Tracking audit & rebuild
GA4 and Tag Manager checked end to end. Duplicate events removed, conversions verified against real orders, bot and internal traffic filtered.
Analytics & CRO
Analytics & CRO
Service / 06
You already have the traffic
Doubling conversion rate has the same effect as doubling ad spend, and costs considerably less. But you can't improve what you're measuring wrongly — which is why this work always starts with the tracking.
The problem
Three tools, three different numbers
The ads platform claims forty conversions. Analytics shows twenty-three. The payment processor says nineteen. Nobody can say which is right, so the monthly meeting becomes an argument about measurement rather than a decision about budget.
Underneath that, the usual culprits: events firing twice, a purchase tag that also fires on refresh, cross-domain tracking broken at checkout, and bot traffic nobody filtered. None of it is visible on a dashboard — the dashboard looks fine, which is the problem.
Every decision you make from the wrong number is a confident decision in the wrong direction.
And without trustworthy numbers, you can't run a test either. A/B testing on broken measurement produces conclusions that feel rigorous and aren't — which is worse than not testing, because now you're acting on them.
What's included
Six parts of the work
In this order. Testing before the measurement is fixed just generates expensive noise.
Server-side tracking
Conversions API and server-side tagging where browser signals have stopped being reliable — which, with current blocking, is most places.
Funnel analysis
Where people actually leave, and what it costs you. Usually one or two steps account for most of the loss, and they're rarely the ones people guess.
Qualitative research
Session recordings, heatmaps and form analytics. Numbers tell you where people leave; watching tells you why, which is what you need to fix it.
Testing programme
Hypotheses from the research, prioritised by expected impact, run to statistical significance rather than stopped when the result looks nice.
Reporting you can read
One view, one set of numbers, in plain language — including the tests that failed and what they ruled out.
Where the money goes
A typical funnel, and its leaks
An illustrative e-commerce funnel. The percentages vary by business, but the shape rarely does — and the biggest single loss is almost always the same step.
Visitors land10,000
View a product4,200
−58% Usually a landing page or relevance problem, not a product one
Add to cart1,050
−75% Price clarity, shipping cost, trust signals, photography
Begin checkout620
−41% Forced account creation and surprise costs, most often
Purchase190
−69% Payment friction, form errors, and mobile layout failures
Lifting checkout completion from 31% to 40% here adds about 55 orders a month without a rupee more in ad spend. That's the arithmetic that makes CRO worth doing — and why we measure the steps before changing anything.
How it runs
The first ninety days
Weeks 1–2 Audit
Find out what's lying to you
Every event checked against real orders. In most accounts we find at least one conversion counted twice and one not counted at all — which is usually enough to change what the business thought was working.
Weeks 2–4 Rebuild
Fix the measurement
Clean GA4 and Tag Manager setup, server-side tracking where needed, one agreed reporting view. Nothing downstream is trustworthy until this is done.
Weeks 3–6 Research
Watch what people actually do
Funnel analysis plus session recordings and heatmaps. This stage usually produces several obvious fixes that need no testing at all — broken mobile layouts, forms that reject valid input.
Week 6 onward Testing
Test the things that aren't obvious
One test at a time, run to significance. Most fail — that's normal and useful, because a failed test rules out a direction you'd otherwise have argued about for a year.
Month 3+ Compounding
Small wins stack
A few percent at each step multiplies through the funnel. And every improvement makes your paid media cheaper, because the same spend now produces more orders.
What we won't tell you
That every test will win. Most won't — somewhere between seven and nine in ten fail or come back flat, across the whole industry. Anyone showing you only winners is showing you a selection, not a record.
We also won't run A/B tests on low traffic. Below a few hundred conversions a month, a test takes so long to reach significance that the market changes before it finishes. At that volume, qualitative research and fixing obvious problems is the honest approach, and it's what we'll recommend.
And sometimes the answer isn't the page. If the traffic is badly targeted, no amount of button colour fixes it — the problem is upstream in the ads, and we'll say so even though it's not what you hired us for.
What you get
Every month, without asking
- A tracking audit listing every discrepancy found and what it was costing you.
- A clean GA4 and Tag Manager setup, documented, in your accounts.
- One agreed reporting view so the numbers stop disagreeing.
- A prioritised test backlog with expected impact and effort marked.
- Full test results including the ones that failed and what they ruled out.
- Direct access to the person running the programme.
Questions
About analytics and CRO specifically
For A/B testing, roughly a few hundred conversions a month — below that, tests take months to reach significance and the result is stale by the time it arrives. The tracking and research work is worth doing at any volume, and at lower traffic it's the whole programme rather than a prelude to testing.
We won't quote a number — it depends entirely on how much is currently broken, and we don't know that until we look. What we can say is that the first round of fixes, the obvious ones found in research, usually produces more than the testing that follows.
Because a test measured wrongly gives you a confident answer that happens to be false — and you then act on it. Duplicate events alone can make a losing variant look like a winner. Fixing measurement first is slower to start and the only way the rest of the work means anything.
For most tracking and test work, no — Tag Manager and testing tools handle it. Permanent changes to winning variants are better built properly into the site, so we'll hand your developers a clear spec at that point, or do it ourselves if we built the site.
Tracking gets built to respect consent, with modes configured so analytics degrades properly rather than breaking when someone declines. Where regulations apply to your market, that shapes the setup from the start rather than being retrofitted after a complaint.
You do. GA4, Tag Manager and testing tools stay in your accounts with you as owner, and the configuration is documented. If we part ways, your developer or next agency can pick it up without rebuilding.
Also available
Pairs well with
Start with the tracking
We'll check every conversion event against your real orders and tell you which numbers you've been trusting that you shouldn't. You keep the findings whether or not you work with us.
No pitch deck, no pressure. Just a conversation about your numbers.