Experiments I'd run next

How the pieces fit. Under Mental Accounting, money is not one pool in someone's head. It gets sorted into labelled buckets, and a labelled bucket gets protected in ways a generic balance does not. Salience is what does the labelling work in practice. A name people can picture (“Save for Phone”) stands out against the interchangeable rows around it (“Savings Pocket 2,” “Everyday Account”), the way one bold figure jumps off an otherwise plain statement.

What already ran. The high school field session put the labelling half in front of a room of Year 8/9 students with paper envelopes. Roughly $30 piled up in the “save” envelopes without anyone consciously deciding to save that amount.

What this isolates. The digital version asks whether a name that stands out in an account list protects a balance the same way a hand-labelled envelope did in that room.

Ask someone to name a new savings pocket when they open it, instead of leaving the auto-generated label in place, and the pocket should behave differently afterward. A smaller share of that balance should hit zero within 90 days. The naming step should not add enough friction to cut onboarding completion in a meaningful way.

Illustrative, not real product data. Assume 30% of customers who open a second savings pocket withdraw its full balance within 90 days. The generically-labelled pocket behaves less like “savings” and more like a second everyday-spend account with extra taps.

What CBA already proved. Buell and Choi ran a live field experiment in CBA's credit card acquisition funnel. Putting each card's drawbacks on equal footing with its perks, a “Good and the Bad” page instead of benefits-only marketing, left overall take-up statistically flat while improving which card people chose. Customers who saw the downsides later spent 9.9% more, cancelled 20.5% less, and missed fewer payments. The full teardown, with the real numbers, sits on this site's Tradeoff Transparency page.

What still sits untested. That work stayed inside one product category: card versus card. It never asked the next-layer question: credit card against the other ways the same person could finance the same purchase, a personal loan or a Buy Now, Pay Later plan. Those options still live on separate pages, owned by separate teams, with no shared drawbacks shown against each other.

Put a customer's real financing options, credit card, personal loan, BNPL, on one screen, with each option's drawbacks named as plainly as its benefits. Some people should move toward the fit that matches how they plan to use the money. Overall take-up should stay roughly flat. CBA's within-category version changed which card people picked without changing how many applied, and this is that mechanism one product-category higher.

Illustrative, not disclosed CBA data. Assume 8% of applicants who land in today's credit-card-only funnel end up choosing a personal loan or BNPL plan instead, entirely on their own initiative. The funnel never surfaces either as an alternative, so this only happens when someone already knew to go looking for one.

Where the line sits. Making a financing alternative visible for the first time changes the decision environment, so the constraint belongs up front. The design is not there to talk people out of revolving credit. For many customers a credit card is still the right tool.

What the design is allowed to do. It surfaces the other real options before commitment: inform the choice without steering it into a preferred product, the same restraint the original CBA study held. See Not Testing Is Still a Bet for the longer case for that line.

What the Uber field test showed. Halperin, Ho, List and Muir ran this at scale on 1.5 million Uber riders after a late ride. A $5 coupon apology paid for itself by restoring future spend that a generic “sorry” alone did not restore. A costly signal can do work a free apology cannot. That mechanism is documented here as Reciprocity.

Why the design needs a second branch. The same paper found that sending the identical coupon again after a second bad ride made riders spend less than if Uber had said nothing. The authors' reading: an apology reads as a promise, and recycling the exact same one a second time reads as that promise already broken. Full numbers sit in the Voltage Effect field session.

After a service failure, a costly and specific apology, a fee credit plus a plain account of what went wrong, should beat a generic scripted apology on post-incident satisfaction. Because reusing an apology verbatim can reverse the gain on a repeat failure, the design also needs a secondary branch on the same sample. Does changing the apology's content and framing keep the advantage when a fixed script would lose it?

Illustrative, not real support-desk data. Assume post-incident CSAT (1–5) currently averages 3.1 after a generic apology message, standard deviation 1.1, typical for a scripted, one-size response with no real cost attached.

Thorndike's claim, compressed. A rater who forms a strong impression on one visible trait spreads that same judgment to other traits of the same target they have not actually scored. See Halo Effect for the full study. Thorndike, and later Nisbett & Wilson, measured that at the level of one person's traits: an instructor's warmth colouring ratings of his unchanged accent and mannerisms.

Moving the same claim onto channels. Here the target is a brand's channels rather than a person's traits. The bank in this scenario has never offered a digital home loan; every application has gone through a broker or a branch appointment. It is about to launch a fully digital application: pre-approval in an afternoon, no paperwork mailed anywhere. Launch day changes nothing else: branch process, broker relationships, underwriting, and call centre stay identical to the day before.

If the halo travels. Awareness alone should matter. Knowing the bank now has one excellent digital channel should lift trust in the channels that did not change, and application intent in those same unchanged channels should rise with it.

People who see messaging about the bank's new digital home loan capability, whether or not they ever open it, should show higher intent to take a home loan with that bank. The lift should show through any channel, digital or otherwise, versus a matched group with no exposure. That is the headline claim. Awareness of one strong channel lifts trust across the relationship.

Held out of the headline. Whether people who later use the digital channel show a still larger lift is a secondary, non-randomised read (see Sub-branch). It stays out of the primary comparison.

Illustrative, not real/disclosed data. Assume 6% of existing customers who are in the market for a home loan in a given quarter say they'd seriously consider getting it with this bank, through any channel. That is before the digital product launches or any awareness messaging goes out.

The failure mode to watch. A halo is trust extended to something nobody actually checked. If digital-channel awareness lifts intent to apply through a branch or broker, and those channels have not been resourced to match the promise the digital product makes, the bank has a problem. It converts people on trust it has not earned in the channel they actually use.

What this test is not permission for. It is not cover to under-invest in non-digital application support because the digital halo is doing the persuading; see Not Testing Is Still a Bet. The outcome guardrails below exist specifically to catch that failure mode before a customer hits it.

