Bottom of funnel
Testing that runs long enough to mean something.
Testing run long enough, on real traffic, to mean something — including the intake steps after the form, where most conversion loss actually happens and no testing tool looks.
Most conversion rate optimization is sold as a practice built for e-commerce traffic volumes — hundreds of thousands of monthly visitors, tests reaching significance in days. A multi-location healthcare group or a self-storage portfolio rarely has that traffic at the page level. Testing a landing page that gets four hundred visits a month will not reach statistical significance before the underlying campaign, season, or price changes anyway, and running the test regardless just produces a result that looks decisive and is not.
The larger problem is where the losses actually happen. Most testing tools stop measuring at the form submission, but a meaningful share of real conversion loss happens after that: how quickly the lead is routed, how long until a callback, what happens on the actual phone call. Optimizing the page while ignoring what happens after the form is optimizing the part that was rarely losing the most.
How the work differs
- Sample size discipline
- Before a test launches, current traffic and conversion rate are used to calculate how long it would need to run to reach significance; if that is longer than the test is useful for, it does not run as a split test.
- Post-form intake
- Time from form submission to first callback, and the outcome of that callback, tracked as part of the conversion funnel.
- Test prioritization
- Candidates ranked by expected impact and by whether the traffic exists to reach a conclusion.
- Qualitative input
- Call recordings and session replay reviewed before a hypothesis is written, so a test targets an observed friction point, not a guess about design preference.
- Multi-location handling
- Where individual-location traffic is too low to test alone, similar locations are grouped for testing, then results checked for hidden per-location differences before rolling out everywhere.
- Inconclusive handling
- A test that ends inconclusive is reported as inconclusive, with the traffic and time a real answer would require.
How it is measured
The number that matters is the conversion once intake is complete — a booked appointment or a qualified handoff, not a form submission. A form fill is a candidate conversion; it is confirmed or rejected based on what happens on the call or the follow-up after it, matched through the same attribution layer used across every other channel.
What this cannot claim credit for: results attributable to a change in traffic quality or volume from the channels sending people to the page in the first place. If paid search sends a different audience one month, page conversion rate moves regardless of any test running, and that shift is separated from the test's own result rather than letting a testing win claim credit for a traffic change.
What we need from you
Access to whatever analytics and session recording tool is already in place, or agreement to add one where none exists. Call recordings or a transcript sample from the intake team, since the post-form review depends on hearing what actually happens on those calls. And a realistic traffic number per page before committing to a testing timeline, so the plan is built against real volume rather than assumed volume.
Who this is for
Built for operators who have real non-brand traffic already arriving through paid search or paid social but suspect the page, the form, or the intake process is where a meaningful share of it is lost — one of several bottom-of-funnel programs aimed at converting demand that already exists. Where a visitor leaves without converting at all, retargeting re-engages that same population on other channels, working from the data CRO already has on where they dropped off. If combined traffic across all pages is under roughly a thousand monthly visits per page being tested, expect to hear that directly rather than have a test launched anyway.
Professional services firms with a long consideration window often find more opportunity in intake than in the page itself. The plumbing that makes any of this measurable — lead routing, deduplication, the systems side — is marketing operations. See client results for a finished example of what this reporting looks like.
Often not at the level agencies typically sell CRO on, and we say so plainly before recommending a formal split test rather than after charging for one that could never reach significance. For lower-traffic pages, the more useful work is qualitative — session replay, call recordings, and structural fixes based on observed friction — rather than a statistical test with too little data to trust. Where similar locations can be grouped to reach usable volume together, we do that and check for hidden per-location differences afterward. The honest answer depends entirely on your actual numbers, not on what a standard proposal assumes.
As long as the sample size calculation says it needs to, based on your actual traffic and current conversion rate — that number is calculated before the test launches, not estimated afterward to fit a deadline. For a well-trafficked page that can be days to a couple of weeks; for a lower-traffic page it can be considerably longer, which is often the point at which we recommend a different approach entirely. We will not end a test early because a result looks favorable partway through, since that is exactly how false positives get reported as wins. The timeline is a calculation, not a guess.
It gets reported as inconclusive, along with the traffic and time that would actually be required to reach a real answer on that specific page. We do not quietly rerun it with a looser significance threshold until something clears the bar, since that produces a result that looks decisive without being one. An inconclusive result is still useful information — it usually means the variation tested was not different enough to matter, or the page was not the actual constraint. Either finding changes what gets tested next.
Both are workable, depending on your team's setup. Where we have implementation access, we build and run variations directly; where your development team owns the codebase, we hand over exact specifications and review the result once it is live. What we do not do is hand over a vague recommendation and call the engagement finished — whichever path fits your team, it ends with something actually running and measured, not a slide deck of suggestions.
Find out if your traffic can actually support testing.
A short traffic and intake review that tells you honestly whether formal testing makes sense yet, or where the real losses are happening instead.