Behavioural economics frameworks: the science of how customers actually decide

Customers do not decide the way the textbook says

Classical economics pictures a customer who weighs every cost and benefit and then picks the best option. Real people decide with tired brains, limited attention, moods and mental shortcuts, and those shortcuts evolved for survival, not for comparing subscription plans. Behavioural economics studies how people actually decide, not how they ought to.

Design for the imaginary rational customer and you will be puzzled when a carefully reasoned offer fails to convert. Design for the way people really choose and the page works with them.

This article takes the core frameworks in turn and says, for each, what the original research found, how well it has held up, where Irish law draws a line, and how to measure it. Every number in the worked examples is invented so I can show each step.

Two systems: fast intuition and slow deliberation

Daniel Kahneman shared the 2002 Nobel memorial prize in economics for bringing insights from psychology into economics, above all about judgement and decision making under uncertainty.¹ His book Thinking, Fast and Slow made one of his framings famous: two modes of thinking, which he calls System 1 and System 2.²

System 1 is fast, automatic and effortless. It recognises a face, reads a billboard and answers 2 + 2. It works by association, produces feelings and first impressions, and is confident even when it is wrong.

System 2 is slow and effortful. It multiplies 17 by 24, compares two products on six attributes and fills in a tax return. It can check System 1, but it does one demanding thing at a time, and distraction or time pressure knocks it out.

Kahneman is clear that the two systems are a useful picture, not two places in the brain.² One popular addition deserves a warning. The idea that effortful thinking draws on a limited reserve of willpower that runs down ("ego depletion") did not survive a large test: a preregistered replication across 23 laboratories and 2,141 participants found an effect of $d = 0.04$, with a 95% interval from $-0.07$ to $0.15$, which includes zero.³ So "System 2 is lazy" is a fair description of how people behave, but not proof that every form field drains a battery.

Why it matters in a shop

Most everyday purchases are made with System 1: familiar, quick, driven by impression. System 2 joins in under particular conditions:

• High stakes, where a mistake is costly.
• Something feels off, such as an unexpected charge or a page that looks wrong.
• Explicit prompting, when someone or something forces a comparison.
• Unfamiliarity, when nothing about the situation matches a known pattern.

So make the System 1 path easy (clear layout, plain words, an obvious next step) and keep the System 2 information ready for when it is called up.

Even a considered purchase starts with a System 1 impression. Social proof is one of the shortcuts System 1 reaches for when it is unsure.

An example: Inistioge Tea Co. (inistiogetea.example), an invented shop in Kilkenny, shows a €24 tea hamper and adds €6.95 delivery on the last checkout step. The surprise wakes System 2, which starts comparing, and the comparison often ends with the tab closed. Showing "€24, delivery €6.95" on the product page keeps the decision on the fast path, and Irish law expects an invitation to purchase to give the price inclusive of taxes and any delivery charges anyway.⁴

Cognitive biases are shortcuts, not bugs

A cognitive bias is a systematic way in which judgement departs from the rational benchmark. Amos Tversky and Kahneman's 1974 paper showed that people rely on a few heuristics (rules of thumb) that are usually useful but lead to predictable errors.⁵ They are features of a mind built for fast, good enough decisions. Here are the ones that matter most for buying.

Confirmation bias

People seek, interpret and remember information in ways that fit what they already believe. Raymond Nickerson's review calls it a ubiquitous phenomenon in many guises.⁶ A customer who likes your brand reads an ambiguous delivery delay generously, one who distrusts you reads it as proof. An early, unambiguous win for a new customer ("your order is packed and leaves today") gives the bias something positive to confirm.

The availability heuristic

People judge how likely something is by how easily examples come to mind.⁵ One vivid complaint can outweigh a thousand quiet successes. Make good experiences concrete and easy to recall, and answer a public complaint quickly, before it becomes the example everyone remembers.

The peak end rule, not recency bias

This is often filed under "recency bias", but the research is more specific. In a 1993 experiment, Kahneman and colleagues had people hold a hand in painfully cold water at 14 °C for 60 seconds, and in another trial for 60 seconds at 14 °C followed by 30 seconds while the water warmed slightly to about 15 °C. Asked which trial to repeat, 22 of 32 participants, or 69%, chose the longer one, which contained all the pain of the short one plus more.⁷ People judged the episode by its worst moment and its end, and barely by its length.

An example: Dungarvan Oak (dungarvanoak.example), an invented furniture shop in Dungarvan, Co. Waterford, has an awkward checkout but a delivery team that assembles the table, takes the packaging away and leaves a handwritten care card. By the peak end rule, customers remember the assembly and the card more than the clumsy form. That is not an excuse to keep the bad form. It says where the last euro of effort pays most: the end of the experience. The conversion and UX psychology article goes into this further.

