**To:** Aisha Rahman | **From:** Ingrid Solberg | **Re:** Loyalty & referral programme — Friday evidence review

**Bottom line:** The programme is running a real, growing cost with no clear revenue upside, and where we initially saw a red flag (refund rate), most of it is a mix effect, not a loyalty effect. Referral is too small to matter yet.

**1. Cost is real and growing (R: "cost by month"; "unit economics").** Outstanding points liability has grown almost every month since 2023, to **€491k by Dec-2025** (redemption runs only ~20–30% of issuance monthly). Looking at realized cost: loyalty members have **redeemed €177.8k in points value against €915k of their own revenue — 19.4% of member revenue**, which is high for a rewards programme (healthy programmes run low single digits). A further €669k of earned-but-unredeemed points sits as a contingent liability. Referral payouts are immaterial (~€650 total over 3 years, 67 redemptions). *Confident.*

**2. Members generate less revenue per head, even accounting for vertical and tenure ("tenure cohort"; "unit economics").** Members average €688 lifetime revenue vs €1,512 for non-members, and order less (3.72 vs 5.55). This persists across every tenure band, so it isn't a "newer accounts" artifact. However, within-vertical comparisons (see below) shrink this gap meaningfully — it's partly a mix issue, not purely behavioral. *Confident the gap is real; less confident about its full size.*

**3. The headline refund-rate gap is mostly a vertical mix artifact — this is the most important correction to our earlier read ("refund rate by vertical").** Globally, members showed 32.4% refund rate vs 15.5% for non-members — a huge apparent red flag. But broken out by vertical, members have an **equal or lower** refund rate than non-members in 4 of 5 verticals tested (Beauty 5.1% vs 6.5%; Home & Living 13.0% vs 15.0%; Sports & Outdoor 22.7% vs 24.1%; Apparel 38.4% vs 38.8%, essentially tied). The reason: 72% of all loyalty members sit in Apparel & Footwear, which runs a naturally high ~38% refund rate for everyone — members and non-members alike. Non-members are more spread into lower-return verticals, dragging their blended rate down. **I do not trust the original global refund comparison as a standalone finding; the vertical-level cut is the one to use.** *Confident in this correction.*

**4. Participation is small, flat, and absent from B2B ("reach by vertical"; "participation over time").** Loyalty penetration has plateaued at ~15–16% of customers for two years, concentrated in Apparel (72% of members) and Sports & Outdoor (highest enrollment rate, 27.6%). Neither B2B wholesale vertical (Industrial Supplies, Office Supplies) has a loyalty programme at all — 0 of 9 tenants — so Tobias's question on whether enrolled B2B accounts pay better **cannot be answered; there's no B2B enrollment to compare.** Referral sits under 1% of customers everywhere. *Confident.*

**5. No gaming signal found, but one test was inconclusive.** Referral redemptions are spread across dozens of referrers (almost all 1 redemption each), and top referrers look like genuine, active customers — no concentration red flag. I tried to test whether refunds cluster on points-redeeming orders specifically, but **zero member orders could be linked to a points redemption via order_id** — likely a data-linkage gap, not evidence of absence. I'd treat the "rewards redeemed without real purchase" hypothesis as untested, not cleared.

**Caveats:** The early-referral-behavior numbers (60 referred customers) are too thin to generalize. B2B-industry matching for one earlier report used fuzzy text matching on tenant names — the reach-by-vertical figures above are the clean version and supersede it.

**Next steps:** (1) Re-run the revenue/order comparison controlling for vertical *and* tenure together, to isolate the true behavioral gap. (2) Get a better redemption-to-order linkage to properly test the gaming hypothesis. (3) Decide whether to extend loyalty to B2B before or after fixing member economics.