evidenced Canadian measurement of the exact mechanism.)
The idle pool and the spread, primary-sourced today. Bank of Canada chartered bank selected liabilities, obs. May 2026: personal chequable/demand deposits C$498.1bn; total personal deposits C$1,609.4bn [62] [Primary]. TD’s own posted rates page, retrieved 13 Aug 2026: TD Every Day Savings 0.010%, all other personal chequing 0.000%, ePremium Savings 0.450% [63] [Primary]. Against EQ Bank Personal Account at 2.75% standard everyday [64] [Secondary] — a documented ~274bp gap.
But the sharpest wedge is teaser decay, not the static spread. Three of the five top rates on the board are explicitly promotional and expire in 3–5 months (CIBC 4.60%/3mo, Simplii 4.60%/5mo, Tangerine 4.50%/5mo) [64]. Capturing the advertised rate requires a recurring act of attention on a date the consumer never sees, the reversion is silent, and the bank has no incentive to warn them. This is (a) unambiguously valuable, (b) impossible for the incumbent to offer, (c) recurring — converting a one-shot comparison into a retention loop, and (d) the cleanest possible demonstration of Adaptive. No comparison site addresses it.
CDIC forces the allocation problem into MBR’s hands. Coverage is $100,000 per category per member institution across nine categories (one-name, joint, RRSP, RRIF, TFSA, RDSP, RESP, FHSA, in-trust); covers deposits and GICs, not mutual funds/stocks/bonds/ETFs [65] [Primary]. For a household with real cash, “move it all to the best rate” is wrong advice — the per-institution limit forces multi-institution allocation. Nine categories × multiple institutions is a genuine optimisation problem no single bank will ever compute.
Transfer-out fees are the hard friction — and are unverified. Reported: TD raising RRSP/TFSA transfer-out $75→$150 (1 Jul 2025), RBC $50→$150, CIBC $100, Tangerine $45→$125, Wealthsimple free. All [Unverified] — snippet-level only; the RBC fee schedule PDF returned 404. Directionally material: $50–$150 per account is 0.5–1.5% of a $10,000 TFSA — enough to make a switch genuinely marginal and more than enough to serve as a rationalisation. Pull each institution’s own fee disclosure before using any figure. Whether Canadian registered transfers are electronic or paper, and the T2033 process detail, could not be verified.
Regulatory — and the answer is favourable
Section titled “Regulatory — and the answer is favourable”Consumer-Driven Banking Act: Royal Assent as part of Bill C-15 on 26 March 2026; proposed regulations pre-published in Canada Gazette Part I on 27 June 2026, 60-day consultation [66] [Primary]. From the Gazette text:
- In scope: deposit, payment, investment (registered and non-registered), and lending accounts — all three MBR target behaviours.
- Excluded: “derived data” — banks’ own scoring and segmentation stay behind the wall.
- Mandated participants: “specified large banks based on a threshold for retail volume.” The numeric threshold could not be retrieved — so whether EQ, Wealthsimple or Simplii are in scope at launch is unresolved, and it matters: MBR’s destinations being out of scope would blunt the framework’s usefulness.
- Accreditation requires Canadian business location, insurance, integrity policies for key personnel, demonstrated security compliance. A real fixed cost and a gate, not a formality.
- Screen scraping: prohibition is in the Act but “will not be brought into force until broader consultation.” Not yet illegal — but legislated to end on an unannounced date. Do not architect on it.
- Write access / payment initiation is explicitly excluded from Phase One — “read only.” No operational date in the Gazette. (DLA Piper
[Secondary]says Phase 2 mid-2027; treat as industry expectation, not commitment [67].)
Strategic reading: this is favourable. Under the framework as drafted, no third party can execute a switch. MBR’s no-custody model is aligned with Phase 1 scope rather than limited by it — “can’t move the money” is currently the constraint on every participant. Build so the product’s value is complete under read-only; treat write access as upside.
Tailwind: the Competition Bureau has twice formally urged Ottawa to cut switching friction — adopt consumer-driven banking swiftly, and eliminate the stress test for uninsured “straight switches” at renewal, noting incumbent-renewers avoid the stress test while switchers do not, an asymmetry that directly subsidises inertia, against ~1.4M uninsured mortgages renewing in 2024–25 [60]. Whether that change was implemented could not be confirmed — verify before claiming it.
