Meta claimed it generated 157 of every 100 sales

VALCA
Niche
Beauty
Platform
Shopify

Valca is an Argentine natural dermocosmetics brand: regenerating creams and stretch-mark treatments, made with natural ingredients, approved by Argentina's health authority and sold through its own online store. It is a large store — in the tens of thousands of paid orders a month — and it grows the way most e-commerce grows today: by buying traffic on Meta.

This story does not end with “the return went up”. It ends somewhere less sellable and more useful: the number they looked at every day was false, and it became true. It is the one kind of improvement nobody can fake, because the number went down.

Results

Before
For every 100 sales the store took, Meta said it had generated 157. With the return inflated, every decision to scale or pull back was off.
What changed
We connected the Escalafy pixel to their Meta account on 5 September, so the platform would receive every sale with the identifying data attached. The score Meta gives each event based on how much data reaches it went from 7.39 to 8.30 out of 10.
Today
Meta claims 93%. Real cost per purchase fell 11% and money-in-over-money-out came out unchanged: 2.455 before, 2.474 after. The ROAS Meta reports fell 32%, and that was the point: what disappeared are sales it was claiming without having caused them.

What was actually broken

When someone buys from your store, the browser tells Meta, and Meta uses that message to decide whether the sale was its own. That is the theory. In practice ad blockers, Safari and iOS cut a good part of the message, and what arrives arrives incomplete: the sale shows up, but without the email, phone, name or city of whoever bought.

When the platform is missing data it does not leave the box empty: it estimates. And it estimates in its own favor. That is how Meta ends up claiming more sales than the store actually had.

Valca arrived with exactly that problem, which anyone who spends on Meta will recognize. Meta showed them a return of 3.3 — that is, 3.3 pesos of sales for every peso of ad spend, the piece of arithmetic the industry calls ROAS. The real one was 2.2.

Decisions get made on that gap every single day. Which campaign to scale, how much to pull back, which ad to switch off. If the platform reports a 50% higher return than you actually have, every one of those decisions is off.

The Escalafy pixel sends that same event from our side, with the identifying data attached, so the platform does not have to guess. It does not invent sales: it takes the ones that already exist and hands them over complete. On 5 September we connected it to Valca's account.

What changed

In the 69 days before the switch, Meta claimed 157% of Valca's real web orders: for every 100 sales the store took, the platform said it had generated 157. In the 69 days after, 93. It went from inventing half a business to sitting slightly below reality, which is where it should be.

Meta scores every event from 1 to 10 based on how much data reaches it to recognize the person behind it. It calls this Event Match Quality, and it is the most direct way to see whether the platform is understanding you or guessing. Valca's pixel went from 7.39 to 8.30 — and that 7.39 is one of the few genuinely clean baselines we have, because it predates anything of ours.

Real cost per purchase fell 11%.

And the honest piece of arithmetic did not move: 2.455 before, 2.474 after. That is total revenue over total ad spend — money in over money out, regardless of what the platform decides to take credit for. It is the one number nobody can inflate.

The other half has to be said: the ROAS Meta reports fell 32%. That was the point. What disappeared are sales Meta was claiming without having caused them, not real sales — money against money came out the same. If you open Ads Manager and compare against before, you will see the lower number. The lower number is the correct one.

And email

An email flujo is an automatic send triggered when someone does something in the store: views a product, fills a cart, starts checkout and does not finish. Klaviyo, the tool Valca uses, only sends it when it knows who that visitor is. Anyone it cannot identify gets nothing — and that never shows up in a report, because an email that does not go out is not counted.

In October we duplicated five of those flujos. The duplicate is a copy that runs alongside the original and uses the pixel's identification to reach the visitors who used to fall outside. Nothing is touched on the flujo they already had.

The clones added 20.4% on top of what those same flujos were already bringing in, reaching 23% more profiles.

Stop operating on incomplete data.

Connect your stores, your payouts and your ads in 5 minutes and see today's real profit — not the one you find at month end.