Every month we have first meetings with a few new clients, and the conversation nearly always starts in the same place: "We're doing things, but we don't know whether any of it works." That sentence describes a measurement problem, not a budget problem.
You can't manage what you don't measure
The great promise of digital marketing was that it could be measured. In practice most teams report rather than measure: impressions, clicks, follower growth. Those are activity indicators, not outcome indicators. "We got thirty thousand impressions" leads to no decision at all.
A metric that drives decisions is one that connects an action to a result. How many people completed the form, how many converted to a sale, what that sale cost. Without those three, everything else is decoration.
Don't start without a baseline
The first thing we do on a new engagement is freeze the current state. Site speed, search visibility, conversion rate, cost per channel. Those numbers are the only way to answer "did we improve?" in month three.
Where work begins without a baseline, everyone tells their own story. The agency says traffic went up, the client says sales didn't, and both are right — because there is no common ground to compare against.
Think about channels together, not separately
Search, social and advertising look like separate budget lines, but to the user they are parts of one journey. If someone arriving from one never sees the brand again in another, trust doesn't form.
In the last two years another layer has been added: AI assistants. Users now put the question straight to an assistant, and whether or not your brand appears in the answer may not even register as a click. Starting to measure that surface means widening your measurement set.
Test small and often
- Change one variable. Change the headline, the image and the audience at once and you'll never know which one moved the result.
- Wait long enough. Deciding on three days of data is gambling. For most campaigns two weeks is the floor.
- Record the losing tests too. Experiments that didn't work are the best answer to whoever proposes the same idea again six months later.
- Open the budget gradually. Don't put a large budget into a new channel from the start; begin small and increase as signal appears.
Don't bury slow channels early
Search visibility and content work show almost nothing in the first two months. Advertising produces numbers in the first week. That asymmetry pushes teams to drift towards advertising, because ads give a measurable feeling of success and content does not.
But the day you stop advertising, the traffic stops too. Content and technical foundations compound: after month six they are the only channel whose cost falls. The right setup runs both — ads fund today, content funds next year.
A practical rule: most of the budget goes to the channel producing sales today, and a fixed slice to the one that doesn't yet but accumulates. Cutting that slice at the first squeeze means starting from zero in the same place a year later.
The line between your team and the agency
The most productive engagements are the ones where it is written down who owns what. Product knowledge, industry language and customer objections sit best on the client side; production, measurement and optimisation on the agency side.
When that line stays vague, one of two things happens: both sides do the same work, or neither does. Drawing up a responsibility table in the first meeting saves half of the next six months.
Keep the report short
A forty-page monthly report is a document nobody reads. Ours has three headings: what worked, what didn't, what changes next month. The raw data sits underneath as an appendix for anyone who wants it.
The format helps discipline as much as readability. Having to fill in "what didn't work" every month stops teams from quietly continuing things that don't.
What to expect when working with an agency
A good agency puts what it will do in writing and shows you what it did at the end of the month. It doesn't take on work beyond its capacity. It opens source files and account access in your name, and holds nothing hostage if you part ways.
None of that is an extraordinary demand; it should be the industry's minimum standard. Ask about these four in the first meeting — the answers explain a great deal.