The sunk cost fallacy

People keep investing in something because of what they have already put in, not because of what it will return. Hal Arkes and Catherine Blumer showed it with theatre season tickets: people who had paid full price went to more plays in the first half of the season than people who had been given a discount at random.⁸ Users who have invested data, settings and effort are reluctant to leave, which is part of why onboarding that asks for effort can raise retention. The honest version makes that investment worth something to the customer. The other version builds switching costs and nothing else.

Status quo bias and default bias

People prefer the current state of affairs and tend to accept whatever is preselected. Both get their own sections below, because they are the strongest effects in this article.

Choice architecture: there is no neutral page

Richard Thaler, who received the 2017 Nobel memorial prize, coined the term "nudge" and, with Cass Sunstein, popularised choice architecture: organising the context in which people choose.⁹¹⁰ The UK's Competition and Markets Authority (CMA) calls the people who design a website choice architects, with examples such as the order of search results, the number of steps needed to cancel and whether an option is preselected.¹¹

Every page has architecture, whether anyone designed it on purpose or not. The CMA calls choice architecture a neutral term: it can help people choose well, or it can hide information and set defaults that do not match their preferences.¹¹

The elements are these:

• Defaults: what is preselected. The strongest lever, covered next.
• Order: where an option sits. Middle options in a line of three look like the safe compromise, a pattern Itamar Simonson documented as the compromise effect.¹² A clearly worse option can make its neighbour look better (the decoy effect), though that one is contested, as the pricing psychology article explains with worked examples.
• Framing: the same fact presented differently. "90% fat free" and "10% fat" describe the same yoghurt. The CMA notes significant evidence that framing, highlighting some aspects over others, can affect decisions.¹¹
• Friction: the effort each step costs, covered below.
• Feedback: progress indicators, confirmations and undo, which help people finish what they meant to do.

Defaults are the strongest lever

A default is the option people get if they do nothing. The CMA describes defaults as one of the strongest and most reliable influences on consumer behaviour, and cites a meta analysis of 58 studies which found that a preselected option is on average 27% more likely to be chosen, out of two options, than when there is no default.¹¹ It gives three reasons: sticking with the default takes less effort, a default can read as a recommendation, and people treat the default as a reference point, as if they already owned it.

The pension study

Brigitte Madrian and Dennis Shea studied a large US company that switched its retirement savings plan from opt in to automatic enrolment, with nothing else about the plan changed. Among employees hired just before the switch, participation at three to 15 months of tenure was 37%. Among those hired under automatic enrolment it was 86%. And 61% of the automatically enrolled kept both the default contribution rate of 3% and the default fund, a money market fund few would have picked actively.¹³ The authors read this as inertia plus employees treating the default as advice.

Britain ran the same experiment at national scale. Automatic enrolment into UK workplace pensions began in 2012. Participation among eligible employees there fell to a low of 55% in 2012 before it took effect, reached 88% in 2019,¹⁴ and stood at 89% in 2024.¹⁵ Ireland started its own version on 1 January 2026: My Future Fund signs up employees aged 23 to 60 who earn €20,000 or more a year and are not already paying into a work pension, and they can opt out after a minimum period.¹⁶ It is too new for Irish participation figures, so the British numbers are the evidence for now.

Organ donation, and what the famous chart really shows

The most quoted default chart compares organ donor rates in opt in and opt out countries. It is often read as showing that opt out countries have near universal donation. That is not what it shows. Eric Johnson and Daniel Goldstein measured the effective consent rate: the share who had opted in, in opt in countries, or had not opted out, in opt out countries. The four opt in countries and the six opt out countries did not overlap at all, and nearly 60 percentage points separated them.¹⁷ But consent is not donation. When they looked at actual donations from 1991 to 2001, controlling for other differences between countries, the default raised donation from 14.1 to 16.4 donors per million, a 16.3% increase.¹⁷ Real, and much smaller than the consent chart suggests. Ireland has since moved to an opt out system too: under the Human Tissue (Transplantation, Post Mortem, Anatomical Examination and Public Display) Act 2024, adults are deemed to consent to donating relevant organs unless they register an objection on the Relevant Organ Donation Opt Out Register,¹⁸ and those sections came into force on 17 June 2025.¹⁹

The lesson for a shop: a default can move what people tick enormously, while moving the outcome you care about by much less. Measure the outcome.

Worked example: a preselected plan at a Galway booking app

Bearna Bookings (bearnabookings.example), an invented Galway company, sells booking software to salons on three plans: Basic at €29, Pro at €79 and Enterprise at €199 a month. Without any preselection, assume 35% of new customers choose Basic, 45% Pro and 20% Enterprise. Assume a preselected plan is 27% more likely to be chosen, in line with the CMA's summary figure,¹¹ and that the extra share comes proportionally from the other plans.

