IN BRIEF
Conversion rate optimisation is the cheapest way to grow sales, because it buys no new visitors at all. Yet it is almost always started from the wrong end: testing a button colour when the biggest leak is three steps earlier. This article covers the order CRO should be done in, and what can be tested reliably even on modest traffic.
What CRO is and what it is not
Conversion rate optimisation is a method for finding out why visitors do not do what they are meant to do, and removing the obstacles in order. It is not a collection of tips, nor a single testing tool. A tool tells you which version won; the method tells you what is worth testing.
The effect compounds. If the conversion rate rises from two to three per cent, sales grow 50 % on the same media budget and the same visitor count. The same improvement cuts cost per acquisition by a third, which opens up channels that were previously too expensive.
Start with measurement, not tests
The first job is not a test but a check that you are measuring correctly. A surprising number of CRO projects start on broken data: a conversion recorded twice, internal traffic included in the figures, or a form’s thank-you page firing no event at all.
- Check that every goal is defined as an event in GA4.
- Exclude your own and your partners’ traffic.
- Make sure the same conversion is not recorded several times.
- Tag campaign traffic with UTM parameters consistently.
Rule of thumb: if you cannot say how many leads the site produced last month and which channel they came from, do not test anything yet. Fix the measurement first, or your test results are guesses.
Find the biggest leak, not the smallest
Draw the path stage by stage and count what percentage moves from one stage to the next. The biggest drop is always the first thing to fix, because a percentage point is worth more there than anywhere else. A typical e-commerce path looks like this.
| STAGE | MOVES ON | TYPICAL REASON FOR THE DROP |
|---|---|---|
| Landing page → product | 35–55 % | message does not match the ad |
| Product → cart | 8–15 % | unclear price or availability |
| Cart → checkout | 50–70 % | delivery costs come as a surprise |
| Checkout → payment | 60–80 % | forced registration, payment methods |
On a B2B site the same exercise is done around the form: how many who opened the page started the form, how many submitted it, and how many of those were sales-qualified. That last number often reveals that the problem is not volume but quality.
Four sources hypotheses come from
A good test starts from an observation, not an idea. Four sources are enough, and they are worth going through in this order, because they answer different questions: where, how, why, and what people say out loud.
- Analytics tells you where on the path visitors disappear.
- Heatmaps and recordings tell you what happens on the page before they go.
- An on-site survey tells you what stayed unclear. One question is enough: what stopped you from continuing?
- Sales calls tell you what the customer asks when the page did not answer.
Prioritise before you test
Once you have dozens of observations, you need an order. A simple score is enough: rate each idea for impact, confidence and effort on a 1–5 scale, and start with those where impact and confidence are high and effort is low. The name of the method is secondary; what matters is that the decision is argued.
Write every test as a hypothesis: if we change X, conversion will rise, because Y. The reasoning is the important half. Without it a win teaches you nothing and cannot be carried to other pages.
How to test on low traffic
Most Finnish sites do not have the traffic for classic A/B testing. A reliable result needs roughly 100 conversions per variant, and if you get 20 leads a month, the test would run for a year. The answer is not to stop measuring but to change the method.
- Test big changes in sequence: a new page for a month, compared to the previous period.
- Measure intermediate goals that happen more often: form starts, video views.
- Combine similar pages into a single test.
- Use qualitative material: five user tests reveal most of the obstacles.
A warning: do not stop a test on the day it looks like it is winning. Stopping early is the most common reason an improvement disappears the following month.
Where the wins usually are
Although every site is different, the same fixes repeat from project to project. These are worth checking before inventing anything of your own.
- Show a price or a price range, even if it is only a starting price.
- Cut form fields and state clearly what happens next.
- Show delivery costs and return terms before the checkout.
- One call to action per page, repeated in several places.
- Put mobile forms and buttons within thumb reach.
Read also: A landing page that converts
Frequently asked questions
Where should conversion rate optimisation start?+
With a measurement check. Once you know how many conversions arrive and from where, draw the path stage by stage and fix the biggest drop first. Testing comes third.
What is a good conversion rate?+
In e-commerce 1–3 % and on a B2B lead form 2–5 % are typical. The figure varies so much by industry that your own baseline is the only sound comparison.
Can CRO be done without a testing tool?+
Yes. Measurement, recordings, an on-site survey and user tests produce most of the observations. A testing tool is needed only once there is enough traffic to compare.
How long should one test run?+
At least two full weeks so day-of-week variation evens out, and long enough for roughly 100 conversions per variant. A shorter test reports random variation.
SOURCES AND DATA
VSCY client project conversion data 2024–2026 · GA4 funnel and path analyses · heatmap and session recording datasets. Reviewed twice a year.