Evidence from cars and price tags. Lacetera, Pope and Sydnor showed that a number's leftmost digit dominates how people judge its size. Crossing a leading-digit boundary, 39,999 miles to 40,000, or $2.99 to $3.00, shifts perceived value far more than the tiny real gap justifies. Their evidence is 22 million real used-car sales; Thomas & Morwitz found the same pattern in ordinary retail prices.

Onto an interest rate. Neither study used a lending rate. This blueprint puts Left Digit Bias on a home loan's headline figure. A rate advertised at 5.99% p.a. and one at 6.00% p.a. cost a borrower almost the same in real repayments. The leading digit still flips: 5.99% reads as “5% something,” while 6.00% reads as “6%” outright.

Expected behaviour. If the mechanism travels the way it does for an odometer reading, 5.99% should be judged, and chosen, as meaningfully cheaper than 6.00%, out of proportion to the real repayment difference.

A home loan advertised at 5.99% p.a. should look cheaper than an otherwise identical loan advertised at 6.00% p.a. It should also convert to more started applications, even when the two rates sit close enough that the monthly repayment gap is trivial.

Illustrative, not real/disclosed data. Assume 4% of visitors to the rate page who are shown a 6.00% p.a. headline rate go on to start an application within the same session.

Already industry practice. This experiment does not invent a new advertising trick. .99 rate framing is already standard, legal advertising across lending, and the real cost stays disclosed via the comparison rate whichever headline number is shown.

The measurement question. How far that convention alone moves behaviour. With real numbers, a bank can decide whether to keep leaning on it, limit how far it leans, or counter it by surfacing the comparison rate more prominently. See Not Testing Is Still a Bet for why measuring this beats assuming it is fine because everyone already does it.

Where the mechanism already shows up in product. Bankwest's real Interesting Rates promotion pays every Easy Saver customer a genuinely competitive 5.75% p.a. everyday rate, then draws 50 winners a month for an 11.50% p.a. rate on top. Probability Weighting is what makes that structure bite: a small, real chance of the better rate gets weighted in people's minds far past its true odds.

Two findings that sharpen the weight. Rottenstreich and Hsee (2001) found people overweight a small chance of something emotionally vivid even more than an equally small chance of plain cash of the same value. That is why a fast-food deal sells “free food for life” rather than its cash equivalent. A year of 0% interest on a real mortgage is at least as vivid: it converts into a large, concrete, easy-to-picture sum. Burger et al. (2020) found a separate effect: simply holding a lottery-style entry lifts happiness before the draw resolves, win or lose. A home loan application is normally a stressful, multi-week wait. A live entry gives an applicant something hopeful to hold during it, a different benefit from the conversion lift Probability Weighting alone would predict.

Why this is not a prize-linked savings swap. Classic prize-linked savings research, including Filiz-Ozbay, Guryan, Hyndman, Kearney and Ozbay's 2015 study, tests a real substitution: part of a saver's own guaranteed return is swapped for lottery expected value. Bankwest's promotion does not do that; its 5.75% rate exists whether or not anyone enters the draw. A home loan version that pairs an already-competitive guaranteed discount with a separate chance at 0% sits even further from that substitution: none of the applicant's real rate funds the prize. The test isolates Probability Weighting's own pull.

What gets compared. Advertising a genuinely competitive guaranteed discount alongside a separate, real chance at 0% p.a. for 12 months, versus advertising that same guaranteed discount alone, with the guaranteed rate held constant between arms. A second prediction runs in parallel: simply holding a live entry should change how applicants feel during the underwriting wait, beyond whether they start applying.

Visitors who see the guaranteed discount plus a chance at 0% p.a. for 12 months should start more applications within 7 days than visitors who see the identical guaranteed discount with no draw. The guaranteed rate never changes between arms. Separate from that conversion lift, applicants in the treatment arm should report less stress and higher satisfaction during underwriting than control applicants. That matches the anticipatory-happiness effect Burger et al. found in lottery ticket-holders, whether or not they win the draw.

Illustrative, not real/disclosed data. Assume 8% of visitors to the new home loan rate page start an application within 7 days when the page advertises the guaranteed discount alone.

Hard constraints on the offer. The guaranteed discount has to be genuinely competitive against the rate the bank would otherwise offer that same applicant. It cannot be a rate held back elsewhere to help fund the draw. It must stay identical in both arms so the draw is the only variable under test. The odds, the number of winners, and the exact prize, 0% p.a. for 12 months on a capped loan amount, need disclosing as plainly as the guaranteed rate itself.

What success is allowed to mean. The question is whether a real, cited effect helps a genuinely competitive offer reach more of the people it is meant for. It is not whether draw excitement can be tuned to pull applicants into borrowing more than they can service. Loan size and serviceability assessment run identically in both arms, and a guardrail metric below checks that directly. See Not Testing Is Still a Bet.

Dual entitlement in one line. The Perceived Fairness finding from Kahneman, Knetsch and Thaler is simple enough to state outright. An identical price rise looks fair when framed as protecting the seller's existing profit from a real cost increase. It looks unfair when it looks like exploiting a shift in demand or bargaining power. The number alone does not decide the verdict; the story behind it does. Their evidence is a telephone survey of Toronto and Vancouver residents judging realistic vignettes. One was a hardware store that raised snow-shovel prices the morning after a storm (rated unfair by roughly 82% of respondents). Another was a landlord charging existing versus new tenants differently for the same unit.

What the vignettes never covered. Those surveys measured stated opinions about hypothetical price changes. They did not measure real behaviour after a real change. None of them used a financial-services fee or rate notice, where the cost being passed on (funding costs, compliance costs) is far less visible to a customer than a snowstorm is.

What a live fee notice can show. If the same dual-entitlement logic holds, a bank that names the real cost driver behind a fee or rate increase should see the change land as fairer. Unlike KKT's survey, this design can check whether that shows up in behaviour: fewer support contacts, fewer customers leaving.