With Pro preselected, Pro's weight becomes $0.45 \times 1.27 = 0.5715$. The weights now sum to $0.35 + 0.5715 + 0.20 = 1.1215$, so the shares are Basic $0.35 / 1.1215 = 31.2\%$, Pro $0.5715 / 1.1215 = 51.0\%$ and Enterprise $0.20 / 1.1215 = 17.8\%$.

For 5,000 new customers a month, expected monthly revenue is a weighted sum:

$$R = N \sum_{i} s_i \, p_i$$

Here $R$ is expected monthly revenue, $N$ the number of new customers, $s_i$ the share choosing plan $i$ and $p_i$ that plan's monthly price. Foundations: expected value explains the weighted sum.

Preselected	Basic	Pro	Enterprise	Monthly revenue
Nothing	35.0%	45.0%	20.0%	€427,500
Basic	40.6%	41.1%	18.3%	€403,109
Pro	31.2%	51.0%	17.8%	€423,979
Enterprise	33.2%	42.7%	24.1%	€456,575

The surprise is the Pro row: preselecting Pro wins Pro the most customers but earns less than no default, because it pulls people down from Enterprise as well as up from Basic. Preselecting Enterprise earns most here, and is also the default most likely to put a small salon on a plan it does not need, which returns later as downgrades and complaints. The right default is the plan most customers would pick if they thought about it. Test it, and follow the customers for months, not days.

# Bearna Bookings, invented: expected shares and revenue by preselected plan
base = {'basic': 0.35, 'pro': 0.45, 'enterprise': 0.20}   # assumed preferences
prices = {'basic': 29, 'pro': 79, 'enterprise': 199}
boost = 1.27        # assumption, in line with the CMA's 27% summary figure
customers = 5000

def shares(default=None):
    weights = {k: v * (boost if k == default else 1) for k, v in base.items()}
    total = sum(weights.values())
    return {k: w / total for k, w in weights.items()}

for default in [None, 'basic', 'pro', 'enterprise']:
    s = shares(default)
    revenue = customers * sum(s[k] * prices[k] for k in s)
    print(f"{str(default):10}", {k: f"{v:.1%}" for k, v in s.items()}, f"€{revenue:,.0f}")

Where Irish law takes the default away

Two defaults are simply not yours to set. Consent to marketing must be an active opt in: it needs a clear affirmative act under the GDPR and, for marketing emails and texts, under regulation 13 of the ePrivacy Regulations 2011,²⁰ and the Data Protection Commission (DPC) spells out that silence, pre ticked boxes or inactivity are not consent.²¹ And an extra paid item cannot be added by default: under section 123 of the Consumer Rights Act 2022, a trader needs the customer's express consent for any payment beyond the main price, and consent inferred from a default option the customer has to reject does not count. If the trader takes the payment anyway, it has to reimburse it, and breaking the rule is an offence.²²

The flow reads top to bottom. The first question is the legal gate: consent and paid extras are never preselected. Everything else must match what customers would choose for themselves, confirmed by a test. The point to take away is the last box: a default that wins on ticks and loses on refunds has not won.

Friction cuts both ways, and the law limits one way

Every extra step, field or click loses some people. Removing friction from the paths customers want (guest checkout, saved payment details, prefilled addresses) helps them and you. Adding friction to the paths you would rather they did not take is a different matter. The CMA calls excessive or unjustified friction that makes it hard for people to get what they want "sludge", and lists barriers to cancellation as the example.¹¹ Irish law now sets a floor for the online exit too. Since S.I. No. 309 of 2026 inserted section 115A into the Consumer Rights Act 2022, a trader whose customer concluded a distance contract online has to offer an online cancellation function, labelled "cancel contract here" or something as clear, prominently displayed and available throughout the cancellation period, followed by a "confirm cancellation" step and an acknowledgement on a durable medium.²³ So no "call to cancel" and no survey walls. Build the exit to be easy, and earn the stay with an honest offer on the way out.

Worked example: checkout steps at a Cork print shop

Barrack Street Prints (barrackstprints.example) is an invented shop in Cork. Suppose, as this example's assumptions, that 85% of people who start the checkout would complete it if it had no steps at all (a notional baseline), and that each step loses 15% of whoever is left. Completion after $n$ steps then shrinks like compound growth in reverse:

$$c_n = c_0 \, (1 - f)^n$$

Here $c_n$ is the share who complete a checkout of $n$ steps, $c_0$ the share who would complete with no steps at all, $f$ the share lost at each step and $n$ the number of steps. Foundations: compounding shows why repeated percentages multiply rather than add.

