Google, Meta and email all claim the same order. How to decide the budget

Google Ads reports half a million, Meta a third of a million, the email tool another few hundred thousand, and the total is higher than the revenue the store actually took that month. Each figure is correct by its own system’s rules; the same order simply appears in several of them. Credited revenue is therefore not separate money you can split the budget by. Set the overall budget from the store’s revenue and margin; whether a particular channel brings additional orders only a controlled test can show.

One order, four claims

Ad platforms and email tools do not count orders; they count contacts with their own channel that were followed by an order. Each sees only its own contacts, and each uses a different time window:

These are defaults; your accounts may be set differently. What matters is that the windows overlap. A customer sees an Instagram ad on Monday, clicks a Sklik result on Wednesday – Sklik is the ad platform of the Czech search engine Seznam – opens a newsletter on Friday evening and buys on Saturday morning after clicking a Google ad. In Shoptet that is one order. Meta claims it for the view only if the purchase came within a day, but Sklik, the email tool and Google Ads each claim the whole order.

Leadhub, a Czech email tool, says so directly in its own documentation: the sum of revenue credited by Google Analytics, Meta and Leadhub is often higher than the store’s actual revenue, because each tool sees only its part of the customer journey. This is not a measurement fault but a property of measurement.

For email, the signal itself is also weak. Both Ecomail and Leadhub credit an order after a mere email open. Since 2021 Apple has prevented senders from knowing when a recipient opens an email in its Mail app. An open therefore does not reliably show that someone read the email, and crediting by opens is the most generous part of the email figure.

Why credited revenue must not be added up

An illustrative example with invented figures for one month, all excluding VAT:

Together the channels claim 118% of what the store took, and store revenue still includes orders from direct visits and organic search. An owner who calculated each channel’s return from its credited revenue would see advertising as far more profitable than it is. Against credited revenue, total spend would come to 11% of revenue; against the store’s actual revenue it is 13%.

The reverse also happens. The sum of credited revenue can fall below store revenue, for example when some purchases go unmeasured because visitors declined cookies. A low sum does not show what each channel really brought in either. The ratio of the sum to store revenue is useful only as a check on whether measurement has changed.

Credited revenue is not additional revenue

For a budget decision, what matters is incrementality: orders that would not have happened without the channel. Attribution does not answer that question, because customers who would have bought anyway also click ads and open emails.

Field experiments at eBay showed that the true return on paid search was a fraction of conventional non-experimental estimates. Ads on the company’s own brand name had no measurable short-term benefit, and ads did not influence frequent customers even though they accounted for most of the spend. A comparison of 15 experiments at Facebook found that common observational methods often failed to match the results of randomised tests.

The error does not only run one way. In 2011 Google estimated from more than 400 paused accounts that 89% of clicks from search ads were not replaced by organic results once the ads were switched off. That measures clicks, not orders, and the analysis came from the ad platform itself, but it shows that incrementality can be high. For email, 70 randomised experiments at a US ticket reseller found that emailed offers raised recipients’ spending by 37.2%, and 90% of the gain did not come through redeeming the offer. Crediting by clicks or codes can therefore understate what email does.

All of this research comes from large US companies, and its figures cannot be transferred to a Czech store. The lesson for decisions is simpler: you do not know in advance either the direction or the size of the gap between credited and caused revenue.

Anchor the budget in store totals

Because credited revenue cannot settle incrementality, the budget needs a fixed point that does not overlap. That point is the whole store’s revenue and margin.

Total spend against total revenue. Add up spend across all paid channels and divide it by the store’s revenue excluding VAT for the same period. This number cannot count anything twice. If you add budget and neither total revenue nor margin moves in the following weeks, the extra spend was probably buying mostly orders that would have come anyway – however many the platform claimed.

Use attribution within a channel. Comparing Google Ads campaigns by Google Ads figures, or newsletters by the email tool’s figures, makes sense because the same rules apply. Comparing Google Ads with email by their own figures does not.

No model is the referee. It is tempting to crown one model as the truth, such as the one in Google Analytics. But GA4 simply divides credit differently: by default it credits paid and organic channels and looks back up to 90 days, while Google Ads credits only Google paid channels. GA4 is useful for comparing channels because it measures them all with the same yardstick. It is no better at answering what would have happened without a channel.

When a test is worth it, and which one

A test costs money and time, and at low volume it often gives no clear answer. Even large experiments produce very wide estimate intervals because customer purchases fluctuate strongly, and Google recommends running experiments for at least four to six weeks, noting that low volumes give inconclusive results. A test pays off where a large share of the budget is at stake, or where you plan to scale a channel up or down substantially.

An email holdout group is usually the most accessible for an online store, because you control the contact list. The eBay research also used a switch-off test, turning off part of its advertising and measuring the result on total sales. Whichever test you choose, set the success criterion and the duration in advance and read the result from store revenue, not from the tested channel’s attribution.

When a test does not make sense, a controlled change remains: adjust the budget gradually, one channel at a time, and watch total revenue and margin. It is not conclusive proof, but it is a better basis than comparing credited revenue.

What to do now

1. Write down the crediting rules. For each channel, the window after a click, after a view or open, and what the channel can see.
2. Stop adding up credited revenue. Add up only spend and compare it with store revenue excluding VAT for the same period.
3. Find the biggest disputed decision. The channel with the largest spend, or the one whose credit is most questionable, such as brand ads or emails credited by opens.
4. Plan one test for it. Decide the test type, duration, the store-side metric and what you will do with each possible result.
5. Between tests, change the budget one channel at a time. Read the result from total revenue and margin, not from attribution.

In its PPC section, Korzaro adds up only the spend across Google Ads, Meta and Sklik and calculates true cost-of-sale from the whole store’s revenue in Shoptet. It shows the revenue each platform claims separately and compares their sum with store revenue only as a check on measurement. Revenue from email comes from Leadhub in its own section. Korzaro does not design or evaluate an incrementality test for you.