Audiences AIThe Half-Second Your CRM Never Sees
The most consequential decision in an FMCG brand's quarter has changed address. It used to be made in front of a shelf, where the brand had at least fought for placement and facings. It is now made inside a quick-commerce ranking algorithm, in the half second after a shopper opens Blinkit or Zepto and finds her usual product out of stock in her pincode. She accepts the suggested substitute or she scrolls to a rival. Sometimes she just closes the app. Brand preference is honored or broken right there, inside software the brand neither wrote nor can read.
Across the FMCG and beauty categories we track, quick commerce carried roughly 60 to 75 percent of online volume by FY26. That band comes from our own category scans, so read it as a cohort finding with its provenance attached; it moves by category and by city. At the top of the range, the out-of-stock moment is the ordinary case. A brand can win every negotiation in modern trade and still watch preference leak away one pincode at a time.
What actually changed
The old shelf was legible. When a store ran dry, a distributor report said so, a merchandiser saw it, and the fix moved at the speed of a delivery truck. The shopper standing in the aisle also gave the brand a fighting chance: she could see the whole shelf at once, and the habits formed there were slow to break.
Dark stores dissolved that arrangement. Assortment per pincode is now set by the platform's demand model. The shelf is a ranked list assembled fresh for every query, and when your SKU is missing, the substitute on offer is chosen by the app's own economics: fill rate, margin, delivery density, conversion. There is nothing hostile in any of this. It is ordinary optimization, pointed at the platform's P&L. Your P&L appears nowhere in the objective function. Beauty shows how far the shift has run. In our scans, beauty and personal care now sits among the largest segments on the biggest quick-commerce apps, which means high-ritual, high-loyalty products are living and dying by the same ranked list as bottled water.
Who holds the decision now
Walk through the candidates. Your CRM has no idea the shelf moment happened. Your loyalty program cannot fire inside another company's checkout. Your email arrives hours later, to a customer who has already bought something. The only parties present at the substitution are the shopper and the algorithm, and the algorithm built the room: the order of the list, and which product gets the similar-item slot.
The subtler cost lands in your data. A lost substitution produces no event in any system you own. It surfaces in the CRM weeks later as a longer purchase interval, indistinguishable from a customer who simply needed less shampoo. The event that cost you the customer is invisible when it happens and mislabeled by the time it appears. Every number your team debates in the Monday meeting sits downstream of an event nobody logged.
Now look where retention money goes in this cohort. The one-rupee trial deal. The hundred-rupee voucher that fires after two quiet days. Those mechanics, visible in any of these apps, rent a habit; they say nothing at the only moment the habit is contested. That is a discount schedule, not a loyalty system. And the more volume shifts to quick commerce, the larger the share of the retention budget spent on moments that no longer decide anything.
The operator's playbook
Start by treating availability as media. In a high-density pincode, an hour out of stock is a paid sampling program for your nearest rival, funded by you. Someone should own out-of-stock exposure the way someone owns share of voice, with the same seniority and the same weekly cadence.
Then choose which customers you will defend at the shelf moment. Fill rate is a finite budget, and spread evenly it protects no one. Concentrating it on the pincodes and cohorts where frequency and margin justify a fight is a real decision, which means it needs an owner, a review cadence, and permission to be wrong.
Finally, negotiate for visibility the way brands once negotiated for shelf audits in modern trade. The platforms sell ad slots readily; availability, ranking, and substitution reporting take more pushing. Brands rarely push, because nobody has priced what the blindness costs. The asymmetry is negotiable, but only for operators who arrive with a number.
The starting point is instrumentation, and none of it needs the platform's permission to begin. Five signals worth standing up before next quarter:
- Out-of-stock hours on your top SKUs, by pincode, weighted by order density, sampled from the apps themselves.
- Daily search rank on your category's head terms in the cities that carry your volume.
- Substitution capture in both directions: how often you inherit a rival's stockout, and who inherits yours.
- Purchase-interval stretch in your CRM laid over out-of-stock windows in the apps; where the two line up, you are looking at substitution booked as reduced demand.
- The share of retention spend that lands before the shelf moment versus after it.
All five can be watched from outside the walls, imperfectly at first, starting this month.
Interpret exists to stand these signals up and to keep reading them long after the first week's novelty wears off.