Conversion
There is no conversion retainer here
Conversion work is built into the design rather than sold beside it. Three things carry almost all of it: speed, accessibility and clarity. Two of those have published evidence behind them and the third is a design judgement, and this page is explicit about which is which.
A/B testing is available and is not the opening move. For a site with 1,000 visitors a month, resolving even a large improvement takes years — and the arithmetic for that is on this page rather than asserted.
Why it is not a separate service
Most conversion problems are built in, not tuned out
A slow site, an inaccessible form and an unclear offer are all decisions made during the build. Selling a monthly retainer to recover from them afterwards is selling a fix for a problem that did not need to exist.
So the work happens in the design: performance budgets before the first page is built, accessible markup as the default rather than an audit finding, and one obvious next action per page. That is not a cheaper version of conversion optimization. For a business at this size it is the version that works.
What actually moves it
Three levers, and what the evidence says about each
Speed
Not a technical vanity metric. The clearest measured relationship between anything a designer controls and whether a visitor converts. It is also the lever that survives every redesign, every platform change and every algorithm update, which is why it is built in from the first build rather than optimised afterwards.
Evidence: A study of 37 brands across 30 million sessions found a 0.1 second mobile improvement raised lead-generation form submissions 21.6% and retail conversions 8.4%. Observational rather than experimental, on large brands rather than small businesses.1
Accessibility
Treated here as conversion work that happens to also be compliance work, because that is the honest order of the benefits for a small business. A form field with no label, a button with no text and text at 3:1 contrast are all conversion faults first. Everyone struggles with them; some people are stopped by them entirely.
Evidence: Automated testing of the top million home pages found 95.9% with detectable WCAG failures, averaging 56.1 errors per page. Low contrast text on 83.9%, missing form input labels on 51%. The authors note automated testing cannot detect every failure, so the real figure is higher.2
Clarity
What you do, who it is for, what it costs, and what happens if they get in touch — answered above the fold and repeated where the decision actually gets made. Most small business sites fail on this before any of the subtler things matter, and it is the cheapest of the three to fix.
No evidence: No citation offered. This is a design judgement built from doing the work, not a measured effect, and a page that cites everything else should say plainly where it has nothing to cite.
The honest arithmetic
Why your site probably cannot be A/B tested
An A/B test needs enough visitors to tell a real improvement apart from chance. How many depends on your conversion rate and on how small an improvement you want to catch3. For most small businesses the answer is more visitors than they will see for years.
| Conversion rate | Improvement to detect | Visitors per month | Visitors needed | Time to a result |
|---|---|---|---|---|
| 2% | 20% | 1,000 | 42,212 | 3.5 years |
| 2% | 10% | 1,000 | 161,358 | 13.4 years |
| 2% | 20% | 10,000 | 42,212 | 4.2 months |
At 95% confidence and 80% power, computed from the standard formula for comparing two proportions. The figures are generous to testing: a 20% improvement is a large effect that few design changes produce, and the calculation assumes every visitor enters the test, which no real test achieves.
The failure mode is not waiting three years. It is running the test for a fortnight, seeing the variant ahead, and shipping it. Stopping a test the moment it looks good is how a coin that came up heads six times becomes a strategy. With samples this small the result is mostly noise, and a confident decision made from noise is worse than no decision at all.
Above roughly 10,000 visitors a month this stops being theoretical and testing starts earning its place — the third row. If that is you, it is worth doing and worth doing properly. If it is not, the work below is what moves the number.
Inside the design work
What gets done instead
None of this needs a test to justify it. A page that loads fast, a form everyone can complete and a clear answer to what you cost are improvements whether or not an experiment at your traffic level could detect them.
It runs inside the web design project, at no separate fee.
- Performance budgets set before the first page is built, not profiled afterwards
- Accessible markup as the default: real labels, real buttons, contrast that passes
- One obvious next action per page, matched to how quickly your customers actually decide
- Forms that ask for what you need to quote the job, and nothing else
- The things buyers check before calling — price range, service area, proof — made findable
- Phone and email treated as conversion paths, because for most of these businesses they are
Fair questions
What people ask before buying this
No. It is how the design work is done rather than something added to it. Speed budgets, accessible markup and a clear next action are part of every build here, and they are not itemised separately because they are not optional extras. If your site was built elsewhere and needs this work retrofitted, that is a project rather than a retainer, and it is quoted as one.
Yes, when there is enough traffic for the result to mean anything, which for most small businesses there is not. A site converting at 2% with 1,000 visitors a month needs roughly 42,000 visitors to detect even a large 20% improvement at conventional confidence — about three and a half years. Running a test anyway and stopping when it looks good is how people convince themselves of things that are not true. Above roughly 10,000 visitors a month it starts being worth doing, and then it is worth doing properly.
Work on the things that do not need a test to justify them. A page that loads in under a second, a form everyone can complete, and a clear answer to what you do and what it costs are improvements whether or not a test could detect them at your traffic level. That is not a compromise position — with small samples it is the more rigorous one, because the alternative is making decisions from noise.
Enquiries, calls and booked work, measured against the same months a year earlier, with the caveat stated plainly: that is a before-and-after comparison, not a controlled experiment, and seasonality and marketing changes are mixed into it. Anyone presenting a year-over-year lift as proof of a design decision is overclaiming, including us. What can be measured cleanly is the mechanism — load time, accessibility failures, form completion rate — and those get reported as what they are.
Possibly, and it depends on your sector, size and where you operate, which makes it a question for a lawyer rather than a web designer. The reason it is on this page is commercial rather than legal: the most common failures found across the top million home pages are missing form labels and low contrast text, and both of those cost you enquiries from people who have no disability at all.
Sometimes, and the honest answer depends on what it is built on. Speed and accessibility are often limited by the platform and theme rather than by anything a consultant can tune, and there is a point where fixing those costs more than rebuilding. Where the site is the constraint that gets said plainly instead of being worked around for months.
Nothing separate. It is built into the web design project, which runs $2,500 to $15,000 depending on scope. Speed, accessibility and clarity are how the site gets designed here rather than a service sold alongside it, so there is no line item to remove and no version of the work that skips them. A standalone audit of an existing site — load time, accessibility failures, and the clarity problems costing you enquiries — is quoted as a one-off from the size of the site.
Sources
Where the numbers came from
Three sources, with their dates and sample sizes, because those decide how much weight a figure can carry. The traffic arithmetic above is not cited: it is computed from the standard formula for comparing two proportions, which is more checkable than a rule of thumb and happens to disagree with the popular one by about 7%.
- 1Milliseconds make millions
Google, Deloitte and 55 (via web.dev) · 24 June 2020 · 37 European and American brand sites, 30M+ user sessions, 30 days
- 2The WebAIM Million: an annual accessibility analysis of the top 1,000,000 home pages
WebAIM · February 2026 · 1,000,000 home pages, automated WCAG testing
- 3How to calculate sample size for A/B tests
Optimizely · 26 April 2023 · Methodology guidance from an A/B testing platform
Methodology from a company that sells A/B testing
Worth naming: the speed study is observational rather than experimental, it is from 2020, and it measured large brands rather than small businesses. It is the best evidence available tying a specific speed change to a conversion outcome, and it is not proof that the same holds for a plumber in Fort Collins. Last reviewed September 2026.
Find out what your site is costing you
Load time, accessibility failures and the clarity problems losing you enquiries — measured on your actual site, with the fixes ranked by what they are worth.