Book a call

Daniel & Friends / Blog / Behavioural science

Behavioural science in marketing: between Kahneman and practice

Kahneman's Thinking, Fast and Slow turned behavioural economics into pop culture and biases into a marketing incantation. Then came the replication crisis and part of that structure collapsed. What is left standing is more valuable in practice than the entire original catalogue of tricks.

Daniel Votruba · 6 July 2026 · 3 min read

Two systems, one buying reality

Daniel Kahneman described decision-making as the interplay of two modes. System 1 is fast, intuitive, associative. System 2 is slow, analytical and reluctant to engage. The overwhelming majority of buying decisions, including rational B2B ones, run on System 1: by recognisability, feeling and mental shortcuts. Marketing that assumes a careful reader of spreadsheets is assuming a customer who does not exist.

Heuristics you meet daily in practice

  • Anchoring (Tversky & Kahneman, 1974): the first number a person sees calibrates every number after it. Which is why price lists start with the most expensive tier and the discount glows next to the original price.
  • Social proof (Cialdini): we resolve uncertainty by imitation. Reviews, customer counts, a best seller label. It works as long as the numbers are true and relevant to the segment.
  • Loss aversion (prospect theory, Kahneman & Tversky, 1979): a loss hurts more than a comparable gain pleases. Hence the power of trial periods, because giving up something I already have is harder than not acquiring it in the first place.
  • The default effect: the preselected option wins. The classic work by Johnson and Goldstein (Science, 2003) on organ donation showed vast differences between opt-in and opt-out countries.

In pricing and e-commerce this adds up to choice architecture: the order of tiers, the highlighted recommended middle plan, the preselected annual payment, cutting sixteen options down to three. None of it changes the product; what changes is how easily the customer reaches a decision.

The replication crisis: a stocktake ten years on

Honesty requires admitting what failed to replicate. Social priming, effects along the lines of words about old age slowing your walk, largely fell apart; Kahneman himself flagged the field as at risk in an open letter in 2012. A meta-analysis of nudges (Mertens et al., PNAS 2022) did find an average effect, but a subsequent reanalysis (Maier et al., PNAS 2022) showed that after correcting for publication bias the evidence for broad effectiveness disappears. And loss aversion faced criticism from Gal and Rucker (2018): for small amounts the asymmetry is weak or absent.

What does hold: defaults are among the most robust interventions there are, anchoring replicates reliably, and the core of prospect theory, the asymmetric weighting of losses at meaningful amounts, survived rigorous multi-lab replications. The practical conclusion is sober. Behavioural science is not a box of magic tricks, it is a hypothesis generator. Every hypothesis needs escalating into an A/B test on your own customers, because an average effect from a meta-analysis will not lift your conversion rate.

The ethical line: nudge or dark pattern?

The same psychology can help a customer or harm them. Fake countdowns, invented scarcity, pre-ticked flight insurance, cancelling a subscription hidden behind seven screens. Those are not nudges, they are dark patterns. In the EU the Digital Services Act explicitly bans them and consumer regulators have long pursued manipulative practices; the US FTC brought a case against Amazon over obstructed subscription cancellation. The test is simple: if the customer saw your intervention described out loud, would they thank you or feel cheated? In the short term a dark pattern lifts conversion. In the long term trust is what sells, and unlike conversion it does not regenerate with another sprint.

Want choice architecture built on data and tests instead of hunches? Here is how we experiment, or write to us directly.

Key takeaways (TL;DR)

  • The overwhelming majority of purchases, including rational B2B ones, run on System 1: fast, intuitive, by shortcuts. Marketing for a careful reader of spreadsheets assumes a customer who does not exist.
  • After the replication crisis, defaults, anchoring, truthful social proof and the core of prospect theory hold up; social priming collapsed and broad nudge effectiveness vanishes once publication bias is corrected.
  • Behavioural science is a hypothesis generator, not a box of magic tricks. Escalate every intervention into an A/B test on your own customers.
  • The ethical test: if the customer heard your intervention described out loud, would they thank you? Dark patterns lift conversion short term, erode trust long term, and in the EU the DSA pursues them.

FAQ

What in behavioural economics genuinely holds up after the replication crisis?

The most robust are defaults, anchoring and truthful social proof. The core of prospect theory survived, though loss aversion is weaker at small amounts than the early literature claimed. Social priming largely collapsed and broad nudge meta-analyses showed strong publication bias, so test effectiveness case by case.

How do you use behavioural science in pricing?

Through choice architecture: an anchor at the top, a recommended middle tier, a preselected default, framing of annual versus monthly payment, fewer options. And always an A/B test, because behavioural effects are sensitive to context and segment.

Where does a nudge end and a dark pattern begin?

Does the intervention help the customer decide in their own interest, or push them against it? Fake scarcity, hidden costs and a seven-click cancellation are dark patterns, and in the EU the DSA penalises them. Conversion in the short term, erosion of trust in the long term.

Sources: Daniel Kahneman, Thinking, Fast and Slow (2011) · Tversky & Kahneman (Science, 1974) · Kahneman & Tversky, prospect theory (Econometrica, 1979) · Johnson & Goldstein, Do Defaults Save Lives? (Science, 2003) · Mertens et al. (PNAS, 2022) and the reanalysis by Maier et al. (PNAS, 2022) · Gal & Rucker, The Loss of Loss Aversion (JCP, 2018) · EU Digital Services Act.