A nonprofit found its donation page's winning argument through disciplined tests

Representative examplesEvery case study in this library is an illustrative composite of the kind of engagement we deliver — written to show our method and standards, not to name clients.

Where they started

Regional food security is this nonprofit's whole work, and its standing in the community is the asset every campaign spends and rebuilds. Individual giving runs through seasonal pushes — a winter appeal, a spring match, a giving-day effort — planned by a leadership team and executed by a two-person fundraising staff. Donation pages had been redesigned over the years by whoever had capacity, and each campaign's messaging reflected the last strong opinion voiced in the room. Fundraising meetings spent as much time relitigating old wording as planning the next appeal.

What it was costing

Donation campaigns ran on conviction and argument — leadership debates about messaging were settled by whoever argued longest. The donation page converted at 11%, and nobody knew which story actually moved donors.

What they could see

• Message debates in leadership meetings ended by seniority or stamina rather than by any evidence.
• The winter appeal's page carried three competing arguments, added in successive years and never removed.
• The fundraising staff could not say which story last season's donors had actually responded to.
• Board members proposed copy changes each campaign from personal reading, and each one received a courteous rewrite.
• Nothing was ever tested, partly from fear that testing would drift into pressuring donors.

The constraints we worked inside

• Donor trust is the asset — no urgency theater, no manipulative framing permitted.
• Campaign windows were fixed by the giving season; tests had to fit real campaigns.
• Traffic was modest, so tests needed honest sample-size planning rather than vanity wins.

What had been tried before

The team once ran a two-version test on the donation button's color during a campaign.

With the season's modest traffic split across a difference too small to matter, the test consumed weeks and taught the team that testing was slow theater.

A board member with agency experience rewrote the appeal's headline each year.

Each rewrite reset the page's voice and results moved within noise, so the exercise produced churn the staff then had to defend in the next meeting.

A fundraising consultant was engaged to recommend storytelling improvements to the appeals.

The advice was sound but generic; without the charity's own data behind it, it became one more opinion in the room, and the loudest voice still won.

What we proposed

Pre-commit the test plan before the season opens: three message hypotheses — impact-per-dollar, local visibility, beneficiary voice — each written down with a sample size and an end date, agreed by leadership in advance. Variants then run inside real campaign traffic, with dynamic text matching each campaign's audience, so results come from actual donors rather than synthetic visits. Every outcome, including the losers, goes into the campaign retro. The discipline is the point: once leadership signs the plan, the argument moves from the meeting to the field.

Just as important is what we ruled out, and why:

• Post-campaign surveys asking donors what moved them — Donors reliably misreport their own motives, and a survey would have recreated the opinion contest in a slower, more expensive form.
• Testing in the off-season with paid traffic — Bought traffic does not behave like a loyal regional donor base; results would have been precise and wrong, and the budget belongs to programs, not synthetic audiences.
• A full page redesign before testing messages — The page's structure was not the dispute — the story was; redesigning first would have eaten the season and answered a question nobody was asking.

How the work ran

Delivered by the growth pod — strategist over 6 weeks, with working increments reviewed with the client every week.

The stack, and the reasoning

Unbounce

A two-person fundraising staff had to build and run variants without a web vendor in the loop; the editor let testing happen as part of the campaign rather than as a technical project.

A/B variants

Split runs were configured with fixed end dates before launch, so the tool's own discipline held the line when mid-campaign enthusiasm argued for peeking early.

Donation platform

Donations completed on the charity's existing platform, so every variant's revenue was recorded where the finance office already reconciles — no new system for the treasurer to learn or trust.

GA4

Years of untagged links had blurred attribution; cleaning campaign tags meant each variant's visitors traced cleanly back to the appeal that sent them.

What went wrong

How we worked together

Cadence

A 30-minute Monday check-in during the season and one written mid-test summary; the post-campaign retro was the commitment everyone could see on the calendar.

Client side

The two fundraising staff ran variants and reporting; leadership attended the pre-season plan review and the retro — and nothing in between, by agreement.

Decisions

The written test plan made the calls; any deviation required a written reason in the shared plan, which kept mid-season second-guessing from reopening settled debates.

They provided

Past campaign data for hypothesis-setting, the fixed campaign calendar, a tone review of donor-facing copy, and consent handling for the beneficiary story.

What changed

The headline: visit-to-donation conversion on the tested campaign versus the prior-year equivalent — 11% → 17%, read from Donation platform records. A second check: hypotheses resolved within the pre-committed windows at 2 of 3.

The argument pattern changed: a claim now arrives with 'that's testable' attached, and the season's plan holds the answer. Leadership stopped relitigating old wording because the retro documents why each choice was made, including the ones that did not work. The fundraising staff describe running experiments like professionals instead of defending instincts like staff. And the losing hypotheses stay valuable — the rejected stories are written down, so the organization remembers what it already learned about its donors instead of relearning it every winter.

The result was read from Donation platform records against the pre-engagement baseline over the stated window, with a guardrail check on hypotheses resolved within the pre-committed windows. Where platform-reported numbers and business outcomes differ, this record says which layer it is quoting.

What they own now

• The season's test plan document with hypotheses, sample sizes, and end dates.
• The Unbounce workspace with variant pages and pre-configured split logic.
• A retro template that records each test's outcome, including inconclusive ones.
• The campaign-tagging convention and untagged-link fixes documented in GA4.
• The consent checklist used for beneficiary-story content.