A sports memorabilia brand with a loyal customer base was reading its ad performance from six places that never quite agreed. We built one dashboard, tracking that holds up, and a marketing mix model, then pointed the same budget at the customers they did not have yet.
Powers Sports Memorabilia · Autographed memorabilia & signings · United States · powerssportsmemorabilia.com
01 · The problem
Powers Sports Memorabilia has been selling authenticated autographs and running athlete signings for years. That builds something most ecommerce brands don't have: a large base of collectors who come back on their own, and a brand people search for by name. It also creates a specific measurement trap.
What we set out to do. Three things, in order. Build one view of the business that puts Shopify, the ad platforms and analytics on the same page. Make sure the data feeding that view is real. Then use a marketing mix model, not platform attribution, to decide where the money should go. The engagement has since become a monthly retainer, which section 05 describes.
02 · One view of the truth
The dashboard came first, because it changes the conversation. Once yesterday's sales by channel sit next to what each platform spent, new versus returning, the question stops being "what does Meta say?" and becomes "what did we get for it?"
The Looker Studio dashboard, six views
A dashboard is only as good as what feeds it, so we rebuilt the measurement underneath.
03 · What the model said
We built a Bayesian marketing mix model in Google Meridian on roughly two years of weekly data: paid spend by channel, email sends, organic and social traffic, search impressions, and holiday flags. It accounts for carry-over (an ad keeps working after it runs) and diminishing returns (the tenth dollar earns less than the first). Then we ran it twice: once against total sales, once against new-customer sales only.
Where revenue comes from, by lens
That split is the whole story, and it is true of any brand with real repeat custom. Score the ads against total sales and they look like a modest contributor to a business that largely runs itself. Score them against new customers only and they are the acquisition engine. Both readings are correct. The one that should drive the budget is the second, because acquisition is the only thing the ads can actually do.
Return per dollar by channel, with uncertainty
Finding
Performance Max ships without brand exclusions by default, so it will happily serve people who already typed the brand name. Their reported return looks strong and their real acquisition work is hidden underneath it. This is the single most common thing we find in well-known brands' accounts, and it is a settings change rather than a spend decision. It had to come first, because until it is fixed every other number is measured against an inflated one.
Finding
Klaviyo and SMS were the most efficient revenue in the account by a wide margin, at close to zero marginal cost. The list had been under-used. Growing it and sending to it more was the cheapest revenue available.
Finding
On the total-sales lens Meta sat around break-even, which is exactly what you would expect from a prospecting channel measured against a business carried by returning customers. The easy call was to cut it. We did not make that call, because the new-customer lens pointed the other way and there was a third way to check.
The natural experiment: new customers when Meta ads were on vs. off
04 · What changed
No increase in spend was required. The reallocation moved money away from where the platforms were claiming credit and toward where the model showed incremental new-customer revenue, then set up the tests to prove it.
05 · The retainer
The project work above turned into an ongoing engagement. This is what lands each month, and it is what any retainer client of ours gets: the analysis, the decisions and the plumbing, from one person who has seen all of it.
| Deliverable | Cadence | What it answers |
|---|---|---|
| Live performance dashboard | Always on | What happened yesterday, by channel, product and customer type. Shopify, Google, Meta, GA4 and Klaviyo in one place. |
| Performance diagnostic | Monthly | New-customer acquisition, paid-media efficiency and site conversion, year over year, with the five things that matter and what to do about them. |
| Google Ads structure & search-term review | Monthly | Brand vs. non-brand spend and return, terms to scale, terms to cut, campaign structure changes. |
| Meta campaign plan & audience architecture | Per launch | Campaigns, ad sets, audiences, exclusions and creative direction, built around new-customer acquisition and the signings calendar. |
| Customer value & break-even analysis | Quarterly | LTV by cohort, cost to acquire, and the ROAS the account must clear at each spend level after fees. |
| Marketing mix model refresh | Quarterly | Re-run on the latest data to confirm the reallocation worked and set the next one. |
| Experiment design & read-out | As needed | Lead-capture tests, ads on/off analysis, campaign experiments, each with a clear verdict. |
| Tracking maintenance | Ongoing | Server-side container, GA4, pixels and UTMs kept clean. Defects found before they reach a decision. |
Methods, briefly. Looker Studio on Shopify, Google Ads, Meta, GA4 and Klaviyo connectors. Server-side GTM on Stape. Google Meridian Bayesian MMM, weekly grain, two years of data, adstock and Hill saturation, two KPI runs (total sales and new-customer sales), convergence checked before any number was reported. Ads on/off comparison using Welch's t-test across five pause windows. LTV from six annual customer snapshots. All figures in this study are rounded or indexed at the client's request; the client's reports carry the exact numbers.
If your brand has real organic pull, no ad platform can separate the demand it created from the demand it merely met. A marketing mix model is the independent check. It starts with a scoping call to see whether you have the data history to support one.
Book an MMM scoping call