Attribution
17 Aug – 13 Sep · vs previous period: 20 Jul – 16 Aug
The platforms count more than the order book does, and this page does not resolve it. Both figures are here, side by side, and the module reports the one it can stand behind.
The rule
Revenue is counted from the shop's own orders, everywhere in this module, including on the screen you keep open during a campaign. The platform's figure is shown beside it and never instead of it, and the two are never averaged.
Over 17 Aug – 13 Sep
257more orders claimed than the order book holds
The platforms count 1,258. The order book contains 1,000. Everything this module reports is the second figure.
Meta Ads
₺57,144 spent — the one figure both sides agree on
Our order book · reported
- Orders
- 194
- Revenue
- ₺105,407
- Return on that spend
- 1.84×
Meta Ads claims · shown, not used
- Orders
- 314+120
- Revenue
- ₺170,760+₺65,352
- Return on that spend
- 2.99×+1.14
62 per cent more than the order book holds, and the two are never added: they are the same conversions counted twice, once by somebody who saw them and once by somebody who was told about them.
A one-day view-through window and a seven-day click window, across two apps and usually two devices. A person who saw the reel on a phone and ordered on a laptop that evening is one conversion to Meta and one order with no advertising on it to us.
Google Ads
₺80,625 spent — the one figure both sides agree on
Our order book · reported
- Orders
- 807
- Revenue
- ₺455,926
- Return on that spend
- 5.65×
Google Ads claims · shown, not used
- Orders
- 944+137
- Revenue
- ₺533,433+₺77,507
- Return on that spend
- 6.62×+0.96
17 per cent more than the order book holds, and the two are never added: they are the same conversions counted twice, once by somebody who saw them and once by somebody who was told about them.
Mostly a thirty-day click window on search, where the intent was already there. The gap is smaller because a search click and an order tend to be the same person on the same machine within the hour.
Where the platforms say the difference comes from
Their documentation about their own windows, applied to our gap. This is the one section on the page that is not a measurement of this shop, and it is labelled on every row.
Meta Ads
120 orders of difference
Seen, not clicked
53 · 44% of the gap
A one-day view-through window. Somebody watched the reel, did not touch it, and ordered that evening from the home page.
Two devices
37 · 31% of the gap
The ad was on a phone and the order was on a laptop. Our order book has one person; the platform has one person too, and they are not joined up in our data.
Nobody can say
30 · 25% of the gap
Cookies that did not survive, app tracking turned off, and modelling the platform does and does not explain. This row is the honest part of the page.
Google Ads
137 orders of difference
Thirty-day click window
80 · 58% of the gap
A click four weeks ago and an order today is still a conversion to Google. In our book it is an order with no advertising on it.
Two devices
30 · 22% of the gap
The same split as anywhere else, smaller because search tends to be one person finishing one thought.
Nobody can say
27 · 20% of the gap
Modelled conversions, where consent was refused and the platform estimated rather than counted.
These shares are the platforms’ estimate, not our measurement. They describe how a window behaves in general, applied to a gap we can only see the size of. Nothing in our data separates a view-through from a second device, which is why the last row of each list is the one that says so.
Both counts, over the period
The same conversions, counted by the order book and counted by the platforms. The distance between the lines is the whole page.
The two lines are never added. They are one set of orders measured two ways, and a total of them would count most of this shop's customers twice.
What this module will not do with two numbers
Written down because every one of them is something an advertising report is routinely asked to do.
It does not average them
One of these is a count of orders and the other is an estimate with a window on it. The number halfway between a measurement and an estimate is not a better measurement; it is a third number nobody can defend.
It does not report the friendlier one upwards
Across this period the platforms claim 257 more orders than the order book holds. Every figure this module sends to management, to the Sales Cockpit and to the weekly summary is the order book's.
It does not switch when the switch would flatter
A campaign reads one way in the order book and better in the platform's report. The desk never quietly moves to the platform's figure for that campaign and leaves the rest on ours — the whole module counts one way or it counts nothing.
It does not claim to know what would have happened anyway
Brand search is the clearest case: most of those people were going to find the shop. Nothing here pretends to separate the orders advertising caused from the orders it merely witnessed, because nothing on either side of this page can.
Trendyol and Hepsiburada both sell advertising inside the marketplace and neither exposes it in the seller API. The panels export a spreadsheet. Until that changes this module does not count that money, and the card does not claim it. That money is real and it is not on this page, which is the other thing this module will not do: count what it cannot read.