There is a finding in our testing program that initially seemed counterintuitive, and that I have come to think is one of the most important patterns in high-consideration funnel optimization.

Late-funnel interventions dramatically outperform early-funnel ones.

When I say late-funnel, I mean tests targeting users who have already progressed significantly through an enrollment process — who have filled out multiple pages of a form, provided personal information, made selections, and are nearing the end of the process. When I say dramatically outperform, I mean the win rates and effect sizes for tests in this stage of the funnel are not incrementally better than early-funnel tests. They are categorically better.

The naive explanation is traffic volume: more users start an enrollment than complete it, so early-funnel tests see more volume and reach significance faster. That is true, but it does not explain the effect size difference. Tests targeting late-funnel users were not just running with smaller samples — they were producing larger percentage lifts when they did win.

The behavioral explanation is more interesting. And it traces back to a principle called the endowment effect.

What the Endowment Effect Actually Says

The endowment effect, documented extensively in behavioral economics research, describes the tendency for people to assign more value to things they already possess than to equivalent things they do not possess. The classic demonstration: people who are given a coffee mug demand more money to sell it than people who were not given a mug are willing to pay to buy it. Same mug, different valuations — the only difference is ownership.

Richard Thaler, who formalized the concept, connected it to Kahneman and Tversky's loss aversion framework: giving up something you own registers as a loss, and losses loom larger than equivalent gains. Ownership, even temporary ownership, creates attachment that distorts the rational exchange calculation.

What is less commonly discussed is how the endowment effect extends beyond physical possession to invested effort.

Research on effort justification — the psychological tendency to value outcomes more highly when they required significant effort to achieve — shows that people who have invested labor, time, or cognitive resources in a process feel a form of ownership over the output of that process. The effort itself creates a psychological stake.

This has been called the "IKEA effect" in some research: people who assemble furniture themselves rate it more highly than equivalent pre-assembled furniture. The effort invested generates a sense of ownership and value that is not present in the effortless acquisition of the same item.

Now think about what a multi-step enrollment form is, behaviorally.

Your Enrollment Form Is Building Psychological Stakes

A user who has spent 10 minutes filling out a multi-step enrollment form has invested something. They have typed their name and contact information. They have answered qualifying questions. They have made product selections. They have disclosed personal details — financial information, health information, preferences. They have, in a meaningful behavioral sense, begun the process of becoming a customer.

That investment creates a psychological stake in the outcome. The form they have filled out is partially "theirs." The account they are creating has, in some anticipatory sense, already begun to exist for them. Walking away from the process at this point is not the same as deciding not to start. It registers, at some level, as loss — the loss of the thing they have partially built.

This is why late-funnel users, in our testing data, respond so much more strongly to intervention. The behavioral conditions that make intervention effective are present. These users are not indifferent to completion versus abandonment. They have skin in the game — psychological skin, created by their own invested effort.

Early-funnel users have invested nothing. They are at the beginning of a process. The question for them is whether to start, which is a low-stakes decision with no sunk cost and no sense of ownership. Early-funnel interventions are trying to motivate action where there is no baseline attachment. The behavioral environment is flat.

Key Takeaway: Multi-step enrollment creates psychological ownership through invested effort. Late-funnel users have a stake in completion that early-funnel users do not. Tests designed to help users complete what they've started operate on a fundamentally more favorable behavioral substrate than tests trying to convince users to begin.

The Confirmation Page That Shouldn't Have Worked — But Did

The most striking illustration of this principle in our testing program came from a test that most teams would not have prioritized.

A user had completed enrollment. Fully enrolled. They had cleared every step of the form, confirmed their selections, and were now on a post-enrollment confirmation page. The test added a widget to this confirmation page: a preview of the dashboard they had just unlocked, showing them content and features they could now access as a new member.

Common sense might suggest this test is irrelevant to conversion. The user has converted. What is there to optimize?

The answer is that the confirmation page test was measuring downstream activation — specifically, how many confirmed enrollees went on to engage with their new account, complete additional setup steps, and reach the "activated user" milestone that predicted long-term retention.

The test produced one of the largest lifts in the entire program, measured by 30-day activation rate.

Here is the behavioral explanation. The confirmation page arrives at the precise moment of peak psychological ownership. The user has just completed a significant effort. They have just received confirmation that the thing they built is real. The endowment effect is at its strongest: they are invested in the outcome and primed to value what they have created.

Showing them the dashboard at this moment — demonstrating the concrete value of what they now own — capitalizes on that peak ownership state. They are not evaluating whether to enroll; they have enrolled. They are evaluating whether the thing they just acquired is worth their continued attention. The timing is perfect, and the test won because it was designed with the behavioral moment in mind.

The Contrast: When Urgency Hits Users Without Stakes

Compare that confirmation page result to another test from the same program: a containment modal.