Customers who get a fee or interest-rate increase notice that names a real, verifiable cost driver should be less likely to contact support to question or dispute the change. They should also be more likely to rate it as fair than customers who get an otherwise identical notice stating only the new number. The actual new fee or rate stays the same between groups.

Illustrative, not real/disclosed data. Assume 18% of customers who receive a bare fee-increase notice, new amount only, no explanation, contact support to question or dispute the change within 30 days.

Honesty constraint. The mechanism only works cleanly when the disclosed cost driver is real and verifiable, a genuine rise in funding costs, compliance costs, or similar that the bank could substantiate if asked. Citing a vague or invented cost story to manufacture fairness would be deceptive framing, and sits outside this experiment's scope.

Scope of the claim. The test asks whether naming a true cost driver changes how a genuine, already-decided fee change lands. It does not ask whether that framing can justify an increase that would not otherwise be defensible on its own. See Not Testing Is Still a Bet for why measuring this matters more than assuming good framing is harmless.

Why the payment method changes the felt bill. Payment Transparency says a purchase's felt cost depends on whether the payment method makes someone rehearse the amount and whether the money leaves immediately. The Credit Card Premium is the sharpest documented case. Bidders told they would pay by credit bid close to double what cash-instructed bidders bid for the identical tickets, because credit strips out both cues.

Suggested in the principles, never isolated. Both principle articles' “Where it can help” sections, and the synthesis in How You Pay Changes What You Spend, argue those two cues can be put back on purpose. One version is a running balance at the moment of tap. Another is a receipt that totals spend by category. Neither study tested whether simply telling someone the mechanism, in plain language, changes what they actually spend. That is a weaker intervention than redesigning the interface. It relies on a customer remembering and acting on a fact rather than on a changed screen.

What the arms isolate. Whether education alone, with no interface change and no restricted product, moves real spending, and whether timing matters: repeated messages close to the moment of use, versus a one-time explanation before any spending has happened.

Customers told the payment-transparency heuristic in plain language should spend less on their lowest-transparency linked payment method (a stored or linked credit card) over the following months than customers told nothing. The cut should be larger and last longer when the heuristic arrives repeatedly as a contextual in-app message near spending. It should be smaller when it arrives once as onboarding content before any transaction. Rehearsal and immediacy operate on what is salient while paying; a single abstract explanation earlier is a thinner version of the same idea.

Illustrative, not real/disclosed data. Assume customers in the control arm spend an average of $780 a month on their lowest-transparency linked payment method in the 90 days after opening a new debit account. Use a standard deviation of $360, reflecting how unevenly card spending is actually distributed across a customer base.

How the message has to read. It must state the real mechanism honestly, something like “paying by credit doesn't feel like spending right away, which is part of why people spend more on it.” Shame-based or fear-based framing is out. The same wording has to apply regardless of income or existing balance, so it reads as financial literacy rather than a judgment about who is trusted to spend.

What the test is for. Whether informing someone of a real, cited mechanism changes their own behaviour once they know it. The question is not whether the copy can be tuned to push spending up or down on demand. See Not Testing Is Still a Bet for why measuring this matters more than assuming a well-meaning message is automatically harmless.

What Barasz and co-authors found. Pseudo-Set Framing groups otherwise arbitrary items into a visible set, and people push harder to finish that set. The pull survives when the reward stays fixed, when finishing costs effort, and even when participants are told the grouping is invented. Five studies covered gambling, effort tasks, giving, and purchasing.

Where product onboarding sits. Those studies never used a live product's own multi-step setup. Banks, telcos, and insurers already ask for several small, useful tasks on day one: verify ID, link a funding source, name a beneficiary, set a goal. Those tasks usually appear as a flat checklist with no shared visual frame.

What the arms compare. The same required steps, pictured either as ordinary checklist rows or as wedges of one circle to complete. Nothing new is required, and no extra step is added. The only change is whether unfinished work looks like an incomplete set. That is a different lever from the Goal Gradient Effect, which pulls people who are objectively closer to a finish line; here the group simply looks unfinished.

New customers who see remaining setup steps as wedges of one “complete your profile” circle should finish every required step inside 14 days more often than customers who see those same steps as a plain checklist. Step count, effort per step, and the finishing reward stay fixed across arms.

Illustrative, not real/disclosed data. Assume 41% of new customers complete all 4 required setup steps (ID verification, linked funding source, a named beneficiary, a savings goal) within 14 days of opening an account under the existing plain-checklist flow.

What each wedge is allowed to be. Every slice of the circle has to map to a step that is already required or already useful to the customer. No filler step exists only to round out the shape. Completion pressure cannot exceed what the checklist arm already uses: no delayed access to core banking, no extra interstitial nags, no dark-pattern blocking.

What the measurement is for. Whether a cited completion mechanism helps people finish setup they already need. The visual is not there to hurry someone through a step they had good reason to postpone. See Not Testing Is Still a Bet.

Two pieces of evidence already on the table. In a controlled lab, people offered a lottery-style bonus deferred payment more often than people offered a guaranteed bonus of the same expected value. Low-balance participants showed the strongest pull. Michigan's Save to Win programme, covered in The Lottery You Can't Lose, moved $8.56 million into new accounts in its first year on that same structure.

What neither source isolates. The lab paid a one-off laboratory bonus. Save to Win is an entire account type. Neither one layers a payout-structure swap onto an existing digital savings feature while holding the rest of the product fixed.

The swap under test. Take part of an existing feature's guaranteed bonus rate and replace it with a prize draw of equal expected value. Same app, same customers, same expected payout. Only the shape of that payout changes.

Customers offered a prize-draw bonus, funded from the same pool that would otherwise pay a guaranteed rate bump, should deposit more net new money into the dedicated savings feature over 60 days than customers offered the rate bump directly. Expected total payout matches across arms. The largest gap should sit among customers with the lowest starting balances, matching the lab pattern.

