Double fires
One event bound twice on the same button, or still bound to a element that shipped out months ago. Both look healthy on a dashboard.
Found in most accounts we inherit
Analytics
you can trust.
Analytics is not a dashboard. It is being able to ask a question about your business and believe the answer. Most teams we meet cannot do that yet. We fix the measurement first, then run the experiments.
- Definitions One owner per event
- Plumbing Fixed before it is charted
- Dashboards One page per audience
- Evidence Losses reported like wins
The measurement plan
We start by writing down what you need to know. Then we name the events that answer it. That document becomes the source of truth for the whole team, and it is yours whether we keep working together or not.
One event, one definition, one owner. Most disagreements about numbers come from definitions that were never written down.
- The question the business needs answered
- The event that answers it, named once and only once
- A written definition and a named owner for that event
- Where it sits in the funnel, and what it is compared against
What we usually find
Before anything gets tested it has to be counted correctly. These four faults turn up in nearly every account handed to us, and every one of them quietly changes a decision someone already made.
Three answers
The same conversion counted in the ad platform, the analytics tool and the CRM. Three numbers, no agreement, and a meeting every week about which one is right.
Pick one system of record
Orphan clicks
An ad click that never links to a CRM record. You can see that traffic arrived. You cannot say what any of it was worth.
Repaired at the handoff
Inflated traffic
Bots, internal sessions and preview loads counted as users. Every rate downstream is then wrong by an amount you cannot measure.
Filtered at collection
Where browsers block or drop data we move collection to the server, which is server side tracking work. The container itself is wired in Google Tag Manager, and the reporting layer is a standard GA4 analytics setup. All of it sits with our own engineers, so it ships in days.
Dashboards people use
One page per audience, and nothing on it that cannot change a decision. Any chart that does not support a decision is removed.
-
Founder view
Revenue, cost per outcome and payback. The three numbers you would quote to a board, on one page, with the definitions attached.
-
Product view
Activation, retention and the step where people stall. It answers what to build next, not how the last quarter went.
-
Channel view
Spend, pipeline created and cost per qualified lead. Enough to decide a budget on a Monday without opening four tools.
When the volume outgrows the reporting tool, the numbers move into BigQuery and warehousing and the dashboards read from there instead.
Before you ask
How long until we can trust the data?
The measurement plan takes about a week to write and a week to ship. After that every number on the page has a definition and an owner behind it.
Do we need a new analytics tool?
Usually not. What is broken is almost always the collection, not the tool. We fix your stack before we suggest replacing any part of it.
Who writes the test?
We do. The change worth testing is usually a code change, so our engineers ship it. No one waits for another supplier's sprint.
What happens when a test loses?
It gets reported the same way a win does. A loss shows where the demand is not, which is also worth knowing.
That is the advantage of one team: the change we want to test is usually a code change, and we write it ourselves — see web development. Clean numbers are also what makes performance advertising safe to scale. The full map sits on the services page.
Start with the measurement
Send us the stack you are running and the question you cannot answer. We reply with what is broken, what it takes to fix it and who would do the work.