With $c_0 = 0.85$ and $f = 0.15$: one step gives $0.85 \times 0.85 = 72.2\%$, three steps $0.85 \times 0.85^3 = 52.2\%$, five steps $37.7\%$ and eight steps $23.2\%$. Real steps differ: an expected address step costs little, a forced account creation a lot. The model's use is to make you put a number on each step and then measure it in your own funnel.

# Barrack Street Prints, invented: completion by number of checkout steps
c0, f = 0.85, 0.15     # assumptions of the example, not measurements
for steps in [1, 3, 5, 8]:
    print(f"{steps} steps: {c0 * (1 - f) ** steps:.1%} complete")

Hyperbolic discounting: now counts for far more than later

Standard economics assumes exponential discounting: a future reward loses a fixed share of its value for every period of delay.

$$V = A \, e^{-rD}$$

Here $V$ is what the reward is worth today, $A$ the reward, $r$ the discount rate per day and $D$ the delay in days. Foundations: discounting covers present value.

People behave more as if they discount hyperbolically: a delay right now costs a lot, the same delay far in the future costs almost nothing. One common form is:

$$V = \frac{A}{1 + kD}$$

Here $V$, $A$ and $D$ are as before and $k$ is how steeply the person discounts. The same Foundations: discounting section explains why a value today and a value later differ. €100 today against €100 tomorrow feels like a real choice. €100 in 365 days against €100 in 366 days barely registers. Richard Thaler reported early evidence of this kind of inconsistency in 1981.²⁴

How well it holds up

The pattern is real, but its size is not a constant. A review of three decades of research by Shane Frederick, George Loewenstein and Ted O'Donoghue discusses the "spectacular variation" in implied discount rates across studies.²⁵ And a careful experiment by Ned Augenblick, Muriel Niederle and Charles Sprenger found very limited present bias in choices about money, and considerably more in choices about effort.²⁶ For a shop, that suggests the pull of "now" is strongest for effort and waiting (setting up, filling in, waiting for delivery) and weaker for pure money comparisons than the folklore says. Popular rules such as "a free trial always beats an annual discount" have no solid evidence behind them that I could find. Test them.

Worked example: five offers through two discount curves

An invented Limerick coffee roaster, Thomondgate Roasts (thomondgateroasts.example), considers five loyalty offers. Assume a steep hyperbolic customer with $k = 0.05$ per day and an exponential benchmark with $r = 0.0003$ per day.

For €100 off in 12 months, $D = 365$. Hyperbolic: $100 / (1 + 0.05 \times 365) = 100 / 19.25 = €5.19$. Exponential: $100 \times e^{-0.0003 \times 365} = 100 \times e^{-0.1095} = €89.63$.

Offer	Face value	Hyperbolic value today	Exponential value today
€10 off now	€10	€10.00	€10.00
€15 off in 30 days	€15	€6.00	€14.87
€25 off in 90 days	€25	€4.55	€24.33
€50 off in 180 days	€50	€5.00	€47.37
€100 off in 12 months	€100	€5.19	€89.63

To this assumed customer, €10 now beats €100 in a year. To the benchmark it is worth a tenth as much. Real customers sit somewhere between, which is the variation the review describes. So lead with what the customer gets now (instant access, delivery tomorrow, visible progress), and test a deferred reward against an immediate one before assuming either wins.

import math

# Thomondgate Roasts, invented offers; k and r are assumptions of the example
def hyperbolic(amount, days, k=0.05):
    return amount / (1 + k * days)

def exponential(amount, days, r=0.0003):
    return amount * math.exp(-r * days)

offers = [(10, 0), (15, 30), (25, 90), (50, 180), (100, 365)]
for amount, days in offers:
    print(f"€{amount} in {days} days: hyperbolic €{hyperbolic(amount, days):.2f}, "
          f"exponential €{exponential(amount, days):.2f}")

Mental accounting: money in labelled jars

A euro is a euro, but people do not treat it that way. Thaler developed the theory of mental accounting: people simplify money decisions by keeping separate accounts in their heads and judging each decision by its narrow effect on one account, not its effect overall.⁹²⁷ A windfall sits in a different jar from wages. A gift card sits in a jar labelled "spend at this shop". Paying by card and paying in cash feel different, and Drazen Prelec and Duncan Simester found people willing to pay more when told to pay by card, by up to 100% in their studies.²⁸ It is a small set of studies, so treat that size as an upper bound.

The theory predicts that gift cards, store credit and points are spent more freely than cash, because they already sit in a spending jar. Plausible, but not a measured constant, so the example below states its figures as assumptions.

There is also a legal line. Online shoppers in Ireland can cancel a sales contract within a cancellation period that normally ends 14 days after the goods arrive,²⁹ and when they do, the Consumer Rights Act 2022 requires the trader to reimburse them using the same means of payment they used, unless they have expressly agreed otherwise.³⁰ So "store credit instead of a refund" cannot be the default. Store credit can only be an option the customer chooses, which is why it needs to be the better deal for them.