Illustrative, not real/disclosed data. Assume customers in the guaranteed-rate control arm deposit an average of $310 in net new funds into the dedicated savings feature over 60 days. Use a standard deviation of $410, reflecting how unevenly savings deposits are actually distributed.

How the prize pool has to be funded. It comes from the exact dollars that would otherwise pay guaranteed interest. The product cannot become worse in expectation. Deposits stay whole win or lose, which is how this structure avoids triggering Loss Aversion on the principal itself. Odds and expected value need the same plain disclosure an interest rate would get, not excitement copy that buries them.

What success is allowed to mean. A cited payout-structure effect helping people save more of their own money. Framing aimed at low-balance customers who feel the lottery pull most cannot be tuned to push deposits past what they can spare. See Not Testing Is Still a Bet.

Kivetz, Urminsky and Zheng's lever. Effort and motivation rise as someone feels nearer a finish line. Felt closeness does the work even when real remaining distance stays fixed. A 12-stamp card with 2 stamps already filled accelerated purchasing the same way a genuinely shorter 10-stamp card did, though both still needed exactly 10 more purchases. The Goal Gradient Effect page's own “where it can help” case is an honest progress bar on a form.

Still untested on a live booking flow. Home loan appointment booking is a long multi-step form people abandon mid-way: contact details, loan and property details, meeting type, time slot. No published test has held that flow fixed and varied only the completion cue.

Two ways to apply the same mechanism. One arm shows an honest step count at the flow's natural, coarse chunking. The other re-chunks the identical real work into more, still-genuine steps so a customer sits further through the count at the same real point. That is the 12-stamp card's lever against a 10-stamp one, applied to a booking form. It is distinct from Pseudo-Set Framing, which pulls people because a group of steps looks unfinished rather than because remaining distance has shrunk.

Visitors shown a real step-progress indicator, at either chunking, should complete and confirm a booking more often than visitors shown the identical flow with no progress indicator. Required fields and appointment options stay the same in every arm. Because the mechanism runs on felt closeness, splitting the same real work into more granular genuine steps should produce a larger lift. At an identical real point, the customer sits further through the count than under an honest step count at the flow's natural, coarser chunking.

Illustrative, not real/disclosed data. Assume 34% of visitors who start the appointment booking flow complete all 4 steps and confirm a booking under the existing no-progress-indicator flow.

Honesty on every screen. Step count and progress bar have to match the real number of steps and the real position at every screen. Compressing or re-ordering to shrink remaining effort is out. Total step count belongs on the first screen, so nobody commits to step one without knowing three or five more sit behind it.

Where re-chunking stays legitimate. Finer sub-steps are allowed only when each resulting step is still genuinely distinct required content. Contact details and identity verification are two real things; splitting one field in half to manufacture an extra step is not. No step at either chunking may be fabricated, reordered to hide what is left, or padded only to inflate the count.

What the cue is for. Helping people finish a booking they already want to make. The count is not there to disguise how long the form takes. See Not Testing Is Still a Bet.

Save More Tomorrow in one paragraph. Employees committed today to a contribution-rate increase that only took effect at their next scheduled pay raise, so take-home pay never shrank on the day it changed. Of 162 employees who had just declined an immediate increase, 78% accepted the deferred version. Average savings rates rose from 3.5% to 13.6% of pay across four raises (about 40 months), and 78% were still enrolled at the end. Full write-up: Smart Defaults.

The employer schedule that banking apps lack. Every tested version of that design starts from an employer who already knows the date and size of the next raise. A savings or banking app has no such calendar. At best it notices, after the fact, that a regular salary deposit just got larger.

If timing against a gain is the real lever. Whether the raise is scheduled or detected should not matter. What should matter is committing to the increase in advance, and only applying it once the extra money has already arrived.

Customers offered a one-tap commitment to raise their auto-save rate at their next detected pay rise should accept some increase more often than customers asked to raise it immediately. The size of the increase on offer is identical in both arms.

Illustrative, not real product data. Assume 9% of customers shown an in-app prompt to increase their auto-save rate by 2% right now accept it.

Deposits that look like raises. A jump in a bank deposit is often something else: a one-off bonus, a tax refund, a temporary higher-paying gig. Treating any of those as a permanent rise would auto-escalate savings against income that is about to disappear.

How detection has to behave. A detected increase should hold across at least two consecutive pay cycles before the commitment fires. Any triggered increase must reverse with one tap and refund in full for a short window afterwards, not only cancel going forward. A wrong detection should cost the customer nothing.

What the arms isolate. Whether advance commitment against a gain, rather than a felt cut today, drives acceptance. Raise-detection cannot be tuned to fire more often than the evidence supports. See Not Testing Is Still a Bet.

When reactance fires. Not Enough Choice ties the backlash to an active removal. An option that was never available does not trigger the same response. The principle's “where it can help” section states plainly that naming the reason should defuse it.

Never isolated on a real notice. That claim has sat untested against an actual feature-retirement message. No published design has held the removal, timeline, and balance migration fixed and varied only whether a reason appears.

What both arms share. Identical removal, identical timeline, identical balance migration. The only difference is whether the notice states why the feature is going.

Customers told why a feature is being removed should contact support or lodge a complaint about it less often than customers told only that it is being removed. What happens to their money is the same in both arms.

Illustrative, not real product data. Assume 22% of customers who receive a plain “this feature is going away” notice contact support or complain about it within 30 days.

Money has to move the same way. Balance migration stays identical, immediate, and lossless in both arms. Which notice a customer receives cannot change anything about their funds.

The reason has to be true. Treatment copy states the actual reason the feature is being retired. A plausible cover story invented for messaging tests is out. See Not Testing Is Still a Bet.

