//Analytics & cro
Fix the signal, then fix the page
Analytics and conversion work is the least glamorous thing we sell and the one that changes the most numbers. First we make the data honest — events, consent, server-side where it matters, one definition of a conversion everyone agrees on. Then we work on the page the traffic lands on, which is usually where the money was going missing all along.
What this work covers...
GA4 and tag setup, rebuilt properly
Best for a brochure site, landing page, or clearly scoped feature. Fixed price, fixed timeline.
Server-side tracking where it earns its cost
Best for a brochure site, landing page, or clearly scoped feature. Fixed price, fixed timeline.
Consent that is legal and still measures
Best for a brochure site, landing page, or clearly scoped feature. Fixed price, fixed timeline.
Dashboards, and a number everyone trusts
Best for a brochure site, landing page, or clearly scoped feature. Fixed price, fixed timeline.
Landing pages rebuilt, not just recommended
Best for a brochure site, landing page, or clearly scoped feature. Fixed price, fixed timeline.
Checkout and form work, measured after
Best for a brochure site, landing page, or clearly scoped feature. Fixed price, fixed timeline.
20–40%
of spend a first audit typically finds going somewhere useless
9 / 10
accounts we inherit have at least one double-counted conversion
2–3 wks
to a measurement setup you can make decisions on
1
definition of a conversion, agreed and written down
//the order
Why measurement comes before optimisation
Running a conversion test on broken tracking is worse than not testing: you get a confident answer that happens to be wrong, and then you build a quarter of decisions on it. So the sequence is fixed, and we will not skip a step because the second one is more fun to talk about.
//always first
Make the data honest
One event fires once, a lead means the same thing in every tool, and the platforms stop claiming the same order three times between them.
events, deduplication, naming
consent mode and what it costs you
a written definition of a conversion
//then
Find where people leave
Funnel by step, by device, by source. Usually one step is doing most of the damage, and it is rarely the one the team was arguing about.
step-by-step drop-off
session recordings, read not collected
speed, because mobile latency is a conversion problem
//last
Change it, then measure
We rebuild the page rather than writing a document about it. On low traffic that means shipping a considered change and watching; on high traffic it means a real test.
page, form and checkout rebuilds
A/B only where the volume supports it
the result written down, including the failures
//what we actually find
The same eight problems, in almost every account
There is very little mystery in this work. After enough audits the same list comes back, which is why the first two weeks are mostly mechanical and the value shows up immediately afterwards — the campaigns start being optimised against numbers that are true.
None of this is exotic. It is just work that sits between marketing and engineering, which is exactly the gap where things rot when the two are different companies.
[01]
The same purchase counted twice
a tag on the thank-you page and a platform pixel, both claiming it
[02]
A newsletter signup counted as a lead
so bidding optimises towards people who will never buy
[03]
Consent blocking more than it has to
a banner nobody configured, quietly deleting half the data
[04]
Internal and agency traffic in the numbers
small sites where the office is a meaningful share of sessions
[05]
Phone and chat conversions invisible
so the channel that drives them looks like it does nothing
[06]
Refunds and cancellations never sent back
revenue in the ad platform that the accountant has never seen
[07]
A mobile checkout nobody has used
on a real mid-range phone on a real network
[08]
Four dashboards, four different truths
and a monthly meeting spent arguing about which is right
//the objections
What you are probably thinking
This is the service clients are least excited to buy and most relieved to have bought.
“Our analytics is already set up.”
Set up and correct are different states. The test is simple: pick last month, and see whether the ad platform, GA4 and your invoices agree on how many orders there were. If they disagree by more than a rounding error, every optimisation decision since then was made on fiction.
“CRO means endless A/B tests.”
Not at your traffic, usually. A statistically real test needs volume most Baltic sites do not have, and running one anyway produces a confident coin flip. Below that threshold the honest method is to fix the obviously broken thing, ship it, and watch a month of data.
“Can you just tell us what to change?”
