Audiences AIThe Borrowed Quarter
The happiest quarter-close call I sat in this spring ran forty minutes, and thirty-eight of them were pleasant. Revenue was up, and full-price sell-through was up with it. The CFO allowed himself a joke, and someone had pasted the topline into the board deck before the call ended. The two minutes that mattered came near the end, when a planner asked, almost apologetically, where the money had come from. Nobody picked the question up. The call closed on schedule. (The scene is a composite of several calls from client work this spring, details altered.)
I kept returning to that question all summer, because the answer surfaced in public reporting within weeks, and it changes what the quarter meant. The money had come from somewhere specific. It had come once.
Where the money came from
Start in the US. CNBC reported that many retailers' strong fiscal first quarters owed part of their strength to higher-than-usual tax refunds. Macy's CEO Tony Spring said the refunds "definitely" helped, and Macy's investor materials paired that admission with cautious guidance for the following quarter, on the expectation of less stimulus ahead. Kohl's, in the same earnings cycle, described its core lower- and middle-income shopper as under acute pressure from gas prices and inflation.
Put the three statements side by side and the print reads differently. The wallet that receives and spends a tax refund belongs to the same shopper Kohl's called stressed. That wallet has no slack. A refund buys a jacket in March, and April inherits the hole where the jacket purchase would have sat. The demand did not grow. It moved.
India was already on the far side
Whatever doubt I had about that reading, India removed it, because India was living the back half of the same cycle at the same moment. Retailers there pulled mid-season sales forward into late May and June, discounting up to 40% as discretionary spending weakened, and executives told the press that sales had fallen as much as 30% over the prior two months. The Retailers Association of India put overall retail growth at 7% for April, with apparel decelerating from the 13% it ran in March.
Strip away the geography and the trigger, and one shopper stands in both stories: a squeezed wallet that spends when support arrives and goes quiet the moment the support fades. Seeing both halves of the cycle in the same news week is what convinced me the spring strength was pull-forward, not new demand. The cohort that lifted the quarter is exactly the cohort that falls silent right after, because the purchase was funded by cash that arrived once.
What the retention system does next
Finance already handles this correctly inside its own frame. FP&A marks the quarter as one-time-helped and braces for tougher comparisons. The damage happens one system over, in the CRM, which has no concept of a macro tailwind. What it sees is a customer who bought in March and has been quiet since. Sixty to ninety days into that quiet, at most retailers I have seen, she gets rescored from active to at-risk, and the win-back machinery wakes up.
The mechanism deserves stating plainly. A recency-based lapse score is a proxy for one question: how likely is it that this relationship has ended. A liquidity trough breaks the proxy. On every field the model can read, last order date, order gap, engagement decay, the refund-window buyer is indistinguishable from a genuine defector. So the system fires a voucher at someone who was always coming back once her cash cycle turned, and the business pays twice. It pays in margin, because the discount lands on a purchase that would have happened at full price. It pays in information, because the reactivation credits a campaign for demand it never created, which flatters the win-back program's measured return and earns it a bigger budget next year.
The math is worth a sketch. Suppose the voucher is 15% and seven in ten of the depleted cohort would have returned unaided; both figures are illustrative. Then most of the campaign's spend is a transfer to customers the business already had, and most of the reported lift is misattribution. The demand curve gets misdated in the same stroke. The model learns that this cohort responds to discounts in month three, when what actually happened is that their money came back in month three.
The operator response is unglamorous. Tag cohorts at the point of sale with the conditions of the sale: refund window, deep-discount window, any liquidity event visible on the calendar. When those cohorts go quiet, hold the reflexive win-back and watch a holdout. If the untouched group returns on its own as liquidity recovers, you have found a depletion trough and kept the margin. If it stays away, the lapse was real and you have lost a few weeks. Either way you now know something the recency score could never tell you: whether the silence was about you at all.
Three questions to put to your own numbers before the next close call ends on schedule:
- Which cohorts inside your best recent quarter bought during an identifiable liquidity event, and how much of the print do they carry?
- When those cohorts go quiet, does any field your retention system reads separate a customer who is out of cash this month from a customer who is gone?
- How much did your last win-back campaign pay to customers who would have returned anyway, and is there a holdout that could tell you?
Separating depletion from churn in a live cohort file is the kind of question we take into Interpret.