Almost every piece of conversion rate optimisation advice assumes you can run a clean A/B test. If your site gets a few hundred or a few thousand visits a month, you cannot. A test on that traffic would need to run for months, and over months your traffic mix, your offer, and the season all change, so the result is not a result.
That is not a reason to leave the site alone. It means the method changes. You take evidence from watching people instead of counting them, you fix what is obviously broken without asking permission from a test, and you ship changes one at a time so each one is readable.
Why the sitewide conversion rate is the wrong number
A single conversion rate for the whole site averages together visitors who have nothing in common: someone who typed your company name into Google, someone who clicked an ad, someone three steps away from buying who landed on a blog post. When that number moves you cannot tell whether anything improved or whether your traffic mix shifted. At small volumes it also swings week to week for reasons that have nothing to do with your site.
Measure steps instead. Landing page to form view. Form view to submit. Submit to a qualified conversation. Conversation to deal. Each is a specific question with a specific fix, and each is readable at much lower volume than a sitewide rate.
Fix the obviously broken things without testing them
Some things do not need evidence. If the form throws a validation error with no visible message, if the phone number is not a tappable link on mobile, if the primary button sits below three screens of scrolling, if the thank-you page is a blank white screen — fix it. Running a test to confirm that broken is worse than working is a way of not doing the work.
Start by going through your own site as a stranger would, on a phone, on cellular data, from the ad or search result rather than from your bookmarks. Write down everything that confuses or stalls you. That list is your first sprint.
Get your evidence from people, not from traffic
With low traffic, qualitative evidence is better evidence. Three sources, in order of how much they are worth:
Sales-call objections. Whoever takes your inquiries already knows what the site failed to explain, because they explain it every day. Ask for the five questions they answer most often on a first call. Those are the five things the site is not saying clearly, and the fix is copy, not layout.
Session recordings. Twenty recordings of people on your highest-intent page show you more than a month of aggregate numbers. Look for specific behaviour: scrolling past the thing you expected them to click, pausing on a form field, hunting for pricing.
A one-question exit survey. On the page where people drop out, ask what they were looking for and did not find. The answers are few and repetitive, which is what makes them useful.
None of this is statistically valid, and it does not need to be. You are not trying to prove a small lift, you are trying to find the sentence that is missing.
Ship changes in sequence, not in parallel
Pick one change. Note the date, what you changed, and the step-level number you expect it to move. Leave it for a few weeks. Then do the next one.
This is weaker than a controlled test and worth saying so out loud: anything that happened in the market during those weeks is in your numbers too. What it gives you is a legible history. Six months later you can read twelve dated changes and see which two coincided with a real shift. Six changes shipped together give you nothing to read.
Sequence by how cheap the change is and how confident you are. Broken things first, then copy, then layout, then anything needing a build.
Cut the form until it hurts
Every field you add costs you submissions and buys you information. Most small-business forms are losing more than they buy.
Go through the fields one at a time and ask who reads this, and when. If the answer is that sales looks it up during the call anyway, or that nobody has used it, delete it. Company size, job title, how did you hear about us, and a required phone number are the usual suspects. You can ask on the call.
Two other things matter more than field count. Say what happens after submit, specifically — who replies, and how fast. And make errors survivable: show the message next to the field, keep everything they already typed, and never clear the form.
Decide what the top of the page has to establish
Before any styling argument, the top of the page has to answer three questions a stranger has: what this is, who it is for, and what to do next. If a visitor cannot answer all three without scrolling, nothing further down gets read.
That usually means plain language over a clever line, one obvious action rather than four competing ones, and whatever proof you actually have placed where it is seen rather than in a carousel further down. If you have no proof yet, say what you do precisely instead — specificity reads as competence.
Treat page speed as a conversion problem
Speed gets filed under SEO, which is why it stays unfixed. The conversion cost arrives earlier: a page that takes several seconds on a phone on a weak connection loses people before they have read a word, and they never appear in your funnel as anything but a bounce.
You will not see this on a desktop on office wifi. Throttle the connection in your browser devtools, or load the site on your phone away from wifi. The usual causes are uncompressed hero images, third-party scripts nobody audits, and font loading that leaves the page blank. Each is fixable without a redesign.
Fix what happens after submit
The highest-leverage conversion work on most small sites is not on the site. A form submission that gets a reply within the hour and a submission that gets one in two days are two different businesses, and only one of them is measured in your conversion rate.
Check three things: that the notification actually arrives at a person who is looking at it, that there is a named owner for replying, and that the first reply does something more than acknowledge receipt. Sales enablement work belongs in the same sprint as the form.
What this will not fix
If the offer is wrong, or you are buying traffic from people who were never going to buy, no amount of page work rescues it. When the step-level numbers are poor all the way across, the problem is usually upstream of the site: positioning, channel choice, or who you are targeting.
Where to start
Walk your own site on a phone, from an ad or a search result, and write down everything that stalls you. Then ask whoever takes your inquiries for the five questions they answer most. You will have a prioritised list by the end of the day without opening an analytics dashboard.
If you would rather have that done properly, the 30-Day Marketing Diagnostic covers the funnel, the data, and the site as a fixed-scope piece of work, and comes back with the sequence rather than a list of ideas. Talk to a CMO if that is the version you want.