The containment modal targeted users who had arrived at a landing page but had not yet begun filling out the enrollment form — pure early-funnel traffic. It deployed a dismissible overlay with an urgency message designed to catch users before they bounced.

This test failed. Decisively.

The behavioral explanation is almost the mirror image of the confirmation page story. These users had invested nothing. They had not filled out a single field. They had no psychological stake in completion. The modal was asking them to feel urgency about something they had not yet started — trying to manufacture a sense of investment where none existed.

You cannot borrow against stakes that have not been accumulated. The endowment effect does not operate at the beginning of a process. It is a function of accumulated investment, and these users had accumulated none. The urgency message landed in a behavioral environment that could not support it.

The modal performed below control, likely because the intrusive overlay generated friction and negative affect that slightly depressed the conversion rate rather than improving it.

This contrast — the confirmation page dramatically winning, the entry modal failing — is the pattern that convinced me effort investment was the right explanation for the late-funnel advantage. It is not simply that late-funnel users are higher-intent. It is that they have built something they do not want to abandon.

Implications for Multi-Step Form Design

If effort investment creates psychological ownership that makes users more likely to complete, this has some non-obvious implications for how you design and optimize multi-step processes.

Progress indicators are not just UX — they are psychological ownership signals. A progress bar that shows "Step 3 of 5" is not only orienting users in the process. It is quantifying their investment. It tells them they have put in 60% of the effort required and have 40% remaining. This makes the completion calculation feel asymmetric in favor of finishing — which is the behavioral outcome you want. Removing progress indicators from long forms, in an attempt to prevent users from feeling daunted, may backfire by removing the ownership signal that motivates completion.

Personalization of partially-completed forms increases stake. When a form remembers what a user previously entered, addresses them by name, or references the specific product configuration they were in the middle of building, it reinforces the sense that this thing is already partially theirs. Re-engagement campaigns for abandoned enrollment flows perform better when they reference the specific state of the user's incomplete application rather than starting fresh.

Front-loading form fields that feel "ownership-creating" may help. If users are most likely to abandon at the beginning of a process, one strategy is to design the early steps to create investment quickly. Not necessarily the most effort-intensive steps — but steps that make the user feel like they are building something personal. Name personalization, preference selection, goal-setting questions. The sooner users feel like the process is "theirs," the stickier the mid-funnel becomes.

Late-funnel is a higher-leverage optimization target than its traffic volume suggests. Many programs over-index on top-of-funnel testing because the traffic is there and statistical significance comes quickly. The endowment effect logic suggests that late-funnel is where behavioral conditions most favor intervention. The user is invested, they are close to completion, and a well-designed nudge operates on a substrate of accumulated psychological ownership. Traffic volume is smaller but impact per test is larger.

The Sunk Cost Connection

I want to draw the connection to sunk cost psychology carefully, because it is relevant but also where things can get ethically complicated.

Sunk cost bias — the tendency to continue investing in something because of past investment, even when the rational decision is to stop — operates through a related mechanism. The endowment effect creates value from ownership; sunk cost bias creates continuation from past investment. Both are driven by the asymmetric weight humans place on losses versus gains.

There is an ethical difference between designing enrollment to help users complete something they genuinely want and designing enrollment to make it difficult to stop because you have engineered artificial sunk costs. The former is good product design. The latter is dark pattern design.

The confirmation page test worked because it helped users get value from something they had genuinely chosen. The endowment effect was serving the user's stated goal: they enrolled because they wanted the product, and the test helped them realize that value faster.

A multi-step form designed to make each step feel like a larger and larger investment specifically in order to make stopping feel more painful — that is sunk cost engineering, and it is not the same thing. The distinction matters for how you design, and it is one I think about every time a "make the form feel bigger so people won't abandon" idea comes across the testing roadmap.

The Practical Takeaway

If you have a multi-step enrollment flow and your optimization roadmap is weighted toward top-of-funnel tests, reconsider the allocation.

Your highest-leverage testing territory is likely in the middle and late stages of the process, where users have invested effort and have psychological stakes in completion. Tests that help these users over specific barriers — clarity gaps, friction points, confusion about what happens next — will outperform tests that try to convince fresh traffic to start.

And do not forget the post-enrollment window. Users who just completed enrollment are at peak psychological ownership. This is the best moment to introduce them to the value they now own. A well-designed confirmation experience is not a nice-to-have — it is a high-leverage activation intervention at the behavioral moment when users are most receptive to it.

If you want to track and compare your late-funnel versus early-funnel test results to see where you're generating the most lift, GrowthLayer lets you tag experiments by funnel stage and visualize win rates by stage over time. The pattern becomes very apparent when you can see it clearly across your entire test history.

The Endowment Effect in Multi-Step Enrollment: Why Users Who've Invested Effort Convert Better