The work is built
in stages.

Conversation and diagnosis

I start with a conversation, not with the ad account. I need to understand where the business makes money, who the customer is, what sets this company apart from the one next door, and what happens after somebody calls. Without that, every campaign is guesswork, even when the numbers look neat.

Only then do I sit down with the data: the ad account, the website, the way conversions are counted, and what competitors are doing on the same queries.

The most common finding at this stage is broken measurement: conversions counted twice, a form that does not report submissions, or a tap on a phone number recorded as a customer. Until that is fixed, every decision rests on misleading data.

Conversation does not replace measurement, it sets it up. Without it you measure the wrong things. After it, it is clear what counts as a real result: not every enquiry is worth the same, and some of them should never appear in the summary at all.

A business does not need online sales to measure results. Calls, forms, visits to  premises, quote requests and the path from first click to signed contract can all be counted, including when the decision comes months later and offline.

Output of this stage: a list of what genuinely works, what is broken and what we simply do not know. The last of those is recorded as unknown, not as zero.

Plan

You get a list of decisions with reasoning and an order of play. For each one I write down how we will know it worked and how much time is needed to  judge it.

I separately list what we are deliberately not doing this month, so that the same ideas do not come back every fortnight. Scope is agreed before we  start, not along the way.

Implementation

Changes go in one at a time and dated. It sounds fussy, but it is the only way to attribute effect to cause a month later. When five things change on the same day there is a result but no knowledge to take from it.

Nothing reaches a client account without approval. For changes that are hard to undo, I record the state beforehand so there is somewhere to return to.

Reporting

A monthly report with a stated source for every number. If two systems disagree, I write down why instead of picking the version that looks better.

The report has to answer whether this is paying off, not bury you in charts. If the month was weak, the report says it was weak, along with what we are doing about it.

Rules I do not negotiate

• No change to a client account without their approval
• Every number in a report has a source I can point to
 
• Missing data is missing data, not zero
• The client retains administrative access; the accounts remain under their control during and after the engagement
• I do not promise positions or results that cannot be delivered

Let us talk about your account →