Competitive Reality — and the dead twin
Section titled “Competitive Reality — and the dead twin”Confidence: Medium-High (78)
| Player | Custody? | Who pays | Note |
|---|---|---|---|
| MaxMyInterest (US) | No — funds stay in the user’s own FDIC-insured accounts [68] [Primary] | User — 16bp/yr, $20/qtr min | Closest structural analogue to MBR. Its “Common Application” for opening online savings accounts is an existence proof that a no-custody third party can attack account-opening friction. Open item: whether Max also takes bank revenue is unresolved (Kitces review returned 403). |
| Nous (UK) | No | Hybrid — £6.99/mo + commissions rebated to members as cash [69] [Primary] | Live template for MBR’s exact revenue design. Explicitly anti-PCW: “We don’t think it’s fair or trustworthy that price comparison sites make all their money from providers.” Not yet proven at scale in what was retrieved. |
| Raisin | Yes (marketplace) | Banks | Supply-side funded — outside the slot |
| Flagstone (UK) | Yes — HSBC holding account in blind trust | Banks, up to 0.30% interest share deducted before rates are displayed [70] [Primary] | Custodial — should not be counted as occupancy |
| Ratehub / Borrowell (Canada) | No | Lenders — ~$50–$500+ per successful referral, “the bulk of revenue” [71] [Partial] | Canada’s consumer-facing layer is supply-side-funded — i.e. not the user’s agent |
The dead twin. Flipper (UK energy auto-switcher): £30/yr membership, charged only once it found ≥£50 of savings, took no commission from suppliers, switched automatically whenever ≥£50 was available. Closed 28 September 2021 [72] [Partial]. That is a near-exact structural replica of MBR’s economics. Look After My Bills — commission-funded — wound down at the same time, citing that cheap fixed deals had disappeared from the market [73] [Partial].
The discriminating fact: both died of the same cause — collapse of cross-supplier price dispersion — despite opposite funding models. The funding model is exonerated. What is indicted is dependence on a spread MBR does not control.
A caution on the fiduciary pitch. The FCA’s thematic review of 14 price-comparison websites (~90% of the GI market) found consumers mistakenly believed the PCW had assessed suitability for them [74] [Partial]. And the disclosure literature says the obvious fix backfires — disclosing a conflict can produce strategic exaggeration and moral licensing, and advisees under-discount disclosed-biased advice [75] [Partial]. No measured evidence that fiduciary/agent positioning increases consumer uptake was retrieved. Lead with measured net benefit and let the structure be the proof rather than the pitch.
MBR Mechanism Cross-Walk
Section titled “MBR Mechanism Cross-Walk”Confidence: High (85)
| MBR mechanism | Behavioural literature it instantiates | Evidence status | Verdict |
|---|---|---|---|
| Hassle Threshold Engine | Alert-fatigue / interrupt-budget discipline; targeting | Suppression: supported. Per-message acceptance falls IRR 0.70 per additional alert per encounter (1.59M alerts, 112 clinicians) [76]; endpoint is 90% override in a system firing on 13.7% of actions (11 studies, 570,776 prescriptions) [77]; redundant later reminders added zero in the only field-replicated savings experiment [26]. Dose: untested. Nothing tested 1–3/yr vs ~12/yr; Karlan’s working effect came from monthly contact. | Keep the logic, drop the “<3/year” claim. Re-gate on P(acts) × dollars, not dollars — targeting on low baseline is detrimental [34]. |
| 1-Click Transfer | Application assistance / hassle-factor reduction (Bettinger); pre-filled return form (FCA) | Strongest evidence in the report. +15.7pp filing, 34%→42% enrolment [9]; 3%→12% switching [10] | Promote to differentiator #1. This is the product. |
| Dollar-denominated net-benefit math | Benefit salience; unit framing | Weakest of the three. Salience works inside a simplified notice (0.23→0.28) [12]. But personalised £-gain + urgency at n=124,000: null [10]; CARD Act 36-mo disclosure +0.5pp on 5.7% base [13]; Keys’ explicit-savings letter arm indistinguishable [11] | Demote to table stakes. Necessary to justify the ask; insufficient to cause the act. |
| Guaranteed Benefit (100% money-back) | — | Unevidenced. No measured evidence on money-back guarantees in financial services retrieved [44] | Untested product hypothesis. It converts a retention problem into a refund liability rather than solving it. |
| Anti-Pitch / Honesty as Marketing | Fiduciary/agent positioning; conflict disclosure | Unevidenced, and the adjacent evidence points the other way. Consumers already assume comparison sites assessed suitability [74]; disclosing conflicts can backfire [75] | Untested. Keep it — it is an ethical commitment and a referral trigger — but do not model it as a conversion lever. |
| Net Worth Thermometer | Goal gradient / endowed progress / switching cost | Unevidenced in finance. Retrieved only as loyalty-card secondary summaries [43] | Untested. Best justification is different: it is the non-interruptive presence layer [37] that solves the zero-alert retention problem. |
| Trojan Horse | Consideration-set entry (Fernandes et al.: “it does not matter changing consumers’ evaluation of options if consumers are not considering those options in the first place”) [41] | Supported conceptually, no effect size | Sound. Keep. |
| Alert-fatigue discipline | See Hassle Threshold Engine | Supported as quality discipline | Keep — but make the argument with the 90%/13.7% numbers, not a vague appeal to “fatigue.” |
| Target user (busy professional, friction-blocked) | Attention-cost model | Supported. Adams et al. conclude the data fit a fixed cost of attention, not per-quote search costs [10]; FCA: 67% of non-switchers never even considered it [39] | Correct diagnosis. |
Three of nine named mechanisms are currently unevidenced (Guaranteed Benefit, Anti-Pitch, Net Worth Thermometer). That is not fatal — they may still be right — but they should be labelled as hypotheses in any investor material, and instrumented so they can be tested.
