Attribution in restaurant marketing is harder than in e-commerce and easier to get wrong. The purchase is fast, often offline, frequently on a different device, and driven by channels that produce no click at all. Last-click attribution, the default nearly everywhere, will confidently tell you that paid search generates everything and direct mail generates nothing.
Last-click gives all credit to the final touch before the order. In a category where the final touch is almost always a branded search or a direct visit, that means brand-building and demand-generating channels are structurally invisible.
The predictable outcome: an operator cuts the mail programme because it shows no attributed revenue, and watches paid search volume fall three months later without connecting the two.
Full multi-touch attribution is overkill for a pizza business. This layered approach is defensible and possible to actually run.
A meaningful share of pizza orders are placed by phone or at the counter and never touch a tracked link. Loyalty enrolment and phone-number matching close much of that gap, which is one more return on identification rate.
Where a customer is identified, the channel that reached them can be linked to the order regardless of how the order arrived. Where they are not, the order joins the unattributed pool, and if that pool is most of your revenue, attribution work is premature and identity capture is the real project.
You cannot attribute what you cannot identify. Attribution sophistication beyond your identification rate is decoration.
Attributed revenue is what platforms claim. Incremental revenue is what would not have happened otherwise. Build the board report on the second, and use attributed figures only for optimising within a channel where the same bias applies to every option.
Practically: keep two columns. Platform-attributed for operational tuning, holdout-measured incremental for budget decisions. Never let the first column be the one that sets the budget.
For most restaurant operators, no. They require clean cross-device identity that a pizza shop rarely has, and they produce precise-looking numbers built on assumptions nobody checks. Holdout testing gives a less granular but far more trustworthy answer at a fraction of the cost.
Match the phone number to a customer record and read the campaigns that customer received recently. Where the caller is unknown, the order joins the unattributed pool, which is another argument for capturing identity at the till.
You get almost no attribution data, and marketplace-driven demand is partly your marketing and partly theirs. Track marketplace volume as its own line and focus your attribution effort on the direct channel, where you can actually see what happened.
Keep reading
Building holdout groups into every campaign, reading the difference honestly, and accepting that most reported marketing revenue was going to happen anyway.
Audience & dataPractical ways to turn anonymous tickets into identified customers (online ordering, loyalty, QR on the box, the till, and wifi) without annoying anyone.
MeasurementSetting a marketing budget as a share of sales, splitting it between acquisition and retention, and running a portion as deliberate experiment.
MeasurementReading retention curves by acquisition month, spotting the second-order cliff, and why total revenue hides everything that matters.
marketing.pizza runs all of it, every night, across every store you have.