Worked example: refund or bonus credit at a Galway knitwear shop

Quay Street Knitwear (quaystreetknit.example), an invented Galway shop, handles 1,000 returns a month with an average value of €75. A refund to the original card is the default. It adds an option: take store credit worth 110% of the order, so €82.50. Suppose, as the example's assumptions, that 30% of customers choose the credit, that 85% of credit is eventually spent, and that 5% of refunded customers come back and spend their refund.

1. Refunds: $700 \times 75 = €52{,}500$ goes back to cards. At 5% coming back, that is $52{,}500 \times 0.05 = €2{,}625$ of later sales.
2. Credit: $300 \times 75 = €22{,}500$ stays in the shop. The credit issued is $300 \times 82.50 = €24{,}750$, of which 85% is spent: $€21{,}037.50$ of goods at retail price.
3. The bonus costs $300 \times 7.50 = €2{,}250$ in extra goods at retail price, and less at cost.

Whether this pays depends on the margin. With a high gross margin the bonus goods cost the shop little, on a thin margin it may be paying people to keep its stock. Run the numbers on margin, not revenue.

Choice overload: the jam study and what happened next

The most famous experiment in this field took place in Draeger's, an upmarket supermarket in Menlo Park, California. Sheena Iyengar and Mark Lepper set up a tasting booth with either 24 or six jams. Of the 242 shoppers who passed the large display, 60% stopped. At the small display, 40% stopped. Then the reversal: 30% of those who stopped at the six jam display bought a jar, against 3% at the 24 jam display.³¹

The 3% is often quoted as the conversion of the large display. It is 3% of those who stopped, not of everyone who walked past. Per passer by, the six jam display sold to $0.40 \times 0.30 = 12\%$ and the 24 jam display to $0.60 \times 0.03 = 1.8\%$. Still a large gap, about six and a half times, from one tasting booth in one shop.

Worked example: was the jam difference more than chance?

The published counts are 31 buyers out of 104 who stopped at the small display and four out of 145 at the large one. The two proportion z test compares them:

$$z = \frac{p_1 - p_2}{\sqrt{p(1-p)\left(\frac{1}{n_1} + \frac{1}{n_2}\right)}}$$

Here $p_1$ and $p_2$ are the shares who bought at each display, $n_1$ and $n_2$ the numbers who stopped, and $p$ the pooled share across both. Foundations: hypothesis tests walks through it.

$p_1 = 31/104 = 0.298$, $p_2 = 4/145 = 0.028$, and the pooled $p = 35/249 = 0.141$. The standard error is $\sqrt{0.141 \times 0.859 \times (1/104 + 1/145)} = 0.0447$, so $z = 0.270 / 0.0447 = 6.06$. The paper reports a chi square of 32.34 with $p < 0.0001$. Within one study, the difference is not luck.

import math

# Iyengar and Lepper (2000), Study 1: buyers among those who stopped
buy_small, stop_small = 31, 104
buy_large, stop_large = 4, 145

p1, p2 = buy_small / stop_small, buy_large / stop_large
p = (buy_small + buy_large) / (stop_small + stop_large)
se = math.sqrt(p * (1 - p) * (1 / stop_small + 1 / stop_large))
z = (p1 - p2) / se
print(f"small {p1:.1%}, large {p2:.1%}, z = {z:.2f}")

The replication problem

Within one study is the catch. Benjamin Scheibehenne, Rainer Greifeneder and Peter Todd pooled 63 conditions from 50 experiments, 5,036 participants in all, and found a mean choice overload effect of virtually zero, with large variation between studies.³² Many experiments found no overload, and some found that more choice helped. A later meta analysis by Alexander Chernev, Ulf Böckenholt and Joseph Goodman pooled 99 observations from 7,202 participants and found the effect is real under specific conditions. It grows with a complex choice set, a difficult decision task, uncertain preferences and a goal of minimising effort.³³

So the honest summary is not "fewer options always sell more". It is: when the options are hard to compare and the customer does not know what they want, a long list can stall them. Simple choices with clear criteria (a size, a colour) can carry many options. Three plans is a reasonable starting point to test, not a law.

Where a long list does stall people, there are gentler fixes than cutting the range: drop variants nobody buys, show the core options first and the rest on request, ask a few guiding questions, label a genuinely popular option as such, and make the remaining options easy to compare.

Status quo bias: inertia keeps customers, and triggers wake them

William Samuelson and Richard Zeckhauser showed that people disproportionately stick with the current option, in experiments and in real choices such as university staff picking health and retirement plans.³⁴ Several forces hold the status quo in place. Changing risks a loss, and losses weigh roughly twice as much as gains: a 2024 meta analysis of 607 estimates put the average loss aversion coefficient at about 1.96, not the 2.25 often quoted.³⁵ Staying needs no evaluation. A bad switch causes more regret than a missed one. And people value what they already have more, the endowment effect.