Where MBR is UNDER-using a proven mechanism
Section titled “Where MBR is UNDER-using a proven mechanism”- Implementation intentions (#19) — date and time, +4.2pp; date-only n.s. Free to build, no custody, no data access.
- Pre-arrangement framing (#21) — “reserved for you” / “waiting for you,” the winning variants across 4.4M people. MBR’s copy currently reads as comparison, not collection.
- Short deadlines (#22) — best-evidenced element available; 1 week beat 3 weeks and matched paying people. Tie to real events (promo expiry, GIC maturity, renewal date). Caution: no evidence retrieved on whether soft or fabricated deadlines damage trust — and manufactured urgency in a subscription product is a recognisable dark pattern that would poison MBR’s core asset.
- Messenger identity (#34) — Gemini’s contribution; MINDSPACE’s first lever, entirely absent from F.A.S.T.
- In-cohort social norms (#35) — with a hard constraint: a truthful national norm says most Canadians don’t switch (12–18%), which is actively counterproductive. Use “of MBR users who saw this alert, N% moved” or don’t use norms.
- Loss framing (#27) — “you are giving up $X per month,” not “you could earn $X.” Free. Untried. (Note Karlan found gain/loss framing n.s. in reminders [26] — test, don’t assume.)
EAST / MINDSPACE vs F.A.S.T.
Section titled “EAST / MINDSPACE vs F.A.S.T.”Confidence: High (90) — both BIT EAST editions and the MINDSPACE primary PDF were fetched and extracted in full.
| EAST element | Sub-principle | F.A.S.T. counterpart | Assessment |
|---|---|---|---|
| Easy | Reduce effort required | Fast | Direct map — strongest external validation MBR has |
| Easy | Simplify messages | Simple | Direct map |
| Easy | Harness the power of defaults | none | Largest gap — and structural |
| Attractive | Attract attention / frame | none (partly Tailored) | Real gap |
| Attractive | Design incentives | none | Gap |
| Social | Show most people do it | none | Gap |
| Social | Power of networks | none | Gap |
| Social | Reciprocity | none | Gap |
| Timely | Choose the right moment | Adaptive (partial) | Adaptive extends this |
| Timely | Immediate costs/benefits | none | Gap |
| Timely | Help people plan their response | none | Gap — and it’s cheap to fix (#19) |
| — | — | Tailored | Partly duplicates Attractive’s “personalised” |
Three findings that matter more than the mapping:
- The biggest blind spot is Defaults, not Attractive/Social. It is EAST’s highest-effect sub-principle — renewable energy as default 3% → 80% in Switzerland, persisting years; pension auto-enrolment 61% → 83% in six months [40]
[Primary]. F.A.S.T. has no Defaults element, and MBR structurally cannot set one — not with no custody, and not under Phase 1 read-only. This is a business-model boundary, not a design oversight. Every competitor with custody has it. Say so plainly rather than letting a sophisticated reader find it. - F.A.S.T. is overweighted on Easy. Two of four elements (Fast, Simple) map onto sub-principles of one of EAST’s four. Half the framework covers a quarter of the evidence base. Defensible as a deliberate bet — friction is the stated diagnosis — but state it as a bet, not as coverage.
- Adaptive has a genuine claim to advancing EAST; Tailored does not. EAST’s “Timely” was built for population-level government campaigns where the moment is a calendar event. Adaptive — firing at an individual’s threshold crossing (this user’s promo reverts on this date; this user’s card balance now exceeds their savings yield) — is a real extension, and exactly what the teaser-decay structure demands. Tailored substantially duplicates Attractive’s personalisation. Make the advance claim for Adaptive alone.
MINDSPACE: six of nine levers unused (Messenger, Incentives, Norms, Priming, Affect, Commitments, Ego — Ego makes seven), one unavailable (Defaults), one partial (Salience). F.A.S.T. is a narrow instrument.
Is F.A.S.T. an evidence framework? No — and BIT’s own text gives MBR its best defence. EAST is explicitly evidence-derived, with a companion evaluation methodology; F.A.S.T. is a design standard. But BIT’s 2014 preface says they moved off MINDSPACE precisely because “its nine elements were hard for busy policy makers to keep in mind (itself reflecting ‘cognitive chunking’),” and because “some of the most reliable effects came from changes that weren’t easily captured by MINDSPACE… simplifying messages, or removing even the tiniest amount of ‘friction’ in a process, can have a large impact” [78] [Primary]. That is a primary-source endorsement of both F.A.S.T.’s form (four memorable elements beat nine) and its emphasis (friction removal is under-weighted by the literature relative to its real effect). Talbot did not invent that reasoning, but independently arrived where BIT arrived.