We can, and we will if you have your own developers. But the audit-as-PDF model is how most CRO money is wasted: the report is read, half of it gets built six months later in a slightly different form, and nobody can tell what worked. We would rather build it.
“Is server-side tracking worth it?”
Sometimes. It recovers signal that browsers now block, and it costs real money to run and maintain. Under a certain spend the recovered data does not pay for the infrastructure, and we will say so rather than sell you a container you did not need.
“Will this get us in trouble with GDPR?”
The opposite, usually. Most setups we inherit are both leaky and non-compliant at the same time. Consent mode configured properly means you measure what you are allowed to measure, model the rest, and stop collecting things nobody ever looks at.
“Our numbers will look worse afterwards.”
Very likely, and it is worth preparing your management for it. Deduplicated conversions mean the dashboard drops while the business does not change. The first honest month is uncomfortable; every decision after it is better.
//case studies
Three fixes, anonymised
None of them involved buying more traffic. Client names withheld under NDA.
#analytics
The dashboard dropped, the revenue did not
Two tags were counting every order. Once deduplicated, the reported conversions fell by a third and the real cost per order turned out to be well above target — which finally explained a year of confusing meetings.
reported conversions
down ~33%
actual orders
unchanged
fix time
4 days
#cro · #dev
A form nobody could finish on a phone
Eleven fields, a date picker that failed on older Android, and no way to see the error. Rebuilt to four fields with the rest asked after the lead was captured. Mobile carried most of the traffic and almost none of the leads.
fields
11 → 4
mobile completion
up sharply
ad budget changed
not at all
#analytics
The channel that looked dead was the best one
Enquiries arrived by phone and were never tracked, so search looked like it produced nothing and was almost switched off. Call tracking and CRM sync put the closed deals back against their source.
untracked before
most enquiries
decision avoided
cutting the channel
setup time
2 weeks
//what it costs
A project, then optionally a habit
Most clients buy the rebuild once and stop there, which is a perfectly good outcome. The ongoing option exists because measurement rots: a developer ships a change, a tag breaks, and nobody notices for two months.
TRACKING AUDIT
free
3–5 days
What is double-counted, what is missing, and which of your numbers can currently be trusted.
MEASUREMENT REBUILD
from €2K
2–3 weeks
Events, consent, deduplication, dashboards and a written definition of every conversion.
CONVERSION WORK
from €990
per month, or per page
Pages and forms rebuilt, tests where volume allows, and a monthly check that the tracking still works.
//questions
Analytics and CRO questions, answered
How much traffic do we need for A/B testing?
Enough conversions, not enough visits — as a rough floor, a few hundred conversions per variant before a result means anything. Most sites we work with are below that, so we ship considered changes and read the month instead of pretending a two-week test was conclusive.
Do you work with our existing developers?
Often, and it works well when there is a named person on their side. We can specify and they implement, or we implement inside your repository if you would rather. What does not work is a specification handed into a queue with no owner.
Can you track offline conversions?
Yes, and for B2B it is usually the whole point. Calls, CRM stages and signed deals can be sent back to the ad platforms so bidding optimises towards revenue rather than form submissions. It needs your CRM to be reliable, which is sometimes the first project.
What do we get at the end?
A working setup in your own accounts, a document naming every event and conversion and what it means, dashboards your team can read without us, and a list of what we deliberately did not track and why.
Is this included if we hire you for ads?
The audit and the ongoing checks, yes — we will not run campaigns against numbers we do not believe. A full rebuild or significant page work is quoted separately, because it is engineering rather than campaign management.
//tools
What we work in
our own dashboards
Marketing support
where this work sits inside a monthly retainer
Google Ads
the channel that suffers most from bad measurement
Web development
when the page needs rebuilding rather than tweaking
//contact us
Give us read access. We will tell you which numbers to believe.
Three to five days, free, and the findings are yours whatever you decide. We reconcile the ad platforms, GA4 and your actual orders for one month, and tell you where they disagree and why.