How to measure a promotion like an analyst, not a marketer

7 July 2026 · 5 min read

Run a happy hour for three Thursdays and someone will ask the only question that matters: did it work? The usual answers — “felt busier”, “the post did well”, “we sold loads of the special” — are all ways of not knowing. The honest answer needs two numbers, and most restaurants already have both.

Lift is a comparison, not a count

Covers during the promotion is not the result; it's half of it. The other half is what those same hours would have done without the promotion — your baseline. The most defensible baseline for a restaurant is the same weekday and window, averaged over the weeks before launch. Thursdays 4–6pm before versus Thursdays 4–6pm after. Lift is the difference. Everything else is decoration.

The traps

Three things quietly corrupt this comparison. A thin baseline — one prior Thursday isn't an average, it's an anecdote; use several. A moving backdrop — if the promo weeks coincided with a heatwave or a festival, some of the “lift” belongs to the calendar, not the offer. And revenue amnesia — a discount that fills seats can still shrink takings, so measure covers and money, not covers alone.

Kill the vanity metrics

Reach, likes, and impressions measure the advert, not the till. They're not useless — they tell you whether anyone saw the thing — but a promotion's verdict is written in covers against baseline and pounds against baseline. If a campaign can't survive that comparison, it wasn't working, however it performed on Instagram.

Decide the verdict before you launch

The discipline that separates analysts from optimists: write the target down first. “This exists to recover about twenty covers on quiet Thursdays” is a claim you can check. Set the target, run it for enough weeks to mean something, read the number, and then do the rarest thing in marketing — stop the ones that didn't work and reinvest in the ones that did.