What it means for retention

Most subscribers are not evaluating alternatives. They stay until something prompts them to reconsider: a price rise, a service failure, a change in their life, a renewal notice. Once prompted, they compare properly. Retention is partly about avoiding needless triggers (a confusing bill, a surprise charge) and handling the necessary ones well (a price change explained early, with the value it pays for).

That is not the same as hiding the triggers a customer ought to see. A clear notice before a renewal payment is the honest version, and a customer who stays only because they never noticed is a complaint waiting to happen.

Worked example: a price rise at a Limerick coffee subscription

Thomondgate Roasts also sells a coffee subscription. Assume, for this example, a base churn of 3% of remaining subscribers a month. A price rise in month 12 is assumed to multiply churn by 3.5 that month and by two the month after. A renewal notice in month 12 is assumed to multiply churn by 2.5 for that month.

Each month the remaining share falls by the month's churn: $S_m = S_{m-1}\,(1 - c\,t_m)$, where $S_m$ is the share remaining after month $m$, $c$ the base monthly churn and $t_m$ the trigger multiplier in month $m$ (1 when nothing happens). It is the same compounding as the checkout example, see Foundations: compounding.

With no trigger: $0.97^{12} = 69.4\%$ remain after a year and $0.97^{24} = 48.1\%$ after two. With the price rise: 64.0% after a year and 43.0% after two. With the renewal notice: 66.2% and 45.9%.

The horizontal axis shows the three scenarios, the vertical axis the share of starting subscribers still paying after two years. One month of heightened churn costs two to five points two years later. Whether the price rise pays is a separate sum: 43.0% of subscribers paying more may beat 48.1% paying less. Do that sum, and never solve it by suppressing the notice.

# Thomondgate Roasts, invented: monthly churn with assumed trigger multipliers
def remaining(months, base_churn=0.03, triggers=None):
    triggers = triggers or {}
    share = 1.0
    for m in range(1, months + 1):
        share *= 1 - base_churn * triggers.get(m, 1.0)
    return share

scenarios = {'No trigger': {}, 'Price rise M12': {12: 3.5, 13: 2.0}, 'Renewal notice M12': {12: 2.5}}
for name, triggers in scenarios.items():
    print(f"{name:20} 12 months {remaining(12, triggers=triggers):.1%}, "
          f"24 months {remaining(24, triggers=triggers):.1%}")

What it means for acquisition

The same bias works against the newcomer: a prospect's status quo is a competitor or doing nothing. Show the cost of staying with honest comparisons, make switching easy with data import and help moving over, offer a trial, and be present at natural moments of change, such as a growing team or a contract ending.

A behavioural audit you can run on your own product

The frameworks turn into a short list of questions. The code below prints the ones worth asking first.

def behavioural_audit(product):
    actions = []
    if product.get('considered_purchase'):
        actions.append('Support System 2: specs, comparisons, references, and a clear first impression')
    else:
        actions.append('Keep the System 1 path short: full price early, few steps, no surprises')
    actions.append('List every default: is each one what customers would choose if asked?')
    actions.append('Consent and paid extras: unticked, active choice only')
    if product.get('subscription'):
        actions.append('Map churn triggers, explain price changes early, keep the exit straightforward')
    if product.get('plans', 0) > 3 or product.get('options', 0) > 6:
        actions.append('Test fewer options, progressive disclosure or guided selling')
    if product.get('returns'):
        actions.append('Refund by the original method, offer store credit only as a better optional deal')
    actions.append('Measure outcomes (refunds, complaints, retention), not just clicks')
    return actions

# An invented SaaS product: considered purchase, subscription, four plans
for action in behavioural_audit({'considered_purchase': True, 'subscription': True, 'plans': 4}):
    print('-', action)

If I had to order the work, defaults come first: they are the strongest lever and usually the least effort to change. Then the churn triggers, then the framing of what the customer gets now, then the number of options.

Design for people, honestly

Behavioural economics is not a box of tricks for manipulating customers. It is a description of how people actually decide, and you are applying it whether you mean to or not, because every page has a default, an order and a level of friction. The question is only whether you choose them on purpose, in a way that serves the customer as well as the business, and whether you measure what they do afterwards.

I have been applying behavioural economics to product design and marketing for over a decade. The frameworks come from the research cited above. The code runs as it stands on invented numbers, so adapt it to your own data before you trust any result, and the statistical significance article covers how to tell a real lift from noise.

