I was recently asked by a small retail chain if a proposed rewards exchange with a local partner — a café offering a free pastry for every £10 spent in store — would actually increase basket size by 10%. The ask felt familiar: founders want a simple, defensible projection before they commit to the operational work and margin trade-offs. The good news is you can prove (or disprove) this with tills and three tightly scoped experiments. Below I walk through a pragmatic, data-first approach I use with SMEs: what to measure, how to run the tests, and how to interpret results so you can make a confident decision.
Why till data is the right source
Till (POS) data is often under-used. It records real transactions with timestamps, SKUs, prices, discounts and — crucially — basket-level detail. That makes it ideal for measuring change in average basket value (ABV) and product mix after introducing a rewards exchange. Compared to surveys or loyalty-app click data, tills tell you what customers actually bought and paid for.
Before any experiment, I always check three things in the till feed:
Define the metric and the causal window
Be precise: the primary metric is average basket value (ABV) per transaction. Secondary metrics include items per basket, incidence of higher-margin SKUs and redemption rate of the partner reward. You want to detect a 10% lift in ABV. That means if baseline ABV is £30, target ABV becomes £33.
Decide on the causal window — the period after a customer becomes eligible for the partner reward where you expect behaviour to change. For a local partner pastry reward this is often short: same-day or next-visit behaviour, and up to 30 days. Keep the test windows consistent across experiments.
Experiment 1 — A/B test at till: control vs. partner offer messaging
Goal: measure the immediate impact of presenting a partner reward to customers at point-of-sale.
Design:
Why this works: it isolates the effect of the communication and urgency (today’s purchase) without yet delivering the redemption mechanics.
What to measure:
Sample size note: use power calculations or a simple rule of thumb — you’ll typically need several hundred transactions per group to detect a 10% change depending on baseline variance. If sample is small, extend test duration or expand stores.
Experiment 2 — Redemption-enabled pilot: coupons issued at till
Goal: measure behavioural lift when customers receive a redeemable coupon for the partner (actual incentive unlocked).
Design:
What to measure:
Interpretation: If ABV in treatment increases by 10% immediately, but redemption is zero, you’ve induced higher spend through messaging. If ABV doesn’t rise but redemption occurs heavily, the program may drive partner footfall without moving your basket size.
Experiment 3 — Partner co-promotion cross-check (reciprocal tracking)
Goal: capture the two-way effect: does the partner reward drive higher basket value at your store when redeemed, and what is the net lift across both businesses?
Design:
What to measure:
This experiment tests the full exchange lifecycle: issuance → redemption → return behaviour. It’s the most realistic test of whether the partner exchange creates sustained additional spend, not just a one-off pastry pickup.
Practical analysis steps using your till data
Once experiments are complete, follow this analysis flow:
Fast ROI modelling
Two numbers matter: gross incremental revenue and cost of the partnership (discount cost + administration).
| Metric | How to calculate |
|---|---|
| Incremental revenue per transaction | (ABV_treatment − ABV_control) |
| Incremental transactions | Change in number of qualifying baskets or repeat visits tied to coupon redemptions |
| Redemption cost | Average cost to you per redemption (if you're subsidising partner item) × redemption rate |
| Net margin impact | Incremental revenue × gross margin − redemption cost − admin |
Run scenarios: conservative (low redemption, small ABV lift), expected, and optimistic. I often model payback period in weeks — how long until the incremental margin covers set-up and promo costs.
Pitfalls and how to avoid them
Common mistakes I see:
Example outcome I’ve seen
In one pilot with a regional grocer and a bakery partner, the A/B messaging test produced a 4% ABV lift. Issued coupons led to a 22% redemption rate and redeemers returned within 10 days, delivering an additional 12% ABV on their return visit. Net of the bakery subsidy (a £0.80 pastry cost) and admin, the program delivered a 6% uplift in incremental margin initially — short of the 10% target. But after adjusting the threshold and promoting higher-margin add-ons at till, the retailer hit the 10% ABV lift in a follow-up test.
If you want, I can draft a tailored experiment plan for your till system, including sample size estimates and a reporting dashboard template you can drop into Excel or Google Sheets. That’s the fastest way to move from “it feels like it should work” to “here’s the evidence and the ROI”.