To make F.A.S.T. evidentially defensible MBR needs three things it doesn’t have: (a) a named measurable outcome per element — not “we reduced friction” but “Fast variant lifted completed switches from x% to y%”; (b) in-product holdout/A-B capability instrumenting the funnel from notification to confirmed action; (c) a public revision log recording where a principle failed and was changed — the discipline the revised EAST demonstrates when it flags non-replicating studies. Until then describe F.A.S.T. externally as a design standard — claiming evidence-derived status is falsifiable, and would be falsified.
RANKED LIST A — MBR can ship this alone
Section titled “RANKED LIST A — MBR can ship this alone”Scored on three axes: (1) effect size after the replication discount; (2) build cost + user friction; (3) deployability with no custody, no employer, read-only open banking.
GATE 0 — do this before ranking anything
Section titled “GATE 0 — do this before ranking anything”0. Run the spread-durability test. Flipper and Look After My Bills both died when cross-supplier dispersion collapsed [72][73]. Model MBR’s subscription against the narrowest historical Canadian dispersion, not today’s 274bp. If the fee exceeds achievable savings in a compressed-spread year, the model fails the way Flipper failed — and no amount of behavioural design rescues it. This is the item that decides whether the rest of the ranking matters.
| Rank | Strategy | Why here | Evidence | Effort |
|---|---|---|---|---|
| 1 | Pre-filled, courier-ready switch package + fee reimbursement (#13, #14, #15) | Only intervention combining a large measured effect with zero custody requirement. 3%→12% and +15.7pp are the two largest third-party-deployable effects in the entire catalog | [FR] [Primary] | High build, highest payoff |
| 2 | Teaser-decay trigger (“your promo reverted on [date]“) | Recurring, incumbent-proof, no competitor does it, converts one-shot comparison into a retention loop, and is the cleanest demonstration of Adaptive. Rides the best-evidenced message class (#20, #21) at a genuine event | [FR] by analogy | Low–medium |
| 3 | Short, real deadlines tied to real events (#22) | 1-week deadline beat 3-week (p=0.01) and matched paying people. Nearly free | [1RCT] [Primary] | Very low |
| 4 | Pre-arrangement framing rewrite (#21, #20) | “Your better rate is reserved — confirm” beats “compare rates now.” Winning variants across 4.4M people. Copy change only | [FR] [Primary] | Very low |
| 5 | Re-gate the Engine on P(acts) × dollars (#29) | Targeting on low baseline is detrimental; intermediate-baseline is best. Treating half the population retained ~75% of benefit. Pure algorithm change | [1RCT] [Partial] | Low |
| 6 | Implementation-intention prompt: pick a date AND time (#19) | +4.2pp vs +1.5pp n.s. for date-only. One extra form field | [1RCT] [Primary] | Very low |
| 7 | Non-interruptive presence layer (#33) — reframe the Net Worth Thermometer as the answer to “zero alerts = unfalsifiable product” | Solves the renewal problem the money-back guarantee currently papers over. Clinical CDS evidence says make low-value alerts non-interruptive, not absent | [FR] analogue | Medium |
| 8 | Simplify the notice itself (#23, #26) | 0.14→0.23 from simplification alone; front page 3%→6%, identical content on the back page zero | [1RCT] at scale | Low |
| 9 | CDIC-constrained multi-institution allocation | $100k × 9 categories × N institutions is a real optimisation no bank will ever compute for a customer. Structurally unavailable to incumbents | [Primary] structural | Medium |
| 10 | Reframe the mortgage pitch at the 37% | Largest named cause of inertia is the relationship default, not ignorance. Show the Bureau’s 6.1bp incumbent premium in dollars over the term | [Primary] | Low |
| 11 | Loss framing + Messenger identity (#27, #34) | Both free; both entirely absent from F.A.S.T.; Messenger is the top blind spot for a non-custodial player | [1RCT]/[UNEV] | Very low |
| 12 | In-cohort social norms (#35) | Only if truthful and in-cohort — a national norm would say most Canadians don’t switch and would backfire | [FR] w/ caveat | Low |
| 13 | Run a megastudy, not a campaign (#32) | Forecasters — including experts — cannot predict winners. Test many variants against one objective outcome with a real control arm | [FR] [Primary] | Medium, ongoing |
| 14 | Instrument realised dollars, not switch counts | The health literature is full of interventions that won the proximate metric and lost the real one. A switch into a reverting teaser is exactly that failure | [Primary] principle | Low |
Do not build: rate-alert badges with no action attached (#40 null), stigma-reduction/reassurance copy (#39 null), naming competitors’ rates in a communication (#37 negative), gamification (#43), or any habit/streak mechanic (no substrate — the behaviour is too infrequent).