Footnotes

1. The Nobel Prize, The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2002, summary, https://www.nobelprize.org/prizes/economic-sciences/2002/summary/

2. Daniel Kahneman, Thinking, Fast and Slow, Penguin, https://www.penguin.co.uk/books/56314/thinking-fast-and-slow-by-daniel-kahneman/9780141033570

3. Martin S. Hagger, Nikos L. D. Chatzisarantis and colleagues, A multilab preregistered replication of the ego depletion effect, Perspectives on Psychological Science 11(4), 546 to 573, 2016, https://research.tilburguniversity.edu/en/publications/a-multi-lab-pre-registered-replication-of-the-ego-depletion-effec/

4. Law Reform Commission, Consumer Protection Act 2007, revised, section 46 (Misleading: withholding, omitting or concealing material information), subsection (3)(c) and (d), updated to 24 November 2025, Revised Acts, https://revisedacts.lawreform.ie/eli/2007/act/19/revised/en/html

5. Amos Tversky and Daniel Kahneman, Judgment under uncertainty: heuristics and biases, Science 185(4157), 1124 to 1131, 1974, https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf

6. Raymond S. Nickerson, Confirmation bias: a ubiquitous phenomenon in many guises, Review of General Psychology 2(2), 175 to 220, 1998, https://pages.ucsd.edu/~cmckenzie/nickersonConfirmationBias.pdf

7. Daniel Kahneman, Barbara L. Fredrickson, Charles A. Schreiber and Donald A. Redelmeier, When more pain is preferred to less: adding a better end, Psychological Science 4(6), 401 to 405, 1993, https://ius.uzh.ch/dam/jcr:5ae9adc9-61ec-4174-b37c-4b752f36c23b/Kahnemann%20et%20al.%20-%20When%20More%20Pain%20is%20Preferred%20to%20Less%20%281993%29.pdf

8. Hal R. Arkes and Catherine Blumer, The psychology of sunk cost, Organizational Behavior and Human Decision Processes 35(1), 124 to 140, 1985, https://econpapers.repec.org/article/eeejobhdp/v_3a35_3ay_3a1985_3ai_3a1_3ap_3a124-140.htm

9. The Nobel Prize, The Prize in Economic Sciences 2017, press release on Richard H. Thaler, https://www.nobelprize.org/prizes/economic-sciences/2017/press-release/

10. Richard H. Thaler and Cass R. Sunstein, Nudge: the final edition, Penguin, https://www.penguin.co.uk/books/310797/nudge-by-thaler-richard-h-and-sunstein-cass-r/9780141999937

11. Competition and Markets Authority (United Kingdom), Online choice architecture: how digital design can harm competition and consumers, discussion paper CMA155, April 2022, paragraphs 1 to 3 and 4.27 to 4.28, https://assets.publishing.service.gov.uk/media/624c27c68fa8f527710aaf58/Online_choice_architecture_discussion_paper.pdf

12. Itamar Simonson, Choice based on reasons: the case of attraction and compromise effects, Journal of Consumer Research 16(2), 158 to 174, 1989, https://econpapers.repec.org/article/oupjconrs/v_3a16_3ay_3a1989_3ai_3a2_3ap_3a158-74.htm

13. Brigitte C. Madrian and Dennis F. Shea, The power of suggestion: inertia in 401(k) participation and Savings Behavior, NBER Working Paper 7682, 2000 (published in the Quarterly Journal of Economics 116(4), 2001), https://www.nber.org/system/files/working_papers/w7682/w7682.pdf

14. Department for Work and Pensions, Workplace pension participation and saving trends of eligible employees: 2009 to 2019, GOV.UK, 2020, https://assets.publishing.service.gov.uk/media/5ee9d19cd3bf7f1eb4a1b56a/workplace-pension-participation-and-saving-trends-2009-2019.pdf

15. Department for Work and Pensions, Workplace pension participation and savings trends of eligible employees: 2009 to 2024, GOV.UK, 31 July 2025, https://www.gov.uk/government/statistics/workplace-pension-participation-and-savings-trends-2009-to-2024/workplace-pension-participation-and-savings-trends-of-eligible-employees-2009-to-2024

16. Competition and Consumer Protection Commission, Auto enrolment, CCPC, accessed November 2026, https://www.ccpc.ie/manage-your-money/planning-for-the-future/pensions/auto-enrolment

17. Eric J. Johnson and Daniel G. Goldstein, Defaults and donation decisions, Transplantation 78(12), 2004, https://hods.org/pdf/Defaults%20and%20Donation%20Decisions.pdf

18. Government of Ireland, Human Tissue (Transplantation, Post Mortem, Anatomical Examination and Public Display) Act 2024, sections 18 (Deemed consent for donation of relevant organs) and 32 (Relevant Organ Donation Opt Out Register), Irish Statute Book, 2024, https://www.irishstatutebook.ie/eli/2024/act/5/enacted/en/html

