Redesigning the In-Booking Donation Flow: Evidence from Three Airline Studies

Status: working paper. All three studies in the program are now complete (combined n = 1,413). The participating airlines are described rather than named while we confirm permission to identify them. One study carries a methodological caveat that we flag wherever its numbers appear.

What we tested

Airlines already ask customers to donate to environmental causes during booking. The question is whether the design of that ask matters. We rebuilt the donation step of three real booking flows — a major European tour operator, an international flag carrier, and a third major carrier — and tested each redesign against that airline's current live design.

The redesign made three changes: a photograph showing what the donation protects, a counter showing how many others had given, and preset one-tap amounts. Every study used an incentive-compatible first phase: participants received a real monetary bonus and chose whether to keep it or donate it during a simulated booking. Donations were real and were transferred to charity. A second, hypothetical phase compared interface variants.

We organize the findings by result rather than by airline, so that what replicates across samples — and what does not — is visible directly.

Giving rose in every study

The redesigned page produced a higher real-money donation rate than the current design in all three studies, and the difference was statistically significant every time. That directional replication on real money is the strongest single claim the program supports.

• Flag carrier (n = 467): 41.9% vs. 18.0% (Fisher's exact p < 0.001, Cohen's d = 0.538) — 2.3× the donations per customer.
• Tour operator (n = 480): 30.0% vs. 18.8% (p = 0.0056, d = 0.264).
• Third carrier (n = 466): 35.1% vs. 20.4% (p = 0.00043, d = 0.331). See the caveat below — this study changed two things at once.

The direction replicates cleanly; the magnitude does not. Effect sizes span a roughly two-fold range, tracking design differences between the studies — chiefly each airline's starting interface (a weak current control inflates the apparent lift) and whether the redesign offered a free-entry amount field or a single fixed ask.

How much people give is a design choice

In every study, donation rates were statistically indistinguishable across the hypothetical interface variants — but the amounts were not. Raising the preset anchors multiplied per-customer yield with no loss in conversion:

• High anchors lifted revenue per 100 customers by 3× to 7.5× over low anchors across the three studies — the single largest revenue lever in the program, and its cleanest cross-study replication.
• The effect is biggest with fixed presets and smaller with a free-entry field. When donors can't type their own number they cluster exactly on the offered anchor; a free-entry field lets them override it, dampening the gain.
• Pre-selecting a higher default did not work. Customers overrode the default; the anchor set, not the selection, drove amounts.
• A round-up mechanic converted in the 30s to low-40s percent — reliable, implementation-sensitive, and never a standout against the high-anchor design.

Anchor amounts don't change whether people give. They change how much.

The ask cost the airlines nothing

After the real-money phase, participants rated six attitudes. In all three studies, exactly two moved — environmental consciousness and climate urgency — and only those two. Brand favorability, future booking likelihood, price perception, and willingness to donate did not shift in any study. Three independent samples producing the same two-item signature, and only that signature, is a clean pattern replication.

The consistency of the null results is the central commercial reassurance: across three brands, asking customers to give — even asking more prominently — created no measurable commercial friction while moving how customers think about the environmental cost of flying.

A caveat we won't bury

The third study changed two things at once between its current and redesigned arms: the interface design and the named recipient charity (a children's cause versus an environmental one). Its donation-rate lift and its attitudinal shifts therefore reflect a combined design-plus-cause package, and no statistical procedure in that design can separate the two. We treat its numbers as directionally consistent with the other two studies, not as independent evidence about interface design — and the clean attitudinal replication rests on the first two studies, with the third as suggestive.

A separate implementation issue compounded it: the donation confirmation step displayed the old charity's name inside the redesigned flow. Because that bug sat only in the redesigned arm, its likely effect was to suppress that arm — so a correctly built redesign might perform better than measured. It should be fixed before any further testing, and it does not rescue the underlying design-versus-cause confound. We're publishing the caveat because honest limitations are how the rest of the evidence earns trust.

What holds people back

Open feedback across all three studies pointed to one dominant barrier, and it was not opposition to climate action: people don't trust where the money goes. Respondents asked for specific outcomes, and several said directly that they would give if they knew exactly what their donation funded. Showing two or three concrete results on the donation page is likely the most effective change available outside of interface design itself.

Stated willingness behaved the same way everywhere, too: it was a clean screen for who donates — people who expressed any willingness gave at many times the rate of those who expressed none — but a poor estimate of how much they would give. Treat it as a targeting signal, not a revenue forecast.

What this means at scale

The three carriers together reach roughly 74 million passengers a year. Built only on real-money behavior, the redesign projects to conservative annual upside of about $5.1M (tour operator), £13.9M (flag carrier), and €2.5M (third carrier — the design-plus-cause figure). If the high-anchor effects from the hypothetical phase hold in production, the per-carrier upper bounds are several times larger.

These figures are reported in native currency and are not summed directly; an illustrative program-wide aggregate at assumed exchange rates is on the order of $25M a year on conservative real-money behavior, rising toward $110M a year if high anchors hold in production. The aggregate is a sum of three separately measured results, not a pooled estimate, and the upper bounds rest on two unproven assumptions (that the rate lift survives at scale, and that high anchors produce real-money giving at the hypothetical levels). A live A/B test inside a production booking flow is the right next step.

What to do

1. Adopt the redesigned flow and confirm with a live A/B test in each booking funnel, to pin each carrier's effect against its own baseline.
2. Make high suggested amounts the primary amount lever — and test the anchor set together with the entry mode, since a free-entry field dampens the gain. Don't rely on a pre-selected high default; customers override it.
3. Put two or three concrete charity outcomes on the donation page. The top barrier in every study was uncertainty about where the money goes.
4. Validate the specific image before launch, especially on single-ask layouts where a weak image can cost conversions.
5. For the confounded study, fix the charity-name bug and re-run with the charity held constant, so design can finally be isolated from cause.

Methods snapshot

1,413 participants were recruited via Prolific across three sequential studies and randomly assigned to each airline's current donation flow or its redesign. The primary outcome was a real-money donation decision; secondary outcomes were donation amount, six attitudinal measures, stated booking likelihood, and stated donation willingness. Binary comparisons used Fisher's exact test, attitudinal items used independent-samples t-tests, the ordinal booking-intent item used Mann-Whitney U, and all donated funds were transferred to charity. Key limitations: participants may not match the airlines' customer bases; the second phase used hypothetical decisions, so its anchor magnitudes need real-money confirmation; the three studies used different baseline interfaces and currencies, so percentage-point uplifts and currency levels are not a clean cross-study ranking; and one study confounded interface design with the named charity, as described above.