Most agency reports are written to survive a meeting, not to inform a decision. Impressions are up, engagement is healthy, the campaign “performed well”. Then your CFO asks what it did for revenue and the room goes quiet. We measure marketing the way your finance team measures everything else: against a baseline, with a clear line to the business goal, and with honest notes on what the data can and cannot prove.
Before any creative, media or search work begins, we agree with you which numbers matter, how each is calculated and where the data comes from. That conversation takes a few hours. It saves months of arguing about whether something worked.
The metric tree we agree before work starts
A metric tree is a single page that shows how every marketing number connects to the one your board actually cares about. It stops teams optimising a figure that does not matter, and it makes trade-offs visible. Ours usually runs in six layers.
The business goal
Revenue, margin, share or a strategic objective, stated with a number and a date. Everything below must explain how it contributes.
Market size
TAM, SAM and SOM, built bottom-up from buyer counts and realistic deal values, so growth targets sit inside a market that can support them.
Demand
How many people or accounts are in the market now, how many know you, and how many are searching for what you sell.
Conversion by stage
Visit to lead, lead to opportunity, opportunity to customer, or basket to checkout. Each step gets its own rate, so you can see exactly where buyers leak.
Unit economics
CAC, CAC payback, LTV and the LTV:CAC ratio, calculated on gross margin rather than revenue, so marketing is judged on profit it creates.
Brand health
Awareness, consideration, share of search and mental availability. The slow numbers that decide what next year’s acquisition costs.
Every term on the tree is defined in our marketing metrics glossary, and you can test your own unit economics with the CAC and LTV calculator before we ever speak.
Which marketing metrics actually matter?
The ones that answer a question someone with a budget is asking. This is the core set we work from. Which of them lead the read-out depends on your business model.
| Metric | What it answers | Formula | When it matters |
|---|---|---|---|
| CAC | What does a new customer cost us? | Sales and marketing cost ÷ new customers in the period | Always; the anchor of unit economics |
| CAC payback | How long until a customer repays their cost? | CAC ÷ (monthly revenue per customer × gross margin %) | Subscription and repeat models; cash planning |
| LTV | What is a customer worth over the relationship? | Average revenue per period × gross margin % × expected periods retained | Setting how much you can afford to acquire |
| LTV:CAC | Is growth creating value or burning it? | LTV ÷ CAC | Board and investor conversations |
| ROAS | What revenue did the platform attribute to this spend? | Attributed revenue ÷ ad spend | Day-to-day channel optimisation |
| MER | Is total marketing paying its way? | Total revenue ÷ total marketing spend | Consumer brands running many channels |
| Conversion rate | Where in the funnel are buyers leaking? | Conversions ÷ visitors (or leads) at that stage | Website, checkout and pipeline reviews |
| Win rate | How often do qualified deals close? | Deals won ÷ deals closed (won plus lost) | B2B with a sales team |
| Share of search | Is demand for our brand growing faster than rivals’? | Our brand searches ÷ brand searches for all brands in the category | Brand investment cases; an early read on share |
| Incremental lift | What happened that would not have happened anyway? | (Test group result minus control group result) ÷ control group result | Before scaling any channel |
Different businesses, different scorecards
A single template for every client is a sign that nobody thought about yours. We build the scorecard around how your money is made.
Consumer brands. ROAS for channel tuning, MER as the honest overall check, plus repeat purchase rate, average order value and share of search. A brand that lives on marketplaces also needs the marketplace’s own numbers in view, because a 200-pixel thumbnail and a review count can move conversion more than a bid change.
B2B and industrial. Qualified pipeline created, win rate, sales cycle length, cost per qualified lead and, for engineering-led businesses, specification wins: the moments your part or material is written into a customer’s design or tender. Form fills are not on the list. A thousand downloads from students will never survive a sales review.
Corporate and internal communications. Reach (did the right people receive it), comprehension (did they understand it) and action (did they do what was asked). For a safety campaign on a factory floor or a change programme across countries, these three tell you far more than open rates.
How we set baselines
Nothing can be called an improvement without a starting point. In the first weeks we pull at least twelve months of history where it exists: ad platform data, web analytics, CRM stages, sales by channel and, if you have them, past brand tracking results. We note seasonality, price changes, launches and stock problems, because a strong March means little if last March had a promotion.
Where data is missing or broken, we say so in writing, fix the tracking first and agree a shorter baseline rather than inventing one. A baseline built on a broken pixel is worse than none.
Could your current agency explain your CAC payback in one sentence?
Send us a briefIncrementality, brand lift and marketing mix modelling
Platform dashboards report what they can see, and they tend to see themselves generously. To find out what marketing truly caused, we use three tools, chosen by budget and question.
- Holdout and geo tests. Hold back a share of your audience, or a set of comparable regions, and compare outcomes against those who saw the campaign. This is the closest marketing gets to a controlled experiment, and it is how we decide whether a channel deserves more money.
- Brand lift studies. Survey exposed and unexposed groups on awareness, consideration or message recall. Several platforms offer these for qualifying campaigns; for larger programmes we design independent studies.
- Marketing mix modelling. A statistical model that estimates the contribution of each channel, alongside price, seasonality and distribution, using several years of weekly data. It earns its cost when spend is large and spread across many channels, including offline. For smaller budgets we will tell you it is not yet worth it.
What attribution cannot tell you
Attribution models assign credit for a sale to the touchpoints a tracking system happened to record. That is useful for tuning campaigns. It is not proof of cause. Last-click models favour whoever was present at the end, usually branded search and retargeting, which often reach people who were going to buy anyway. Privacy changes, consent banners and cross-device behaviour mean a growing share of paths to purchase is never seen at all.
Attribution also cannot see the conversation in a buying committee, the trade show stand someone walked past, or the article an engineer forwarded to procurement. So we treat attributed numbers as directional, confirm the big decisions with incrementality tests and say plainly in every read-out which figures are measured and which are modelled.
The monthly read-out
One document, written for someone who has ten minutes before a board meeting. It opens with the business goal and where you stand against it. Then the metric tree, layer by layer, with movements against baseline and last month. Then what we believe caused the changes, how confident we are, and what we will do differently next month, including where we recommend spending less. Your live dashboard sits behind it for anyone who wants to go deeper.
The same read-out covers paid media from our media planning and buying team, organic and AI search from our SEO work, and the brand health questions that come out of brand strategy. One view, one set of definitions.
Why our fee stays flat whatever we measure
Measurement is only honest when the people measuring have nothing riding on the answer. Our fee is a flat project or monthly fee agreed up front, never a percentage of your media spend and never tied to a metric we also report on. If the numbers say a channel should be cut, our income does not change when we tell you. The detail is on How we charge, and what to expect week to week is on Working with us.
If you want an outside read on your current numbers, send us a brief with whatever data you can share. We will tell you what we see, including what we cannot yet see.