Should the Flash Sale Be Smaller?
Ops, economics, and customer experience collide on the year's biggest day.
The Prompt
Your client, a top-3 Indian e-commerce marketplace, runs an annual festival flash-sale week that generates 22% of annual GMV. Last year the week also produced: 9% order cancellations (vs 3% normal), delivery times of 8–11 days (vs 2–3), a courier-network meltdown, and a measurable dip in repeat-purchase rates among first-time festival customers. The COO asks a heretical question: should we deliberately cap the sale?
Opening exchange
The question behind the question: does the marginal festival order create or destroy value once fulfilment degrades? I need three things: the contribution of a festival order versus normal, the cost of the degradation (cancellations, support, re-delivery), and the LTV effect on customers who had a bad first experience.
Reframes "cap or not" into "find where the marginal order turns value-negative" — that's the analytical move the case wants.
Festival orders average ₹1,450 AOV at 6% contribution after discounts and logistics — versus 11% normally. A cancelled order costs ~₹70 in logistics and support. First-time festival customers with delayed orders repeat at 18% over six months versus 34% for on-time.
So festival orders are already half-margin, and the tail of them — the ones that overflow capacity — carry cancellation costs and an LTV penalty. The structure: find the capacity threshold where degradation begins, price the degradation per order beyond it, and then design the better answer — because the real options aren't binary cap/no-cap; they're shaping demand and flexing capacity.
Structure & Hypothesis
Analysis & Data
Price the degradation. Last year: 31 lakh orders/day average across the week against your 24 lakh threshold; first-time customers were 35% of festival orders.
Seven lakh orders a day ran "hot." Direct costs: cancellation delta 6% × ₹70 ≈ ₹4 per hot order — small. The LTV term dominates: 35% first-timers × (34% − 18%) repeat-rate drop × say ₹260 contribution per future repeat ≈ ₹15 per hot order — and that's one repeat cycle only; compounding cohorts make it ₹25–40. So each over-threshold order earns ₹87 of contribution but burns ₹20–45 of future value, plus brand drag we haven't priced. The marginal order is still weakly positive on paper — but barely, and the average customer experience degrades for everyone, including the 24 lakh below the line. That externality is the real argument.
The subtle point: over-capacity orders damage other orders' experience too — a congestion externality, like the airline case's network effects inverted.
So — cap or not? The CEO will want one sentence.
"Don't cap revenue; cap velocity — spread the same GMV across more days and pre-sold slots, and the problem dissolves." Concretely: extend the event from 6 to 10 days with category-staggered deal drops; sell delivery-date slots at checkout (with a small reward for choosing the slow lane); pre-position the top-500 SKUs in regional warehouses based on wishlist data; and contract gig-courier surge capacity at 130% of last year. GMV holds or grows, the peak flattens under the threshold.
Recommendation
Recommend to the COO
- Reject the binary: cap velocity, not revenue — stretch to 10 days with category-staggered drops and pre-booked delivery slots.
- Pre-position top SKUs regionally using wishlist signals; contract courier surge to 130% of last year's peak.
- Offer an incentivized slow lane ("get it in 6 days, earn ₹50 credit") — customers self-sort, and the credit is cheaper than a cancellation.
- Track the festival on contribution including LTV deltas per cohort, not GMV — make the externality visible in the metric the org optimizes.
Key Takeaway
What this case teaches
Past a capacity threshold, marginal volume taxes every other order — a congestion externality most GMV-obsessed metrics never see. The elegant answer to "should we do less?" is usually "do the same, differently distributed" — flatten the peak, don't shrink the area under the curve.