Guéguen and Pascual on the street. Adding “but you are free to accept or to refuse” to an otherwise identical request for bus fare raised compliance from about 10% to about 47.5%. That finding now sits inside this site's Not Enough Choice write-up.

From one cash ask to a digital setting. The original test was a single, low-stakes, in-person request. Whether the same freedom-naming line moves a digital, repeatable financial decision is still open.

The ask under test. Turning on round-up savings: a genuinely optional feature that is easy to feel nagged into. Same offer in both arms; only whether the prompt names the freedom to decline differs.

Customers shown an in-app prompt to turn on round-up savings that explicitly names their freedom to decline should accept it more often than customers shown the identical prompt without that line.

Illustrative, not real product data. Assume 11% of customers shown a plain round-up savings prompt turn it on.

The founding reactance finding. Hammock and Brehm showed that actively restricting an option makes it more wanted, beyond simply making it harder to get. Marketing already leans on that lever informally. The mechanism sits inside Not Enough Choice.

Honest upsell versus invented scarcity. This site has not yet tested that lever as an upsell where the restriction named is real and checkable. Manufactured urgency belongs to a different, dishonest category.

What the message is allowed to name. A real, existing eligibility gate. Interest in upgrading is the outcome. No invented scarcity, no fake headcount, no countdown without a deadline behind it.

Standard-tier customers shown a premium-tier message that names a genuine eligibility restriction should click through to check their eligibility more often than customers shown the identical benefits with no restriction framing.

Illustrative, not real product data. Assume 6% of standard-tier customers shown a plain premium-benefits message click through to “see if you qualify.”

Where this goes wrong. A restriction that reads as real but is not is exactly the manufactured urgency this site's Scarcity and Signpost Effect articles warn against.

What the copy may state. Only a restriction that is true at the moment it is shown: a current eligibility gate or a genuinely limited rollout. Fabricated headcounts and countdowns with no real deadline stay off the screen.

What the arms measure. Whether naming a real restriction changes interest. Inventing a restriction to manufacture demand is outside the design. See Not Testing Is Still a Bet.

Khan and Dhar's lab sequence. People first imagined committing to volunteer work, then chose between a pair of jeans and a vacuum cleaner at the same price. Imagining the virtuous act pushed choice toward the jeans (the indulgent option) compared with people who imagined no such act. A second study found the same shift toward luxury sunglasses over practical ones. Full write-up: Licensing Effect.

Imagined virtue versus a live product action. Both studies used a hypothetical virtuous act in a lab. Nothing on this site has tested whether a real, completed responsible action inside a live product changes a real choice made minutes later.

Why a loan repayment should transfer. If the mechanism is a temporary boost in feeling like a responsible person, the virtuous act here being financial (setting up an extra loan repayment) rather than social (volunteering) should not block the effect. A completed action should shift a real redemption choice the way an imagined one shifted a hypothetical purchase.

Customers who have just set up an automatic extra repayment on their credit card then reach a reward-points redemption screen. That screen offers a practical option (a statement credit) and an indulgent option (a dining or travel voucher) of equal point value. They should choose the indulgent option more often than customers who reach that same screen without having just taken the repayment action.

Illustrative, not real product data. Assume 35% of customers reaching the redemption screen without a prior prompt choose the indulgent voucher over the statement credit.

How the finding could be misused. This measures a real effect on real spending choices. Read carelessly, the same result could be taken as licence to stage a virtuous moment purely to sell more discretionary redemptions, which works against the customer's own financial interest.

What has to stay true of the setup. The extra repayment opt-in already exists as a useful feature offered on its own merits. It is not invented or timed to manufacture the licensing effect. Both redemption options stay equally visible and equally easy to pick in every arm. A guardrail metric tracks whether the extra repayment is cancelled soon after, which would signal it was ticked as a box rather than adopted.

What the arms isolate. Whether completing a responsible action shifts a later, unrelated choice. The responsible-action prompt itself is not there to appear more often than product usefulness justifies. See Not Testing Is Still a Bet.

Read and van Leeuwen's snack choice. Office workers picked fruit or junk food for the following week, either seven days early or on the day itself. A week out, hunger barely moved the choice. On the day, hungry workers took junk food far more often than workers who had just eaten, even though both groups faced the same two snacks. That timing reversal is the core of Present Bias.

Food versus cash on hand. Their reward was a snack. This site has not yet checked whether the same advance-versus-in-the-moment flip appears when the prize is money that could be spent today instead of saved.

Why a savings share should travel. If value collapses once “now” is on the table, swapping a snack for a savings percentage should not erase the pattern. Naming a share a week before a bonus or refund arrives should favour saving more than asking the identical question once the cash is already in the account.

Customers asked a week before an upcoming bonus or tax refund how much of it to move to savings should commit a higher average percentage than customers asked the identical question on the day the money arrives.

Illustrative, not real product data. Assume customers choosing on the day the money lands commit an average of 15% of it to savings.

Where an early lock-in hurts. A savings share chosen a week early can be wrong by the time the money lands if circumstances change. A surprise bill, a job loss, or an emergency can make the earlier number a bad fit.

What stays reversible. Only customers who already opted into a windfall-savings feature on its own merits enter the test. The advance commitment stays editable or fully cancellable until the money lands, one tap, no fee. A mistimed ask should cost the customer nothing.

What the arms isolate. Whether ask timing changes the committed share. Hardening advance commitments so they are harder to reverse is outside the design. See Not Testing Is Still a Bet.

Arkes and Blumer's theater season. People who paid full price for a season of plays attended more of it than people handed the identical season at a random discount. Future attendance tracked what had already been paid, not what the remaining shows were worth going forward. That is the mechanism behind Sunk Cost Fallacy.

Attendance after the fact versus a renewal prompt. Their field study measured shows attended once the ticket was already bought. This site has not yet tested whether naming a customer's own sunk cost, at the moment they would otherwise shop around, changes the next tap.

