The Revenue We Deliberately Don't Claim

In short: Our attribution window is 24 hours from a widget click, with cart continuity required. Behind that public number we keep a private one: a ledger of every order line that looked widget-influenced but failed the test. Over six weeks on one store it collected 177 lines. Exactly four of them could be proven: the same cart, carrying a widget click past the window, twice by barely an hour and once for twenty-two days. We decided merchants will never see that second number, and this post is about why.

The question the last two posts left open

Post 13 in this series argued that you should read any app's attributed revenue with the window in one hand and the methodology in the other. Post 15 showed that on many Indian stores the order population itself is partial before any window math starts.

Both posts point a question back at us. If a strict 24-hour window is the honest choice, it must be leaving real revenue unclaimed. How much? We didn't want to estimate. So we built an instrument to measure it.

What the shadow ledger records

A claimed sale at Angadi has to pass every test: a shopper clicked a product in the widget, that product reached a completed order within 24 hours, and the cart chain connecting the click to the order is intact. Fail any one of those and the sale earns zero credit in the merchant's dashboard.

But failing the test is information too. So every near miss gets written to a ledger the merchant never sees, with its evidence attached: which test it failed, the time gap, and whether the cart chain survived.

What the evidence shows	Example	Where it goes	Counted as revenue?
Click to purchase inside 24 hours, cart chain intact	Shopper clicks a pairing at noon, orders that evening	Claimed. The dashboard number.	Yes
Same cart provably carries the click past 24 hours	A cart held for weeks, then purchased	Ledger, proven tier	No
Ordered product matches a clicked product, chain unprovable	A bestseller bought four days after someone clicked it	Ledger, unproven tier	No

The third tier is the important one to be suspicious of. Bestsellers land in orders whether a widget exists or nothing does. Treating "the product matches" as "the widget did it" is where inflated attribution begins, and most of the ledger lives in that tier.

Six weeks of near misses

The numbers, from one store we work with, as of 4 August 2026, covering the six weeks since attribution went live there:

177 near-miss lines total. The overwhelming majority, 173 of them, sit in the unproven tier: a clicked product turning up in a later order with no surviving chain between the two events. Some of those were probably real influence. Some were certainly coincidence. The ledger can't tell them apart, which is exactly why none of them are claimed.

Four lines had proof. The cart chain was intact, the clicked product was in it, and the purchase simply arrived after the window closed. Two of the four missed by barely an hour. One arrived four and a half days later, one after twenty-two days. Against the store's claimed attribution lines over the same period, those four amount to just under nine percent, and a near-identical share by value. That is the measured cost of the strict window: small, and known, which beats large and imagined in either direction.

The cart that waited three weeks

One of the four deserves its own paragraph. A shopper clicked a piece inside a look, added it, and left. The cart survived. Twenty-two days later the same cart checked out, that piece still in it.

By any common-sense reading the widget influenced that sale. We still don't claim it. A rule you bend for the sympathetic case stops being a rule, and the whole value of the 24-hour number is that nobody has to wonder what got bent to produce it. The three-week cart is the price of that, paid in public, right here.

The two lines that missed by an hour are the harder test, and they get the same answer. A window with a negotiable edge is not a window.

The second number we chose never to show

The obvious product move is to surface the ledger. Call it "assisted revenue" or "extended influence," put it next to the strict number, let the merchant feel the full impact.

We wrote that option down, argued about it, and killed it. A few reasons, in order of weight:

The two numbers get added. Every merchant, every time, in their head if nowhere else. The strict number stops meaning anything the day a bigger sibling stands beside it.

Every inflated convention in this category began life as an honest-looking secondary metric. View-based credit was once a footnote too.

Even the proven tier can't demonstrate the sale would have been lost without the widget. That gap between attribution and incrementality was the core of post 13, and it applies to our own near misses hardest of all.

So the ledger stays internal. One number faces the merchant. The rest is instrumentation.

What the ledger is actually for

Two jobs, both internal.

It audits the window. If proven-beyond-24h lines were a third of the ledger instead of four lines out of 177, the window would be wrong and we'd change it. The point was never that 24 hours is sacred. The point is that the choice is measured, and the measurement currently says the window costs very little truth.

It's an early-warning system. If shoppers start holding carts longer, in festival season for example, the proven tier grows before anything else moves. We'd rather learn about a behavior shift from our own ledger than from a merchant asking why the numbers feel low.

How to use this on any app you evaluate

You don't need our ledger. You need the question it answers. Ask a vendor: what happens to sales that almost matched your attribution model but didn't?

A vendor with no answer has never measured their own miss rate, so their claimed number has an error bar nobody has looked at. A vendor whose answer is a second, bigger number on your dashboard is inviting you to add the two together. The answer you want to hear is a model, a documented decision about the near misses, and one number.

---

Full disclosure: Angadi is our product. The figures in this post come from a store we work with, shared in aggregate and anonymized, and were re-verified against our database on 4 August 2026, the publication date.