RANKED LIST B — requires a partner or institution
Section titled “RANKED LIST B — requires a partner or institution”(Chilton / Jeff pitch material — this is the “what we unlock with a partner” slide.)
| Rank | Strategy | Partner required | Evidence | Why it’s worth the deal |
|---|---|---|---|---|
| 1 | Default / auto-enrolment into a better-rate product | Bank or employer | 37.4%→85.9% [15] [Primary] | The field’s single highest-effect intervention; permanently unavailable to MBR alone |
| 2 | Write-access execution (Phase 2 open banking) | Regulatory timeline + accredited status | Phase 1 read-only [66] | Converts “here is $1,880” into “done.” Turns rank-1 of List A from paperwork into one tap |
| 3 | Incumbent-lender-branded reminder at a rate event | A lender partner | 8.9% → 15.7% (+76%) [25] — but the standing came from being the customer’s own lender | The +76% is a ceiling MBR cannot inherit as a stranger. A co-branded letter might |
| 4 | Tax-refund-moment savings prompt | Tax software (Wealthsimple Tax, H&R Block, Intuit) | +42–64% relative deposit rate [20] [Partial] | The single best-timed windfall moment in the Canadian year |
| 5 | Save More Tomorrow / auto-escalation | Employer + payroll | 3.5% → 11.6% over 28 months [16] | The famous one. Needs a payroll relationship, full stop |
| 6 | Distribution via a trusted messenger (Chilton) | The relationship itself | MINDSPACE Messenger [38]; FCA suspicion finding [74] | Directly fills F.A.S.T.’s Messenger blind spot — a named, trusted human is the highest-value version of a lever MBR otherwise cannot pull |
| 7 | Commitment products | Bank with custody | +81% savings, 28% take-up [18]; CARES 11% take-up [49] | High per-user effect, low take-up. A niche feature by construction — never the main funnel |
| 8 | Round-ups / prize-linked savings | Custodian; PLS legality varies by province | Round-ups: no causal estimate [22]; PLS Kenya RCT not significant [21] | Listed for completeness. Weakest items on this list — do not lead a pitch with them |
Proposed USP Revision (for Talbot’s decision — MBR/Strategy/*.md NOT edited)
Section titled “Proposed USP Revision (for Talbot’s decision — MBR/Strategy/*.md NOT edited)”What changes and why: the three differentiators are individually sound; their ordering is inverted relative to the evidence. #3 does the work, #1 is table stakes. The 20-word pitch also currently leads with rate outcomes (“Higher savings rates. Lower debt rates.”) and buries the execution promise — which is the only part with a large measured effect behind it.
Draft — The 20-Word Pitch
Section titled “Draft — The 20-Word Pitch”We don’t just find you a better rate. We put the paperwork in front of you, signed and ready. Benefit guaranteed.
Draft — The Three Differentiators (reordered)
Section titled “Draft — The Three Differentiators (reordered)”1. One-Click Execution — Most tools tell you what to do. MBR does it. We pre-fill the transfer forms, coordinate the signature, and reimburse eligible closing fees. Your job is to click once. (The evidence: across four independent field experiments — college aid, savings switching, mortgage refinancing, benefit take-up — pre-filled paperwork moved behaviour 4× more than being told the dollar amount. In a 124,000-person trial with the UK regulator, a “just sign here” form moved switching from 3% to 12%; personalised savings figures moved it by an amount indistinguishable from zero.)
2. We Only Interrupt You When It’s Worth It — The Hassle Threshold Engine monitors continuously and stays silent unless a verified, personalised opportunity clears your threshold — including the promo rate that reverts next month and the bank will never mention. Every alert that arrives means the math is unambiguously in your favour.
3. Benefit Guaranteed — If the market never moves enough to trigger an alert, your full annual fee is refunded. We only earn when you gain.
Deliberate changes from the current version:
- Execution promoted to #1, net-benefit math folded into #2 as the gate, not the pitch.
- “Most users are alerted fewer than 3 times per year” removed. No retrieved study tests that dose; Karlan’s working reminder effect came from monthly contact. Replace with a claim about quality, not count, until MBR’s own holdout arm gives it a number.
- Teaser reversion added to #2 — the one trigger no incumbent will ever send and no comparison site tracks.
- “MBR is not a rate comparison site” framing kept and strengthened: the differentiator is not better information, it’s that the form arrives filled in.
Also recommended, not drafted here: move the Anti-Pitch out of the USP’s supporting evidence and into brand/values, where it belongs. It is a genuine ethical commitment and a plausible referral trigger — but there is no measured evidence that agent positioning increases uptake, and the FCA found consumers already assume comparison sites act in their interest. Selling it as a conversion lever risks it not being one.
