Open Meta Ads Manager, then Google Ads, then GA4, then Shopify. Add up the revenue each one says it drove. If the total is bigger than the money that actually landed in your bank account, you've met the core problem with modern marketing measurement. Every tool is telling you a slightly different story, and most of them are flattering themselves.
The good news is that you don't need perfect attribution to make good decisions. You need a handful of numbers you trust, a few cheap tests, and the discipline to look at them every week. This is how I'd set that up for a small or mid-size brand.
Attribution was always a bit of a guess
It's tempting to blame the last few years of privacy changes for breaking attribution. Tracking opt-outs on phones, cookie consent banners, shorter lookback windows and people switching between devices have all made it worse. But it was never as clean as the dashboards suggested.
Last-click models give all the credit to whatever happened right before the sale, which usually means branded search or email. Platform reporting does the opposite. Meta and Google both count conversions they had any plausible hand in, including people who scrolled past an ad and bought anyway. Each platform is marking its own homework.
For a big brand, the answer is often a full marketing mix model and a data team to run it. For a brand spending a few thousand to a few tens of thousands a month, that's overkill. What you want is to be roughly right, quickly, rather than precisely wrong after three months of modelling.
Anchor on the numbers that can't lie
Start with figures that come from your own books, not from an ad platform. These are hard to fudge because they don't depend on tracking at all.
The first is MER, or marketing efficiency ratio. Total revenue divided by total marketing spend, across every channel, for the same period. It doesn't tell you which channel worked. It tells you whether marketing as a whole is paying its way.
The second is new customer acquisition cost. Total spend divided by the number of first-time customers. Returning customers would have come back through email or a bookmark anyway, so this is the cleaner read on whether you're actually growing.
The third is contribution margin after marketing. Revenue, minus cost of goods, minus fulfilment and returns, minus ad spend. This is the one that tells you if you're building a business or just buying revenue.
Work out your breakeven point before you look at anything else. As a rule of thumb, if you keep 50p of every pound after product and fulfilment costs, you need a blended return of at least 2x just to stand still on first orders. Anything above that is margin you can reinvest or keep.
When I owned the full P&L for the Swan & Edgar and Anthony James watch brands at Fields Luxury, this was the lens that mattered. Turning loss-making brands profitable meant looking at shipping, packaging, returns and average order value alongside the ad accounts. No attribution report was going to show that the fix for a channel's numbers sat partly in the returns rate.
Use platform data for direction, not for truth
This doesn't mean ignoring what Meta or Google tell you. It means using their numbers for the right job.
Inside a single platform, comparisons are fairly reliable because the same tracking flaws apply to everything. If one ad set reports a 4x return and another reports 1.5x, the first is very likely better, even if neither number is accurate. Use platform data to choose creative, audiences, keywords and bids.
Where it falls apart is comparing across platforms. A 5x on Meta and a 5x on Google are not the same 5x. They count different things over different windows with different assumptions.
At Money Clothing we ran PPC, paid social and affiliate side by side. That's exactly the setup where every channel wants credit for the same order. Platform ROAS was useful for managing each account, but the decisions about where the next pound of budget went needed the blended view and some testing.
Run cheap incrementality tests
The question attribution is really trying to answer is simple. What would have happened if I hadn't spent this money? You can answer that directly, and you don't need a data scientist.
A few tests that work for smaller brands:
- Pause test. Switch off a channel or campaign for two to three weeks and watch total revenue and new customers. If nothing moves, it wasn't doing much. Branded search is the classic one to try, because people searching your name were often going to find you anyway.
- Geo holdout. Run a campaign in some UK regions and not others, then compare sales by region against the previous period. Pick regions that normally behave similarly.
- Budget step test. Raise spend on one channel by a clear amount, say 30-50%, hold everything else steady, and see whether blended revenue follows.
A few rules make these worth doing. Change one thing at a time. Write down what you expect to happen before you start, so you can't rationalise the result afterwards. Avoid running tests across a big sale or a stock problem. And accept that small brands will get noisy results, so look for big, obvious effects rather than splitting hairs over a few percent.
Ask your customers
The most underrated measurement tool is a single question after checkout. How did you hear about us?
It's cheap, it takes an afternoon to set up, and it picks up things that tracking will never see. Word of mouth, a podcast mention, a TikTok someone watched on their partner's phone, an influencer post they saw a week before buying.
People don't remember perfectly, and they'll often say Google when they mean they searched your name after seeing an ad. That's fine. You're looking for trends over time, not a court-ready record of each customer. If you start spending on a new channel and it never appears in the survey, that's worth knowing.
Keep the options short and include a free-text box. Read the free text every month. It's often more useful than the dashboards.
Watch the signals around the edges
Good upper-funnel marketing tends to leave footprints in places it doesn't get credit for. Branded search volume rises. Direct traffic creeps up. Email sign-ups grow. The share of new customers in your sales mix changes.
If you increase spend on paid social and your branded search impressions go up a couple of weeks later, that's a decent sign the social spend is creating demand, even if Meta's own reporting looks thin. If you cut a channel and branded search drops, you've learned something too.
None of these signals is proof on its own. Put two or three together and you'll usually have a clearer picture than any single attribution model gives you.
Turn it into a weekly habit
Measurement without attribution only works if it's routine. One sheet, updated every week, with the same columns every time:
- spend by channel
- total revenue and MER
- new customers and new customer acquisition cost
- contribution margin after marketing
- top answers from the post-purchase survey
- branded search and direct traffic trends
- a notes column for anything that changed, like a price rise, a new product, or a test starting
That notes column matters more than it looks. Half of the confusing weeks in any business are explained by something nobody wrote down.
Then agree some simple decision rules. If MER is above target for three weeks, increase spend in measured steps. If new customer cost is rising while MER holds, you're probably leaning on returning customers and need to look at acquisition. If a channel survives a pause test, keep it. If it doesn't, cut it and reinvest.
It's not glamorous. It does mean you'll make better calls than a brand staring at four dashboards that disagree.
FAQ
What is a good MER for a small brand?
It depends entirely on your margins. Work out your breakeven point from your gross margin after fulfilment and returns, then set a target comfortably above it. A brand with thin margins needs a much higher MER than one selling high-margin products with strong repeat purchase.
Do small brands need marketing mix modelling?
Usually not yet. MMM needs plenty of historical data, steady spend across channels and someone who knows how to run it. Blended metrics, simple holdout tests and a post-purchase survey will get most small and mid-size brands far enough to make sound decisions.
Should I stop looking at GA4 and platform reporting?
No. They're still useful for optimising inside each channel, spotting site issues and comparing campaigns against each other. Just don't treat their revenue figures as the truth, and don't use them to compare one platform against another.
If you're trying to work out what your marketing is really doing and the numbers don't add up, I'm happy to talk it through. I'm open to Head of Marketing and Head of Growth roles, particularly with DTC and consumer brands, and I take on consulting work too. You can get in touch via the contact link on tomosborne.me.