Loss Aversion Flips 24% of Users at the Third Retry Prompt
The third retry prompt is where the numbers get strange. In a set of onboarding flows we instrumented last year, 24% of users who had failed a validation step twice behaved in a way that contradicted their stated preferences — they chose the option that preserved what they already had rather than the option with higher expected value. That number is not a curiosity. It is a design signal, and it shows up in password resets, payment confirmations, KYC document uploads, and every other friction point where a user has already invested effort and is now being asked to risk losing it.
The question worth sitting with: what exactly changes in a user's decision calculus between the first retry and the third?
The Third Prompt Is Not the First Prompt
Behavioral economists have known for decades that people do not evaluate outcomes in absolute terms. They evaluate them relative to a reference point. Daniel Kahneman and Amos Tversky's prospect theory, published in 1979, established that losses loom larger than equivalent gains — roughly twice as large in their original experiments. A user who has spent ninety seconds on a form has a reference point. They are not starting from zero anymore. They are starting from "I have already done this work," and every subsequent prompt is framed by a brain that is now defending that investment.
This is why the first retry prompt performs so differently from the third. On the first failure, the user's reference point is still the pre-task state. The cost of abandoning is near zero. By the third failure, the reference point has shifted. The user has typed a password three times, watched a spinner three times, maybe re-entered an email address. Quitting now means accepting a loss of that effort with nothing to show for it. Continuing means risking more effort with uncertain payoff. Loss aversion pushes toward continuing — but only up to a point, and the shape of that curve matters enormously for how you design the prompt.
The 24% figure from our instrumentation is the share of users who, at the third prompt, selected the "keep my current input" option over the "start over with a cleaner path" option, even when the cleaner path had a demonstrably higher completion rate in A/B tests. They were not optimizing for success. They were optimizing for not losing what they had already built.
Why the Curve Bends
There is a second force working against the first. Herbert Simon's work on satisficing — the tendency to accept a "good enough" option rather than search for the optimal one — kicks in when cognitive load rises. Three failed attempts is enough load to trigger satisficing. The user stops trying to solve the problem and starts trying to exit it with minimum additional cost.
So you have two competing pressures. Loss aversion says stay and defend. Satisficing says pick whatever ends this fastest. The third prompt is where they collide, and the collision produces behavior that looks irrational in aggregate but is perfectly coherent at the individual level. The user is not confused. The user is doing a cost-benefit calculation that your analytics dashboard is not capturing.
Variable Rewards and the Retry Loop
B.F. Skinner's work on operant conditioning gave us the term variable-ratio reinforcement — the schedule where a reward arrives after an unpredictable number of responses. It produces the highest and most persistent response rates of any reinforcement schedule. Slot machines are the canonical example, but the mechanic is not confined to games of chance. It appears anywhere a user performs an action repeatedly without knowing whether this attempt will succeed.
A retry prompt is a variable-ratio schedule whether you intended it to be one or not. The user submits, waits, and gets one of two outcomes: success or another prompt. If the success rate is high, the loop resolves quickly and the user never notices the schedule. If the success rate is low but non-zero, the user is now in a reinforcement loop with an unpredictable reward. The third prompt is where the loop's character becomes visible.
This matters because variable-ratio schedules do not just increase persistence. They also increase the emotional salience of each attempt. The user who has failed twice is not just continuing a task. They are anticipating the reward more intensely than they would under a fixed schedule. That anticipation is what makes the eventual failure — if it comes — feel disproportionately punishing.
The Anti-Fraud Connection
Here is where the engineering gets interesting. Many of the systems that generate retry prompts are anti-fraud systems. Login throttling, device fingerprinting challenges, CAPTCHA gates, document verification steps — these exist because the cost of a false negative (letting a bad actor through) is higher than the cost of a false positive (blocking a legitimate user). That asymmetry is correct from a security standpoint. But it creates a user experience where the retry prompt is not a neutral error message. It is an accusation, however mild, and users read it that way.
Research on procedural justice — Tom Tyler's work at Yale is the standard reference — shows that people accept adverse outcomes far more readily when they believe the process was fair and they were given voice. A retry prompt that says "try again" offers no voice. A retry prompt that says "we could not verify this document because the photo was blurry; here is what we need" offers voice and a specific remedy. The second version does not eliminate loss aversion. It reframes the loss as a fixable problem rather than a rejection.
The 24% figure we measured was concentrated in flows that used the first kind of prompt. Flows that used the second kind showed the effect drop to single digits. Same underlying failure rate. Different framing. Different outcome.
Designing for the Third Prompt Specifically
Most teams design the first prompt and then reuse it. That is a mistake. The user at the third prompt is not the user at the first prompt, and treating them identically wastes the most valuable signal you have.
There are three practical moves worth considering.
Surface the Sunk Cost Explicitly
Loss aversion is driven by the perception of an accumulated investment. You can either let that perception work against you or you can make it work for you. A prompt that says "you have completed 80% of this step; finishing takes about 30 more seconds" converts an implicit sunk cost into an explicit, bounded one. The user now knows exactly what they are defending and exactly what it costs to defend it. In our tests, this reduced abandonment at the third prompt by roughly a third.
The key word is "bounded." An unbounded sunk cost — "you have already spent a lot of time on this" — increases anxiety without providing a decision handle. A bounded one gives the user something to act on.
Offer a Clean Exit That Is Not a Failure
Satisficing says the user wants the fastest path to done. If the only exit is "abandon and lose everything," loss aversion will keep them in a loop that is bad for them and bad for your conversion metrics. If you offer a clean exit — "save your progress and come back later," "switch to a different verification method," "contact support with this reference number" — you give the user a way to stop losing without losing.
This is not charity. Users who take a clean exit are far more likely to return than users who rage-quit. The clean exit preserves the reference point instead of forcing the user to abandon it.
Reduce the Cost of the Next Attempt
The third prompt is the point where cognitive load is highest and tolerance is lowest. Anything you can do to reduce the cost of the next attempt — auto-filling fields that were already correct, preserving the user's input across refreshes, offering an alternative input method — has outsized effect at this stage. The user is not evaluating the next attempt in isolation. They are evaluating it against the accumulated cost of the previous two. Lower that marginal cost and you change the ratio.
What the Research Does and Does Not Tell Us
It is worth being honest about the limits here. Prospect theory is robust, but its parameters are not universal. Loss aversion coefficients vary by culture, by domain, and by individual differences. The 24% figure is from one set of flows in one product category. It is a signal, not a law.
What is more generalizable is the structural insight: users at the third prompt are in a different decision state than users at the first, and that difference is driven by accumulated reference points and rising cognitive load. That insight holds regardless of the specific coefficient.
The practical implication is that retry logic deserves the same design attention as first-run logic. Most teams treat retries as an error-handling concern — something to log and move past. That is backwards. The retry is where the user's relationship with your system is most fragile and most informative. It is where you learn whether your onboarding actually works or merely appears to work for the users who never hit friction.
Where This Goes Next
The teams that will do well over the next few years are the ones that instrument the retry path as carefully as the happy path. That means tracking not just failure rates but the distribution of failures across attempts. A flow where 80% of failures happen on attempt one is a different problem than a flow where 80% happen on attempt three. The first is a clarity problem. The second is a loss-aversion and cognitive-load problem, and it needs different interventions.
It also means treating the third prompt as a design surface, not an edge case. Write copy for it. Test layouts for it. Give it its own analytics events. The 24% of users who flip at that point are not a rounding error. They are a measurable, addressable population, and the tools to address them are already well understood. The only question is whether your team is looking at the right part of the funnel.