Where Your D2C Repeat Buyers Go: Quick Commerce Leakage

The cohort chart that looks like a product problem

The sequence is familiar. A D2C brand lists on Blinkit, Zepto and Swiggy Instamart. Total brand revenue rises, so the quarter looks good. Two quarters later the website cohorts start telling a different story. Ninety day repeat rate on new customers, which had been steady around 30 percent, drifts to 24 percent. Email flows underperform. Someone proposes a formulation review and a fresh round of customer interviews.

In most cases nothing is wrong with the product. The customer still buys the brand, just not from the brand. She buys it in ten minutes at 9 pm from an app that already has her address, her payment method and her basket. The D2C cohort is not churning, it is migrating, and the two things need completely different responses.

Why platform reporting hides it

Quick commerce reporting gives a brand sell out by store, by city and by SKU. It does not give the customer identity, so there is no way to see from the platform side that the buyer in Powai who bought your refill pack last Tuesday is the same person who ordered from your website every six weeks for a year. Your own analytics sees only the disappearance.

That asymmetry is why the leak gets misdiagnosed for so long. Blended metrics make it worse. Total revenue is up, blended customer acquisition cost looks fine, and the aggregate view smooths over the fact that the highest contribution channel is quietly being drained by the lowest. The diagnosis only appears when you cut the data by geography, because geography is the one dimension both channels share.

Three tests that size the leak

Run these in order. The first is an afternoon of work and usually settles the question.

• Pin code overlay. Tag every D2C customer pin code as served or unserved by your quick commerce listings, using the platform serviceability list. Compare ninety day repeat rates for the two groups, for cohorts acquired before the listing and after. If served pin codes decay by five to ten points while unserved pin codes hold flat, the leak is real and you have just measured it. If both decay equally, you have a genuine product or lifecycle problem.
• City stagger. If you are still rolling out, hold one comparable city back for six to eight weeks. Compare D2C repeat and total brand contribution in the launched city against the held city. This is the cleanest read available and it costs nothing except a short delay in a single city.
• Phone number match. Where you have consent and a clean identity layer, match your D2C customer phone list against any first party data the platform shares in campaign or sampling programmes. Coverage is partial, but even a partial match tells you the direction and rough magnitude.

Look at frequency as well as repeat rate. A common finding is that migrated customers actually buy more often overall, because a ten minute delivery removes the planning friction that used to gate the purchase. Higher frequency at lower contribution per order can still be a good trade. You cannot know until you do the arithmetic.

The margin question you have to answer

Compare contribution per repeat order, not gross margin. On a Rs 700 D2C repeat order at 60 percent gross margin, subtract shipping of about Rs 70, payment fees, packaging and a small allocation of retention marketing, and you are usually left with 42 to 48 percent contribution, or roughly Rs 300 to Rs 340.

On quick commerce, take a Rs 250 order value at the same gross margin, then subtract platform margin of 22 to 30 percent, listing and fill related charges, damages and expiry deductions and visibility spend of 6 to 12 percent of sales. Contribution typically lands between 18 and 26 percent, or roughly Rs 45 to Rs 65 per order.

Now put frequency on top. If a migrated customer buys 2.4 times as often, the annual contribution comparison closes considerably but rarely fully. Run this per category, because a Rs 199 impulse pack and a Rs 1,200 refill pack produce completely different answers. The output you want is a single number per cohort: annual contribution per customer, D2C versus migrated. Everything else is commentary.

Defending without delisting

Delisting is almost never the answer, because the demand you would be walking away from does not return to your website. It goes to whichever competitor is on the shelf at 9 pm. The workable defences are all about giving the two channels different jobs.

Pack and format differentiation is the strongest lever. Put the small, impulse, single use and trial formats on quick commerce, where the shopper is buying for tonight. Keep the multi month refill, the value pack, the bundle and the limited edition on your own site, where the shopper is buying for the quarter. A brand selling a 30 day pack on both channels has built a leak on purpose.

Subscription is the second lever. A customer on an active replenishment plan with a scheduled delivery and a locked price has a reason not to reopen the question every month. Subscription penetration among repeat buyers is the single best predictor we see of how much a D2C cohort survives a quick commerce launch.

The third is to treat quick commerce as an acquisition surface rather than only a sales channel. Pack inserts, on pack codes and sampling programmes can bring a shopper who discovered you on an app into your own first party base at a cost far below paid social. If you are going to fund the platform anyway, get an identity out of it.

What to track every month

Four lines on one slide. Ninety day repeat rate, split by served and unserved pin codes. Subscription share of D2C repeat orders. Annual contribution per customer by channel. And the share of quick commerce buyers who have entered your first party base. If the first two hold and the last two rise, the leak has become a portfolio, which is the outcome you actually want.