Method

1. Record the baseline first

The measurement taken before anything changes is the only one that cannot be reconstructed afterwards. It is the first deliverable on every engagement, and where a baseline cannot be established we say so before starting rather than reporting on activity instead.

2. Write the expectation down

Every change carries a stated expectation: what should move, by roughly how much, and by when. Written in advance, it is a test. Written afterwards, it is a description — and descriptions always fit.

3. Measure at 7, 14 and 28 days

Three windows, because a single reading cannot separate a result from ordinary variation. This is also why retainers carry a three-month minimum: a one-month engagement cannot demonstrate anything, and charging for one would be selling a report rather than a result.

4. Measure independently of the channel being judged

An advertising platform reporting on its own performance has an obvious interest in the answer. Conversion tracking is set up so the figures we report can be checked outside the system that produced them.

5. Publish the negative results

An approach that did not work is a finding. We record it, stop, and say so. A programme that only ever reports success is not measuring anything — and it makes every other number it reports worth less.

What we will not do

• Report platform-attributed conversions as though they were incremental
• Publish a client name, logo, testimonial or result without written permission and evidence on file
• State an industry average as though it were our own result
• Present a projection as a measurement, even when it is labelled as one
• Judge a programme by how much it published

Where this is being tested hardest

We are building a system that runs this loop with as little human time as possible, and testing it on our own site before anywhere else. Its baseline was recorded at launch. It has produced no results yet, and this site claims none.