What the renewal screen adds. Put the customer's real year-to-date spend on the plan next to the renew button. The question is whether that figure alone cuts how often they open a cheaper alternative before renewing.

Customers shown their own cumulative spend on the current plan at renewal should click “compare other plans” less often than customers shown an identical renewal screen with no spend figure. The same cheaper plan stays available in both arms.

Illustrative, not real product data. Assume 22% of customers shown a plain renewal screen click "compare other plans" before renewing.

How the finding could be misused. The same lever could trap customers on a worse deal by leaning on the fallacy the Sunk Cost Fallacy article already flags. Measuring the effect is not the same as recommending it as growth copy.

What the screen is allowed to show. Only the customer's own real, checkable spend, never rounded up or estimated in the company's favour. “Compare other plans” stays equally visible and fully working in both arms. Switching never gets harder; only the accompanying figure changes.

What a positive result means. The arms ask whether a true sunk-cost figure moves behaviour by itself. They do not ask whether comparison shopping should be buried. Confirmation is a reason to avoid shipping the pattern, not a green light to ship it. See Not Testing Is Still a Bet.

Gourville's pennies-a-day finding. Restating a cost as a small daily amount, rather than one lump total, shifted which other expenses people compared it against and raised willingness to pay. That comparison shift is the heart of Temporal Reframing.

Daily framing alone versus full disclosure. Those lab arms put the daily figure in place of the total, not beside it. Legal and trust reasons usually force real checkouts to keep the true total on screen, so whether the effect survives that disclosure is still open.

What the checkout arms change. Keep the real annual total visible in both arms. Add a truthful daily-equivalent figure next to it in one arm only. Nothing else about disclosure moves. The question is whether that addition still pulls more people toward the annual plan.

Customers shown “33¢/day, billed annually at $120” should pick the annual plan over the identical monthly alternative more often than customers shown “$120/year” alone. The true total stays visible in both arms.

Illustrative, not real product data. Assume 18% of customers choose the annual plan over the monthly plan when shown the plain annual total.

Where the reframe turns dishonest. A daily-equivalent that is not a true, correctly rounded division of the real total misleads. That is the failure mode the Temporal Reframing article already names as the dishonest use of this lever.

What the copy must compute. The daily figure is the exact annual total divided by 365, rounded normally. The annual total appears in the same type size, never as fine print. The monthly plan's price and visibility stay fixed across arms.

What the arms isolate. Whether a fully disclosed daily reframe still moves plan choice. Shrinking disclosure to make the reframe hit harder is outside the design. See Not Testing Is Still a Bet.

Yadav and Monroe on transaction value. An extra discount stated once on a combined bundle price lifted perceived transaction value more than the identical dollar amount split across the bundle's separate items. That framing difference is the load-bearing claim in Bundling.

Described goods versus a live cross-sell. Their scenarios used physical products people imagined buying. This site has not yet tested whether the same framing shift changes real take-up when the “items” are financial add-on services.

What the arms hold constant. Two add-ons, one total $2 saving. State that saving once on the bundle in treatment, or $1 on each line in control. The question is whether the single bundle statement lifts how often customers take both.

Customers offered “save $2 when you bundle ID theft protection and device insurance together” should take both add-ons more often than customers offered “save $1 off ID theft protection” and “save $1 off device insurance” as separate line items. The total saving stays $2 in both arms.

Illustrative, not real product data. Assume 11% of customers take both add-ons when the two discounts are shown separately, one per item.

Where framing oversells. A bundle line that makes a saving feel larger than it is can push add-ons a customer would otherwise skip, then cancel once the bill arrives and the framing has worn off.

What the discount math must match. The bundle discount equals, never exceeds, the sum of the two item-level discounts. No inflated “bundle-only” saving may be invented to stack the framing test.

What the arms isolate. Whether framing alone changes take-up for an identical real saving. Pricing a bundle to look better than it is sits outside the design. See Not Testing Is Still a Bet.

Bertrand and Morse at the payday counter. Across 77 stores, 1,441 real borrowers saw either ordinary terms or one of several disclosure formats at the counter. APR figures, and comparisons to credit-card APRs, barely moved borrowing. Naming the real dollar fees a borrower would rack up if the loan renewed repeatedly cut borrowing over the next four months by about 11%. Full teardown: The Fine Print Nobody Reads.

Paper at a counter versus a digital confirm screen. That disclosure arrived in person, on paper, for a payday loan. Whether the same dollar-fees-over-time format, delivered inside a digital checkout instead of by a clerk, moves Buy Now, Pay Later behaviour the same way is untested. BNPL is built around the repeat, compounding late fee the original study measured.

Which decision the arms target. Like Bertrand and Morse, this is about an existing high-cost-credit borrower's next borrowing decision, not whether to finish the purchase sitting in front of them. The dollar figure is a projection from the current order's own real late-fee schedule, mirroring the original format rather than a personal history lookup.

Existing BNPL users with at least one prior late payment who see the real dollar late-fee cost of missing payments on their current order at confirm should miss fewer payments and take out fewer new BNPL purchases over the following months than users shown only the standard terms link.

Illustrative, not real BNPL provider data. Assume 24% of users with a prior late payment miss a payment again on their next order when shown only the standard terms link.

Why the disclosure is not a neutral add-on. Showing a real dollar-cost projection to someone who has already missed a payment changes the emotional temperature of the checkout. The design is not built to talk anyone out of BNPL when they intend to pay on time. Plenty of BNPL purchases clear without extra cost.

What the figure is allowed to be. A real projection tied to the current order's own instalment terms, never an invented worst case. It sits beside the purchase, never in place of it. Checkout speed and tap count stay the same across arms.

The backfire this specifically watches for. Honesty alone does not guarantee help. That is why Disclosure Backfire sits in the design: a late-fee warning that reads as an accusation, or that shoves a borrower toward a less transparent product instead of more careful repayment, would be the intervention failing.