Next Steps
Section titled “Next Steps”- Run Gate 0 — the spread-durability test. Model MBR’s subscription against the narrowest historical Canadian HISA/GIC/mortgage dispersion over a full rate cycle. Flipper is the precedent and it is the only finding here that can invalidate the business rather than reorder its roadmap.
- Decide on the USP reordering above, then update
MBR/Strategy/Unique Selling Proposition.md(deliberately untouched by this research pass) and propagate to the pitch deck and Chilton proposal. - Fix the Nobel line everywhere it appears in marketing copy, using the verified wording in Premise Verification §1.
- Ship the four zero-cost items this quarter — short real deadlines, pre-arrangement framing rewrite, date+time implementation-intention prompt, loss framing. All copy or one form field; all evidence-backed; none require custody.
- Re-gate the Hassle Threshold Engine on P(acts) × dollars and stand up a holdout arm so the alert-dose question becomes an owned dataset rather than an assertion. This is the Data Moat — currently claimed but not yet instrumented.
- Verify the three open Canadian items before any of them appears in investor material: the retail-volume threshold defining mandated open-banking participants (does EQ/Wealthsimple/Simplii qualify at launch?); whether the Bureau’s straight-switch stress-test exemption was implemented; and actual registered transfer-out fees from each institution’s own disclosure (all figures here are
[Unverified]). - Resolve whether MaxMyInterest takes bank compensation — it is the closest structural analogue to MBR, and if the flagship “user-paid agent” is quietly supply-side-funded, no pure version of the model has ever been proven.
Sources
Section titled “Sources”- [1] RCTs to Scale: Comprehensive Evidence from Two Nudge Units — DellaVigna & Linos, NBER WP 27594 / Econometrica 2022 [Primary]
- [2] No evidence for nudging after adjusting for publication bias — Maier et al., PNAS 2022 [Primary]
- [3] Left-truncated effects and overestimated meta-analytic means — Bakdash & Marusich, PNAS 2022 [Primary]
- [4] Megastudies improve the impact of applied behavioural science — Milkman et al., Nature 2021 [Primary]
- [5] (see [28])
- [6] Bottlenecks for Evidence Adoption — DellaVigna, Kim & Linos, NBER WP 30144 [Primary]
- [7] Effect of Reminder Devices on Medication Adherence (REMIND) — Choudhry et al., JAMA Intern Med 2017 [Primary]
- [8] Interventions to Improve Medication Adherence: A Review — Kini & Ho, JAMA 2018 [Partial]
- [9] The Role of Application Assistance and Information in College Decisions (H&R Block FAFSA Experiment) — Bettinger, Long, Oreopoulos & Sanbonmatsu, NBER WP 15361 / QJE 2012 [Primary]
- [10] Testing the Effectiveness of Consumer Financial Disclosure: Experimental Evidence from Savings Accounts — Adams, Hunt, Palmer & Zaliauskas, JFE 2021 [Primary]
- [11] Failure to Refinance — Keys, Pope & Pope, JFE 2016 [Primary]
- [12] Psychological Frictions and the Incomplete Take-Up of Social Benefits — Bhargava & Manoli, AER 2015 [Primary]
- [13] Regulating Consumer Financial Products: Evidence from Credit Cards — Agarwal, Chomsisengphet, Mahoney & Stroebel, NBER WP 19484 [Primary]
- [14] Megastudy shows that reminders boost vaccination but adding free rides does not — Milkman et al., Nature 2024 [Primary]
- [15] The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior — Madrian & Shea, NBER WP 7682 / QJE 2001 [Primary]
- [16] Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving — Thaler & Benartzi, JPE 2004 [Primary — distributed copy; the widely-quoted “13.6% by the fourth pay raise” was NOT verified this session]
- [17] Automatic Enrollment with a 12% Default Contribution Rate / Simplification and Saving — Beshears, Choi, Laibson & Madrian, NBER [Partial]
- [18] Household Decision Making and Savings Impacts: Commitment Savings in the Philippines — Ashraf, Karlan & Yin, HBS 09-100 (restates QJE 2006) [Primary]
- [19] Earmarking and Partitioning: Increasing Saving by Low-Income Households — Soman & Cheema [Partial]
- [20] Refund to Savings: tax-time savings experiment — Washington University Center for Social Development [Partial]
- [21] Effect of a Prize-Linked Savings Intervention on Savings and Healthy Behaviors Among Men in Kenya — PMC [Partial]
- [22] Keep the Change Savings Program — Bank of America [Partial]
- [23] Nudging COVID-19 Vaccine Uptake by Changing the Default: An RCT — PubMed [Partial]
