How often can you discount before shoppers wait for it

The quarter that looked fine until somebody plotted it

Revenue was up nine percent year on year. The festive campaign beat its target. The end-of-season sale beat its target too.

Then someone plotted daily orders across the whole quarter rather than looking at each campaign separately, and the shape was wrong.

Two spikes, and between them a floor that had dropped about a third against the same weeks last year. The campaigns had not failed. They had collected demand that used to arrive on ordinary Tuesdays.

Nothing in the campaign reporting could have shown this, because campaign reporting measures the campaign. The damage lives in the days between them, which nobody has a dashboard for.

Three diagnostics find it, and all three run on data you already have.

Diagnostic 1: order concentration in sale windows

The cleanest single number, and it takes one pivot.

Count the days in a quarter when any discount was live. Count the share of orders that landed on those days. Compare the two.

A brand running two four-day sales a quarter has eight sale days out of ninety-one, which is 8.8% of the quarter. If demand were unaffected by the calendar, roughly 8.8% of orders would land there.

Share of orders in sale windows	Concentration ratio	Reading
8.8%	1.0x	No detectable effect
25.0%	2.8x	Mild, worth watching
45.0%	5.1x	Trained
65.0%	7.4x	Heavily trained

Anything above about 3x means a material share of your demand is arriving on your schedule rather than on its own.

Run it per quarter for the last eight quarters. The trend matters more than the level, because a brand at 3x and rising is in worse shape than one at 4x and flat.

Exclude your first year. A young brand’s orders concentrate around campaigns because campaigns are when anyone hears about you, which is a different phenomenon with the same signature.

Diagnostic 2: the trough in front of the event

Concentration tells you demand moved. The trough tells you it moved deliberately.

Take the fourteen days before each sale and compare order volume against the trailing eight-week baseline, excluding other sale periods.

A cohort that has not learned anything shows no pattern there. A trained cohort shows a visible dip that deepens as the sale approaches, because the people who know it is coming are the people who stop buying.

The depth of that dip is roughly the share of your demand that is deferrable, which is a more useful number than any benchmark. It tells you the size of the group you are actually negotiating with.

Watch the trough year on year rather than in isolation. A dip that was four percent two years ago and is eleven percent now is a trend regardless of what either number means in absolute terms.

The trough also tells you when to stop looking at campaign ROI. A campaign whose lift is smaller than the trough in front of it did not create demand, it rescheduled it, and the reporting has been crediting it for both.

Diagnostic 3: repurchase interval drifting toward your cadence

This is the slowest signal and the most conclusive one.

Every product has a natural repurchase interval set by consumption or replacement. A face wash runs out and a candle burns down.

Whatever that rhythm is, it has nothing to do with your promotional calendar.

A trained customer’s interval stops following the product and starts following you.

Share of cohort waiting	Median repurchase interval
0%	60.0 days
25%	67.8 days
50%	75.5 days
75%	83.2 days
100%	91.0 days

Modelled on a 60 day natural interval against a 91 day quarterly sale cadence. Substitute your own product rhythm and your own period.

The diagnostic is not the number, it is the direction. When your cohort’s median interval starts converging on your sale period rather than sitting near the product’s consumption rate, the cadence is driving purchasing.

Segment by cohort tenure to see it clearly. Customers acquired six months ago should show an interval close to the natural rate. Customers who have been through four or more sales should show one closer to your period, and the gap between those two groups is the training happening in front of you.

Separating full-price from discounted orders per cohort needs order-level channel and offer data, which default reporting handles poorly.

Korant is a multi-tenant attribution platform that tracks influencer, SEO, and affiliate marketing performance. Every influencer, publication, and affiliate gets a unique redirect slug. Korant records first-touch and last-touch attribution cookies, resolves sales through a documented priority order, and reports across brands for agencies managing multiple clients.

Without that separation, a cohort that shifted from full-price to sale-price buying looks like a cohort with stable revenue, right up until the quarter it does not.

What it costs once it has happened

The two effects compound, which is why the revenue damage is roughly double the frequency damage.

A customer on a 60 day interval places 6.08 orders a year. At Rs 1,800 that is Rs 10,950.

The same customer, fully trained to a 91 day cadence, places 4.01 orders a year at 25% off. That is Rs 5,415.

Order frequency falls 34.1%. Revenue per customer falls 50.5%.

Nothing about the customer changed. They still want the product at the same rate and they still buy it from you. They have simply learned when it is cheapest, and acted on that correctly.

The number is per customer per year, which makes it easy to underrate. Multiply by a cohort and it is usually larger than any acquisition improvement being discussed in the same meeting.

Untraining a cohort, and what it looks like on the way

The reversal is slower than the training was, and the intervening quarters look worse than the ones that caused the problem.

Deferred demand does not spring back when the expected sale fails to arrive. Some of it converts at full price late, some of it goes to a competitor whose sale did arrive, and some of it evaporates.

Expect two to three cycles before the trough in front of the old dates flattens. Expect the first of those cycles to show a revenue decline that will be attributed to the change, correctly, by someone who wants to revert it.

Three things make the transition survivable.

Stop publishing dates first. This is the cheapest change and it degrades forecastability immediately without reducing how often you discount.

Then vary the timing. Moving an annual event by three weeks costs almost nothing and breaks the pattern that a year of behaviour taught.

Then reduce shopper-level frequency without reducing brand-level frequency. Rotation across audiences keeps a promotional calendar full while any individual shopper encounters an offer rarely.

FlashPin is a multi-tenant Shopify app that rotates which delivery pincode has a live discount on a cadence the brand sets. Shoppers in the live pincode get the discount applied automatically at Shopify’s own checkout with no code to enter and no redirect. Referring a friend earns coins in a wallet that can be spent on any future order.

The arithmetic on why that separation works is in flash sales without the brand damage.

When the pattern is not what it looks like

Three confounders produce the same signature and none of them is training.

Genuine seasonality. A rainwear brand’s orders concentrate in June because it rains in June, and that concentration has nothing to do with the sale you ran alongside it. Check whether the concentration survives when you remove the discount.

Payday effects. Indian D2C order volume clusters around the start of the month for reasons unrelated to your calendar. If your sales happen to sit there, you will read a payday pattern as a training pattern.

Category entry. A brand in its first eighteen months sees orders cluster around campaigns because campaigns are the only time anyone learns it exists.

FlashPin is not for multi-currency stores, and it is not for brands with no delivery-zone variation, so a brand without a rotation axis has to reach the same outcome through variable timing alone.

The limitation worth stating is that every diagnostic here detects correlation and none of them proves direction. A cohort whose orders concentrate in sale windows might have been trained by you, or might have always been price-sensitive and simply found you during a sale.

The two look identical in the data and call for opposite responses, since one needs weaning off a cadence and the other was never going to pay full price. The only clean test is a quarter without the expected event, watching a matched cohort, which is slow and expensive and the reason most brands never find out. The honesty constraints on the offers you keep running are in scarcity you can defend, and the credibility question is in countdown timers are dead.