19. Irish Statute Book, Human Tissue (Transplantation, Post Mortem, Anatomical Examination and Public Display) Act 2024, commencement information (sections 1 to 36 commenced 17 June 2025 by S.I. No. 250 of 2025), https://www.irishstatutebook.ie/eli/isbc/2024_5.html

20. Government of Ireland, S.I. No. 336 of 2011, European Communities (Electronic Communications Networks and Services) (Privacy and Electronic Communications) Regulations 2011, regulation 13, Irish Statute Book, 2011, https://www.irishstatutebook.ie/eli/2011/si/336/made/en/print

21. Data Protection Commission, FAQ on Consent for Electronic Direct Marketing, April 2020, dataprotection.ie, https://www.dataprotection.ie/sites/default/files/uploads/2026-05/FAQ_Consent_Electronic_Direct_Marketing_April2020.pdf

22. Government of Ireland, Consumer Rights Act 2022, section 123 (Additional payments), Irish Statute Book, 2022, https://www.irishstatutebook.ie/eli/2022/act/37/section/123/enacted/en/html

23. Government of Ireland, S.I. No. 309 of 2026, European Union (Distance Contracts for Financial Services) Regulations 2026, regulation 6 (new section 115A of the Consumer Rights Act 2022, online cancellation function), Irish Statute Book, 2026, https://www.irishstatutebook.ie/eli/2026/si/309/made/en/html

24. Richard Thaler, Some empirical evidence on dynamic inconsistency, Economics Letters 8(3), 201 to 207, 1981, https://econpapers.repec.org/article/eeeecolet/v_3a8_3ay_3a1981_3ai_3a3_3ap_3a201-207.htm

25. Shane Frederick, George Loewenstein and Ted O'Donoghue, Time discounting and time preference: a critical review, Journal of Economic Literature 40(2), 351 to 401, 2002, https://www.aeaweb.org/articles?id=10.1257/002205102320161311

26. Ned Augenblick, Muriel Niederle and Charles Sprenger, Working over time: dynamic inconsistency in real effort tasks, Quarterly Journal of Economics 130(3), 1067 to 1115, 2015, https://econpapers.repec.org/RePEc:oup:qjecon:v:130:y:2015:i:3:p:1067-1115

27. Richard Thaler, Mental accounting and consumer choice, Marketing Science 4(3), 199 to 214, 1985, https://econpapers.repec.org/article/inmormksc/v_3a4_3ay_3a1985_3ai_3a3_3ap_3a199-214.htm

28. Drazen Prelec and Duncan Simester, Always leave home without it: a further investigation of the credit card effect on willingness to pay, Marketing Letters 12, 5 to 12, 2001, https://link.springer.com/article/10.1023/A:1008196717017

29. Government of Ireland, Consumer Rights Act 2022, section 113 (Duration of cancellation period), subsection (4), Irish Statute Book, 2022, https://www.irishstatutebook.ie/eli/2022/act/37/section/113/enacted/en/html

30. Government of Ireland, Consumer Rights Act 2022, section 117 (Obligations of trader in event of cancellation), subsection (5), Irish Statute Book, 2022, https://www.irishstatutebook.ie/eli/2022/act/37/section/117/enacted/en/html

31. Sheena S. Iyengar and Mark R. Lepper, When choice is demotivating: can one desire too much of a good thing?, Journal of Personality and Social Psychology 79(6), 995 to 1006, 2000, https://faculty.washington.edu/jdb/345/345%20Articles/Iyengar%20%26%20Lepper%20%282000%29.pdf

32. Benjamin Scheibehenne, Rainer Greifeneder and Peter M. Todd, Can there ever be too many options? A meta analytic review of choice overload, Journal of Consumer Research 37(3), 409 to 425, 2010, https://econpapers.repec.org/article/oupjconrs/v_3a37_3ay_3a2010_3ai_3a3_3ap_3a409-425.htm

33. Alexander Chernev, Ulf Böckenholt and Joseph Goodman, When product assortment leads to choice overload: a conceptual review and meta analysis, Journal of Consumer Psychology, 2015, Kellogg School of Management, https://www.kellogg.northwestern.edu/faculty/research/detail/2015/when-product-assortment-leads-to-choice-overload-a-conceptual

34. William Samuelson and Richard Zeckhauser, Status quo bias in decision making, Journal of Risk and Uncertainty 1(1), 7 to 59, 1988, https://econpapers.repec.org/RePEc:kap:jrisku:v:1:y:1988:i:1:p:7-59

35. Alexander L. Brown, Taisuke Imai, Ferdinand M. Vieider and Colin F. Camerer, Meta analysis of empirical estimates of loss aversion, Journal of Economic Literature 62(2), 485 to 516, 2024, https://www.aeaweb.org/articles?id=10.1257/jel.20221698