- [24] Effect of the Polypill on Adherence and Prevention of Cardiovascular Diseases — PMC [Primary]
- [25] The Last Mile of Monetary Policy: Inattention, Reminders, and the Refinancing Channel — Byrne et al., NBER WP 31043 [Primary]
- [26] Getting to the Top of Mind: How Reminders Increase Saving — Karlan, McConnell, Mullainathan & Zinman, NBER WP 16205 / Mgmt Sci 2016 [Primary]
- [27] Using Implementation Intentions Prompts to Enhance Influenza Vaccination Rates — Milkman et al., NBER WP 17183 / PNAS 2011 [Primary]
- [28] A 680,000-person megastudy of nudges to encourage vaccination in pharmacies — Milkman et al., PNAS 2022 [Primary]
- [29] A megastudy of text-based nudges encouraging patients to get vaccinated at an upcoming doctor’s appointment — Milkman et al., PNAS 2021 [Primary]
- [30] The effect of deadlines on cancer screening completion: an RCT — Scientific Reports 2021 [Primary]
- [31] The Fresh Start Effect: Temporal Landmarks Motivate Aspirational Behavior — Dai, Riis & Milkman, Mgmt Sci 2014 [Primary]
- [32] Message framing and buying behavior: A field experiment — Ganzach & Karsahi [Partial]
- [33] The MPG Illusion — Larrick & Soll, Science 2008 + energy-label framing studies [Partial — Science full text inaccessible, expired TLS cert on hosted PDF]
- [34] Machine Learning Who to Nudge: Causal vs Predictive Targeting in a Field Experiment on Student Financial Aid Renewal — arXiv [Partial]
- [35] An Evaluation of the Impacts and Implementation Approaches of Financial Coaching Programs — Theodos et al., Urban Institute 2015 [Partial]
- [36] Choice Proliferation, Simplicity Seeking, and Asset Allocation — Iyengar & Kamenica, J Public Econ 2010 [Partial]
- [37] Drug-drug interactions that should be non-interruptive to reduce alert fatigue in EHRs — Phansalkar et al. [Primary]
- [38] MINDSPACE: Influencing behaviour through public policy — Dolan, Hallsworth, Halpern, King & Vlaev, Institute for Government / Cabinet Office 2010 [Primary]
- [39] DP18/6: Price discrimination in the cash savings market — Financial Conduct Authority [Primary — the +3%/+9% RCT results are restated within DP18/6 from FCA Occasional Paper No. 19, not fetched directly]
- [40] EAST: Four simple ways to apply behavioural insights (revised edition) — Behavioural Insights Team [Primary]
- [41] Financial Literacy, Financial Education, and Downstream Financial Behaviors — Fernandes, Lynch & Netemeyer, Management Science 2014 [Primary — accepted MS hosted on a third-party site; page numbers differ from the journal]
- [42] Gamification Can Help Consumers Reach Their Saving Goals — CEPR [Partial]
- [43] The Goal-Gradient Hypothesis Resurrected — Kivetz, Urminsky & Zheng, JMR 2006 [Partial]
- [44] Money-back guarantee literature (Service Science 2014 and related) [Partial/Unverified — no financial-services evidence retrieved]
- [45] Lower Pill Burden and Once-Daily Antiretroviral Treatment Regimens for HIV: A Meta-Analysis of RCTs — Nachega et al. [Primary]
- [46] Community Pharmacy Automatic Refill Program Improves Adherence to Maintenance Therapy — AJMC [Partial]
- [47] Discontinuing financial incentives for adherence to antipsychotic depot medication: long-term outcomes — PMC [Partial]
- [48] Randomized Trial of Four Financial-Incentive Programs for Smoking Cessation — Halpern et al., NEJM 2015 [Primary]
- [49] CARES: Commitment Savings for Smoking Cessation (Philippines) — Giné, Karlan & Zinman, AEJ: Applied 2010 [Partial]
- [50] The Prize in Economic Sciences 2017 — Press release — Royal Swedish Academy of Sciences [Primary]
- [51] ideas42 homepage [Primary]
- [52] Irrational Labs — Financial Services [Primary]
- [53] Common Cents Lab / Duke Center for Advanced Hindsight partner reports [Partial]
- [54] BEworks financial-services sponsors (Newswire release) [Partial]
- [55] Behavioural Insights Team annual update / ownership [Partial]
- [56] FCAC behavioural research (Refund-to-Savings, credit-card quiz) [Partial — canada.ca research index returned 403]
- [57] Financial Education Affects Financial Knowledge and Downstream Behaviors — Kaiser, Lusardi, Menkhoff & Urban, NBER WP 27057 / JFE 2022 [Primary]
- [58] Does Financial Education Impact Financial Literacy and Financial Behavior, and If So, When? — Kaiser & Menkhoff, World Bank Economic Review 2017 [Primary]
- [59] Exploring consumer behaviour in mortgage renewal — FCAC, Monthly Financial Well-Being Monitor, 2 Mar 2026 [Primary]
- [60] Strengthening Competition in the Financial Sector: Submission — Competition Bureau Canada, 21 Mar 2024 [Primary]
