We run conversion programs on research rather than opinion, and we call tests on pre-agreed sample sizes instead of stopping the moment a variant looks good. When the honest answer is that your traffic can’t support testing yet, we say so and fix the obvious problems instead.
“We ran tests and nothing ever won.”
Usually a sample size problem, not an ideas problem. The audit checks whether testing is viable at all. Start with the audit
“We know the funnel leaks but not where.”
Session recordings, funnel analysis, and form analytics locate it before we touch design. See the research
“Our traffic is modest.”
Then a testing program is the wrong purchase. We’d do a research audit and fix the found problems directly. Read who this isn’t for
“We want more leads without more ad spend.”
That’s the point of this. Same traffic, better conversion. See the plans
These figures cover every Gatilab service, not this one alone. We’d rather say that plainly than imply a per-service record we haven’t published.
Brands we’ve delivered work for
A testing program needs roughly 1,000 conversions a month to produce trustworthy results at reasonable speed. Below that, research and direct fixes beat A/B testing, and we’ll tell you which situation you’re in.
For any traffic level. This is the honest first purchase and the only one that works at low volume.
One funnel · 2 to 3 weeks
Includes a straight answer on whether your traffic can support testing at all.
For sites with roughly 1,000+ monthly conversions where compounding small wins is worth real money.
Needs ~1,000+ monthly conversions · 3-month minimum
We report losers and inconclusive tests in full. Roughly half of all tests lose.
For teams where CRO touches pricing, onboarding, and lifecycle, not just landing pages.
3-month minimum
We’ll recommend the smaller tier if your conversion volume can’t justify this.
Most failed CRO programs fail on statistics, not creativity. This process is built around that.
Weeks 1 to 2
Analytics, session recordings, form analytics, and support tickets get read together. Support tickets are the most underused conversion research there is, because customers already told you what confused them.
You leave with: A ranked list of real problems, each with the evidence behind it.
Week 3
Each hypothesis gets a predicted effect size, then a required sample size and test duration. Anything that would need eight months to reach significance gets built directly instead of tested.
You leave with: A test plan with sample sizes and stop dates fixed in advance.
Per test
Variants get built and checked across browsers, devices, and screen readers. A variant that breaks on mobile Safari doesn’t just lose, it poisons the data.
You leave with: A tested variant and a QA record, before a single visitor sees it.
Per test
Tests run to their pre-agreed sample size and get called against the rule set at the start. No peeking and stopping early, which is how most agencies manufacture winners.
You leave with: A written result, including the losers and the inconclusive ones.
This service is unusually easy to fake, because a stopped-early test always looks like a win. Two of these are reasons we decline work.
Something not covered? Ask directly and you’ll get a straight answer, not a calendar link.
Roughly 1,000 conversions a month to run a testing program that reaches significance in two to four weeks per test. Below that, tests take months and you’ll be tempted to call them early, which is worse than not testing. The research audit works at any traffic level and is the right purchase if you’re under that line.
We won’t quote one before research, and treat any agency that does as a warning sign. What we will commit to is honest measurement, pre-agreed sample sizes, and full reporting of losers. Across a well-run program, roughly one test in three wins, one loses, and one is inconclusive.
Because if you watch a running test and stop when it looks good, you will find a winner almost every time, and it usually won’t hold. Fixing the sample size and stopping rule before launch is the difference between measurement and storytelling.
Yes, in full, with the same detail as winners. A losing test that tells you customers do want the longer form is worth real money in decisions avoided. An agency that only shows you wins is either lucky, selective, or stopping early.
Usually yes, through a client-side testing tool for most changes. Anything touching checkout, pricing logic, or server-rendered content needs your developers, and we’ll say which is which during the hypothesis stage rather than discovering it mid-build.
Then you’ve saved yourself several months of retainer, and you still have a ranked list of real problems to fix directly. That’s a good outcome for you and an honest one for us. It happens on a meaningful share of audits.
Send your funnel, your rough monthly conversion count, and where you suspect the problem is. We’ll tell you whether a testing program makes sense for you or whether the audit alone is the better buy.
