Analytics and Tracking
Measurement comes before optimization, because every decision downstream rests on it. Almost every account we open has at least one conversion action that is wrong, and until that is fixed nobody in the room is discussing real numbers.
What goes wrong with conversion tracking
The same handful of faults, repeatedly, and all of them are invisible unless somebody checks.
- A thank-you page counted on every load. One enquiry becomes four when the visitor refreshes or returns via the back button.
- Spam counted as conversions. Bot submissions inflate the total until the real rate is unknowable.
- A form that reports success without sending. The tracking fires on the button, not on the delivery, so the numbers look healthy while nothing arrives.
- Every form treated identically. A newsletter signup and a quote request recorded as the same event, so the metric averages two unrelated things.
- Tracking that never fires on mobile. Frequently a script that fails silently on one browser.
The consequence is not just bad reporting. Ad platforms optimise toward whatever you tell them a conversion is, so a broken signal actively directs budget at the wrong people.
How to check whether tracking is honest
By reconciling against something outside the analytics. Count the enquiries that actually reached a person over a month and compare that with what the platform reported. If those numbers are far apart, the reported one is not usable.
That comparison is cruder than a technical audit and it settles the question faster. Everything else follows from knowing whether the number can be trusted.
What a B2B site should measure
| Event | Why it earns its place |
|---|---|
| Qualified enquiry | The one that matters. Defined so it excludes spam and job applications |
| Phone call started | On a considered industrial sale, often the main route, and usually untracked |
| Specification or document opened | A genuine buying signal well before anyone makes contact |
| Form started but abandoned | Distinguishes "nobody wanted to" from "the form defeated them" |
| Traffic source, honestly grouped | Assistant referrals frequently land in Direct, so a plain channel report understates them |
Resist tracking everything. An account with forty events has no hierarchy, and the important one gets lost among scroll depth and video plays nobody acts on.
Platform numbers and analytics disagree by design
Because they attribute differently and they count different things. An ad platform credits itself for a conversion within its attribution window; your analytics assigns it to the last channel it saw. Neither is lying and they will never reconcile exactly.
The workable approach is to use platform figures for optimising inside that platform, analytics for comparing channels, and enquiries that actually arrived as the number the business is judged on. Trouble starts when one of the three is presented as the truth.
The traffic you cannot see
A significant share of visits arriving from AI assistants carry no referrer and land in Direct. Analytics also only ever records a citation somebody clicked; a brand mentioned in an answer that shaped a decision leaves no trace at all.
So Direct is not a synonym for "typed the URL", and a channel report is a floor rather than a measurement. Where the assistants are identifiable we group them explicitly, and where they are not we say so rather than reporting a number we cannot support.
What the work involves
- Audit what is installed and what each event actually records, as opposed to what it is named.
- Define conversions that map to business outcomes, with spam and internal traffic excluded.
- Implement and verify, including on mobile, by submitting real test enquiries end to end.
- Group channels so assistant and AI referrals are visible rather than absorbed into Direct.
- Reconcile monthly against enquiries that genuinely arrived.
Where this fits
It is the first stage of conversion optimization and the prerequisite for testing, which cannot work against an unreliable signal. It also underpins paid media, since campaign optimisation is only as good as the conversion data feeding it.
Frequently asked questions
Do we need a tag manager?
Usually, once there is more than one thing to track. It keeps tracking out of the site's code and lets changes happen without a deployment. It is also easy to fill with forgotten tags nobody will admit to owning, so it needs occasional pruning.
Will adding analytics affect our privacy policy?
Yes, and it has to be updated before the tools go live. This site currently runs no analytics at all, which the Cookie Policy states plainly. Adding any changes that, and in some jurisdictions triggers consent obligations.
Can you fix tracking without rebuilding the site?
Almost always. Tracking sits alongside the site rather than inside it. The exception is a form that reports success without delivering, which is a fault in the form.
See where you stand
We will review your current search visibility and show you what buyers find when they look for what you do.
Contact Us