Peck and Shu, then Atasoy and Morewedge. Brief physical contact with an object raised how much people valued it and what they would pay. Atasoy and Morewedge ran the mechanism the other way: identical content lost value and willingness to pay once it became a purely digital good. A digital copy generated a weaker sense that it was “mine,” the same ownership pull that sits under the Endowment Effect. Broader case: Printing Costs More Than Digital. The Paper Copy Still Wins.

Lab objects versus a dated renewal mailer. Those studies used everyday consumer goods and generic digital content: mugs, sunglasses, souvenir photos, books, films. Neither tied a recurring benefit mailer to a real, dated renewal decision for a paid product. Whether physical ownership still shows up in renewal months later remains untested.

What the card programme arms compare. A premium card's annual member guide, either as a printed booklet or as the digital-only equivalent already in the app. Renewal at the next annual fee date is the outcome that matters, and the same premium-card context as Credit Card Premium.

Cardholders who receive a printed annual member guide on top of the digital version every cardholder already gets in the app should report higher perceived value of the card at 90 days and renew at a higher rate at their next annual fee date than cardholders who receive the digital-only version.

Illustrative, not real card-programme data. Assume the premium card's current annual renewal rate is 85%, with the digital-only guide as the existing standard.

Cost is not an excuse to thin the control. A physical mailer costs materially more per cardholder than a push notification. Testing it against a cheaper default is fair only when the digital-only arm still gets the full, real content, not a deliberately thinner version built to make print look better.

What both arms receive. Identical content. The only difference is the medium the guide arrives in, and every cardholder still gets the in-app digital version regardless of arm.

The fairness read this watches for. A cardholder who notices peers got a printed booklet and they did not could read the difference as being treated as a lesser member, the unfairness pattern covered under Perceived Fairness. The complaint-rate guardrail below catches that directly.

Benartzi and Thaler's 1/n pattern. Retirement savers roughly split contributions evenly across however many funds their employer's plan listed. The resulting stock/bond mix tracked the mix of funds on offer far more closely than any stated risk preference. That is the core of Diversification Heuristic.

Employer 401(k) menus versus an editable app list. Their evidence came from employer-chosen plan menus, not a screen a product team can redesign and re-test inside its own app. It also predates today's single-screen, pick-as-many-as-you-like investing menus like CommSec Pocket's. Whether trimming a live investing app's option list changes how evenly a new user's first contribution spreads is still open. A shorter default list also brushes against Choice Overload, the companion finding that more options can make choosing itself harder before the 1/n split kicks in.

What the signup arms compare. Replace a flat 10-option “Invest in” menu with three pre-grouped options, each already diversified inside itself. Track how many portfolios a new user selects, and whether the resulting split still tracks pure equal-weighting by count, inside the same signup flow for the same app.

New users shown the trimmed, 3-option menu should end up with a genuinely diversified first portfolio more often than users shown the original 10-option menu. Both groups should still tend to split money close to evenly across however many options they pick. The 10-option group should be more likely to overweight several overlapping single-country or single-theme funds by picking, for example, 4 of the 10 and splitting evenly across them.

Illustrative, not real/disclosed data. Assume 45% of new signups shown the existing 10-option menu end up with a first-session portfolio classified as genuinely diversified. That means spanning at least two distinct, non-overlapping asset classes rather than several near-duplicate single-country or single-theme picks split evenly.

What the menu redesign may not do. Arms may regroup options and change how many appear by default. They may not push a user to invest more than intended or hide a legitimate option's existence. Every option from the original 10-item menu stays reachable through the curated screen, just organised differently. Any “already diversified” curated option must actually be diversified, not marketed that way while concentrated.

What the arms isolate. Whether restructuring the menu reduces the accidental overlap naive equal-weighting produces. Funneling new users into house-preferred, higher-fee products via a shorter, more persuasive menu sits outside the design. See Not Testing Is Still a Bet.

Khan and Dhar's imagined volunteer act. People first imagined committing to volunteer work, then chose between two equally priced products, one practical, one indulgent. Imagining the virtuous act shifted choice toward the indulgent option compared with people given no such prompt. A completed (or even merely imagined) responsible act can license a later indulgence. Full write-up: Licensing Effect.

Earned points versus new money. The reward-redemption experiment already on this site tests the mechanism on a small, already-earned resource: points, choosing between two redemptions of equal value. Nothing here yet tests it on new money entering someone's life, a real lump sum with no assignment yet, and a real choice about whether it goes toward debt.

Why confirmation should matter. If making a virtuous act visible is what does the licensing, two customers whose refund hits their card balance identically should still spend differently afterward. The only difference is whether the app told them so as its own moment.

Customers already enrolled in Refund Auto-Payoff who get an explicit confirmation that their tax refund hit their card balance should show a larger rise in discretionary card spending over the next 30 days. Enrolled customers whose identical refund lands with no distinct confirmation, only a line in ordinary transaction history, should show the smaller rise.

Illustrative, not real product data. Assume 40% of customers whose refund is applied without a distinct confirmation show a rise in discretionary card spend above their own prior 30-day average in the month that follows. That is ordinary post-windfall spending, refund or no confirmation.

Where the bank's incentive warps the find. A confirmation that lifts discretionary spending after a debt paydown can win back, in interest and fees, some of what the customer just saved. That incentive tempts redesigning the confirmation to maximise the effect rather than to inform honestly.

What stays identical across arms. Every enrolled customer's refund hits their balance the same way regardless of arm. Only whether they are told about it, not whether it happens, its size, or its timing, is under test. The confirmation states only true, specific facts: the dollar amount applied and the new balance. No encouragement to spend, celebrate, or reward the moment.

What the arms isolate. Whether making a real debt reduction visible changes later spending. Redesigning the confirmation to push that spending further sits outside the design. See Not Testing Is Still a Bet.