- [61] Your Data, Your Control — Competition Bureau Canada, 15 Jan 2026 [Primary]
- [62] Chartered bank selected liabilities: Month-end (formerly C2) — Bank of Canada, obs. May 2026 [Primary]
- [63] TD Canada Trust posted account rates — retrieved 13 Aug 2026 [Primary]
- [64] High-interest savings account rates — Ratehub.ca, updated 13 Aug 2026 [Secondary]
- [65] How deposit insurance works — CDIC [Primary]
- [66] Consumer-Driven Banking Regulations — Canada Gazette Part I, Vol. 160 No. 26, 27 Jun 2026 [Primary]
- [67] The new Consumer-Driven Banking Act explained — DLA Piper, Apr 2026 [Secondary]
- [68] MaxMyInterest FAQ [Primary]
- [69] Nous pricing / how I fund this [Primary]
- [70] Flagstone — How it works [Primary]
- [71] Borrowell / Ratehub referral-commission model [Partial]
- [72] Flipper closure and business model (Money To The Masses; Energy Live News) [Partial — both full-text fetches returned 403; details from search snippets]
- [73] Look After My Bills wind-down [Partial]
- [74] TR14/11 Price comparison websites: thematic review — FCA [Partial]
- [75] Conflict-of-interest disclosure literature (Loewenstein/Cain, PLOS Medicine; BEUC MiFID II response) [Partial]
- [76] Effects of workload, work complexity, and repeated alerts on alert fatigue in a CDSS — PMC [Primary]
- [77] Override rate of drug-drug interaction alerts in CDSS: systematic review and meta-analysis — Health Informatics Journal 2024 [Primary]
- [78] EAST: Four simple ways to apply behavioural insights (2014 original) — Behavioural Insights Team [Primary]
- [79] Impact Canada — Increasing Take-Up of the Canada Learning Bond — Privy Council Office Impact and Innovation Unit, 2018 [Partial — CLB take-up 5.7%→8.8% (+55%), n>140,000 children; validates the behaviourally-informed letter format in Canada]
Explicitly not retrieved this session (stated rather than inferred): a randomized just-in-time-vs-advance financial-education comparison; a head-to-head cost-effectiveness study of education vs nudges/defaults; measured evidence on money-back guarantees in financial services; measured evidence that fiduciary/agent positioning increases consumer uptake; a causal estimate for round-up saving; Canadian data on credit-card-balance-vs-savings behaviour; the T2033 transfer process mechanics; the open-banking retail-volume threshold; and whether the Competition Bureau’s straight-switch stress-test recommendation was implemented. Morningstar’s behavioral research team, Ontario BIU, US OES and Impact Canada’s behavioural science pages were inaccessible (403/404/TLS) and are deliberately not characterized.
This report is frozen prior art (see Behavioural-Solutions), unfrozen only for this Notes glossary plus acronym links below — every occurrence, not just first use (Talbot, 2026-08-15) — no other body edits. Evidence-tier tags ([Primary]/[Secondary]/[Partial], [FR]/[1RCT]/[LAB]/[NULL]/[UNEV]) are defined inline at their own first use (see §“Evidence tier” near the strategy tables) and are not repeated here.
- FCA — UK Financial Conduct Authority (financial-services regulator)
- CASS — Current Account Switch Service (the UK bank-switching guarantee scheme)
- BIT — UK Behavioural Insights Team (“Nudge Unit”) — originated MINDSPACE, then EAST
- EAST — BIT’s current framework: make it Easy, Attractive, Social, Timely
- MINDSPACE — BIT’s earlier, retired 9-lever framework: Messenger, Incentives, Norms, Defaults, Salience, Priming, Affect, Commitments, Ego
- USP — Unique Selling Proposition
- RCT — Randomized Controlled Trial
- SD — Standard Deviation; CI — Confidence Interval
- pp — percentage points (an absolute, not relative, difference); ns — not statistically significant
- AER — American Economic Review; JFE — Journal of Financial Economics; QJE — Quarterly Journal of Economics; JPE — Journal of Political Economy; PNAS — Proceedings of the National Academy of Sciences; NEJM — New England Journal of Medicine; JAMA — Journal of the American Medical Association; NBER (WP) — National Bureau of Economic Research (working paper) — journal/publisher names, citation context only
- FAFSA — Free Application for Federal Student Aid (US)
- EITC — Earned Income Tax Credit (US)
- CARD Act — US Credit Card Accountability Responsibility and Disclosure Act (2009)
- REMIND — name of the cited medication-reminder-device trial (Choudhry et al.); not a further acronym
- CDIC — Canada Deposit Insurance Corporation; FCAC — Financial Consumer Agency of Canada