Cohorts: the report that shows whether the business is actually getting better

Total revenue is the least informative number a restaurant produces. It combines new customers, returning customers, price, frequency and seasonality into a single figure that moves for reasons you cannot see. Cohort analysis takes it apart: group customers by the month they first ordered, and follow each group forward.

Building the report

Rows are acquisition months. Columns are months since acquisition. Each cell holds the share of that cohort who ordered in that month, or the revenue per original customer.

Read it two ways. Across a row shows how a single cohort decays over time. Down a column compares cohorts at the same age, which is how you tell whether the customers you acquired in June are better or worse than the ones from March, independent of how long they have had to come back.

The second-order cliff

The first thing every restaurant cohort chart shows is a steep drop between month zero and month one. A large share of first-time customers never return, and this single transition dominates lifetime value more than anything that happens afterwards.

That is why the second-order journey is the highest-return automation available, and why the metric worth watching weekly is the share of first-time customers who place a second order within 30 and 60 days.

What the curve tells you that revenue does not

• A flattening tail: the cohort has found its core of genuine regulars. Where it flattens is your real retained base.
• Cohorts getting worse at the same age: you are acquiring lower-quality customers, usually from a discount-led channel. This is invisible in total revenue while volume is growing.
• Cohorts getting better at the same age: retention work is landing. The most satisfying line on any restaurant chart.
• A cohort that decays then partially recovers: a win-back programme working, and one of the few places you can see it cleanly.
• A cliff at a specific month across all cohorts: something happened to the business that month. Price change, staff change, a competitor, a bad review cycle.

Revenue tells you what happened. Cohorts tell you what is going to happen.

Segment cohorts by acquisition source

The most useful cut is by how the customer was acquired. Customers acquired through a deep discount routinely retain worse than customers acquired through referral, organic search or a fundraiser: often dramatically so.

Once you can see that, cost per acquisition stops being the right way to compare channels. A channel with a higher acquisition cost and a much better retention curve is the better buy, and only cohort analysis makes that visible.

Questions