Solnick and Hemenway's two worlds. People chose between earning $50,000 while others earned $25,000, or earning $100,000 while others earned $200,000. Most picked the poorer absolute income as long as it left them ahead of the people around them. That relative ranking is the load-bearing claim behind Positional Concern.

Survey worlds versus real cards. Their design was a hypothetical questionnaire: no real dollars changing hands, no product anyone could walk away with. This site has not yet asked whether the same ranking pull appears when both options are real cashback rates on a card a customer can actually choose.

Why a peer line should move the choice. If relative standing weighs on a live purchase, adding a truthful peer benchmark beside two unchanged rates should shift some people off the higher-paying card and onto the one that merely ranks better against similar spenders. That is a ranking question, distinct from Social Norm, which tracks what most people do rather than whether someone ends up ahead of them.

When two real cashback rates sit side by side on a card-choice screen, adding a truthful benchmark of what similar-spend customers typically earn should raise the share who pick the lower-rate card. The same two rates with no benchmark should leave that share lower.

Illustrative, not real product data. Assume 6% of customers pick the lower-rate card when shown only the two rates, most likely people choosing on a card feature other than cashback, rather than misreading the numbers.

Where the issuer's margin sits. The lower-cashback card earns the issuer more. A rank benchmark that steers people toward it, packaged as helpful context, would cost real customers real dollars for the issuer's gain.

What stays fixed across arms. Both rates are real, unchanged products either arm could choose without this test. The benchmark is an independently computed figure, identical in method for both arms; only whether it appears on screen differs. It is never tuned to manufacture the result.

What the arms isolate. Whether visible relative standing pulls a real choice away from the higher-paying option. Building that pull into how the cards are sold sits outside this design. See Not Testing Is Still a Bet.

John, Blunden and Liu on messengers. People who deliver bad news get rated less likeable, even when a random process produced the outcome rather than the messenger. Signalling a benevolent motive while delivering the news shrank that penalty. The full mechanism write-up is under Shooting the Messenger.

A person versus a template. Their finding came from a human delivering news to another human. A loan decline is usually the reverse: a templated, automated letter with no person in the room. Whether benevolence-framing still softens the hit when the “messenger” is a system-generated notice remains open on this site.

What should transfer to automation. If an honest statement of care softens the penalty for a live messenger, the same idea should travel to an automated decline. Opening an otherwise identical decline with that statement should soften how a declined applicant feels about the bank, while the underlying decision stays identical.

Applicants who receive an identical loan decline, same decision and same reason codes, should report less negative sentiment toward the bank on a short post-decision survey when the decline opens with an explicit, genuine statement of care. Applicants on the current generic decline message should score worse on the same survey.

Illustrative, not real product data. Assume 34% of applicants who receive the current generic decline rate their experience 1 or 2 stars out of 5 on a short post-decision survey.

Where warmer copy can paper over a bad call. Banks gain from fewer complaints and better sentiment scores whether or not the underlying decision was fair. Softer language can make a weak or biased decision feel more acceptable, or discourage someone with a real grievance from escalating it.

What stays identical. The credit decision, the reason codes shown, and the applicant's appeal rights match in both arms. Only the opening framing sentence changes. That sentence states true next steps and real resources, and never implies the decision will change just because someone asks.

What the arms isolate. Whether framing changes how a customer feels about an unchanged decision. Using framing to dampen complaints about decisions that are actually wrong sits outside the design. See Not Testing Is Still a Bet.

Gollwitzer and Brandstätter's if-then plans. Naming exactly when and where you will act on a goal raises follow-through far above holding the same goal with no specific plan. That is the core claim of Implementation Intentions.

What CommBank's flow already says. The live appointment booking path (commbank.com.au/appointment) lets a customer book a home loan appointment by phone or in-branch. Both paths already tell the customer the confirmed time and what to have ready, everything an implementation intention needs, but they state it to the customer rather than asking it back. An in-branch slot also adds a barrier a phone call never has: getting to a specific address on time.

Four prompts, two appointment types. The confirmation screen tries four ways of asking for a plan: a vivid, specific moment; the whole plan restated in the customer's own words; naming what might get in the way and planning around it; and a light, optional suggestion with nothing to fill in. Running the same four against phone and in-branch appointments separates which approach works at all from whether the ranking holds for a remembering problem and a travelling problem alike.

Customers who book a home loan appointment, by phone or in-branch, and then see any of the four confirmation-screen prompts should no-show less often than customers who are only ever told the details. How the four prompts rank against each other may differ by appointment type. A travelling problem could favour a different prompt than a remembering problem, and that ranking difference is part of what the run would reveal.

Illustrative, not real CommBank data. Assume 22% of booked home loan phone appointments end in a no-show under the current reminder-call script.

Illustrative, not real CommBank data. Assume 28% of booked in-branch home loan appointments end in a no-show, higher than the phone rate. Travelling somewhere by a set time is a real added barrier a phone call never has.

Where a plan prompt overreaches. A prompt that assumes a specific time, travel mode, or personal obstacle can feel intrusive if it asks for more than a customer wants to give before they have even arrived. A prompt with no real payoff to the customer can also read as the bank directing their thinking rather than helping them plan something they would benefit from planning.

What stays optional and identical. The appointment, the specialist or branch, and the information required match across all five arms for whichever appointment type is shown. Every confirmation-screen prompt is optional and skippable. Options offered are neutral and non-exhaustive, and none presume which obstacle, travel mode, or moment matters most for a given customer.

What the arms isolate. Whether asking the customer to state their own plan changes follow-through on an identical appointment. Whether the confirmation screen or reminder call should exist at all sits outside this design. See Not Testing Is Still a Bet.

The reminder-fatigue risk. The reminder call itself carries a separate cost this experiment does not measure: repeated contact can lift today's response while raising how many customers opt out of future contact altogether. See Reminder Fatigue.