dotCREA
TREN

CRO: raising conversion rate in a measurable way

The way to grow revenue without growing traffic: measure where the loss happens, and only implement changes you can measure.

What does this service cover?

The most common mistake in conversion work is making many changes at once without seeing which one worked. Even if the result improves, it cannot be repeated because nobody knows why.

We start by establishing where the loss is: how many people drop at which step, on which device, from which traffic source. Changes are ordered against that data and measured one at a time.

Measurement accuracy first

Badly configured analytics is the source of bad decisions. We always start here:

Verifying that conversion events fire once and correctly
Separating device and traffic source breakdowns
Reporting campaign periods separately from normal ones
Tracking net conversion after returns as well

Where the returns are biggest

The patterns that repeat across the stores we measure:

Unexpected cost appearing late at checkout: shipping should be stated in the cart
Guest checkout made difficult or hidden
Breaks in the form and 3D Secure flow on mobile
Missing decision information on the product page: size, delivery time, return terms
An unclear delivery promise
No cart abandonment recovery chain at all

What these have in common: none of them is a matter of design taste, and all of them are measurable.

How we work

01
Analytics audit
The accuracy of existing tracking is checked; missing or duplicate events are fixed.
02
Loss map
The drop at each step of the funnel is established, broken down by device and source.
03
Hypothesis list
Expected impact and measurement method are written for each fix, and work is ordered by impact.
04
Implementation and testing
Changes are made one at a time, with A/B tests set up wherever traffic is sufficient.
05
Reporting
Results are reported before and after; anything that did not work is rolled back.

Frequently asked

A meaningful result needs a certain order volume. On low-traffic stores, rather than running tests we apply the fixes whose value is already known and measure before and after — and we say so from the start.