Ad Reporting That No Longer Adds Up

Three numbers, no errors

The advertising platform reports forty conversions. Your analytics reports twenty-six. Your own order system shows thirty-one sales.

The instinct is that somebody is wrong, usually the platform, usually deliberately.

In fact all three are reporting accurately about different questions, over different periods, using different rules about who gets the credit.

Understanding which question each answers is the difference between usable reporting and a monthly argument.

They are counting different events

The platform counts conversions it can connect to somebody who saw or clicked its advertising.

Your analytics counts conversions on your website and assigns them to whichever source it can identify at the moment of the visit.

Your order system counts orders, including phone orders, repeat customers, and anybody who arrived by a route nothing was measuring.

The third is the only complete count of the thing you actually care about, and it is the one businesses look at least often.

Attribution windows

The largest single cause of disagreement, and the one that has just changed.

A window is the period after seeing or clicking an ad during which a resulting sale is credited to it.

Platforms historically used generous windows, counting a purchase up to four weeks after a click and often crediting a view that led to no click at all.

Analytics tools typically credit the last source before the visit, over a shorter period.

So one tool credits an ad seen three weeks ago while another credits the search that brought them back yesterday, and both record the same sale.

Those windows have now shortened, which is why a campaign can appear to have declined without anything about it changing.

The same sale counted twice

A consequence worth stating explicitly, because it produces impossible-looking totals.

If somebody sees an ad on one platform, sees another elsewhere, then searches for you and buys, each platform may claim that sale in its own reporting.

Neither is being dishonest. Each is answering did somebody who saw my ad go on to buy, and for both the answer is yes.

Add the platform reports together and you get more conversions than you had sales, which is the clearest possible demonstration that these figures are not designed to be summed.

Modelled numbers

The newer element, and the one causing most of the current confusion.

Where a platform can no longer observe an outcome directly, it estimates it, using the conversions it can still see to infer the ones it cannot.

This is statistically reasonable and it produces a number that behaves differently from a count.

It is more reliable in aggregate than in detail, so a monthly total may be sound while a single day is noisy.

And it usually cannot be broken down, so asking which specific customers those conversions were is a question with no answer.

Where reporting distinguishes observed from estimated figures, look at both, and be more cautious about small numbers than large ones.

What to actually do about it

• Pick one number as the truth, and use your own records.
• Compare monthly, not daily.
• Track spend against total sales, not per campaign.
• Use the platform's figure for direction, not for magnitude.
• Ask customers how they found you.
• Never add platform totals together.

The first is the discipline that resolves this. One number is the business's number, and everything else is an indicator.

A worked example

A specialist retailer spent about four thousand a month across two platforms, which between them reported roughly a hundred and forty conversions.

Their actual order count for the month was ninety-one, including nineteen phone orders no platform could see.

Rather than reconciling, they changed what they looked at.

They recorded total monthly orders, total monthly spend, and the cost of the two together, and added one question to the checkout and the phone script asking how people had heard of them.

Platform reporting was still consulted, but only for direction: which campaigns were rising and falling relative to each other.

Within three months they could say what a customer cost them and could not say which platform deserved credit for any individual one.

The second question turned out not to matter, because the decision they actually faced was whether to spend more in total.

The half-visible half of your business

Worth naming, because it explains why online reporting understates local advertising badly.

Phone calls, walk-ins, people who saw an ad and mentioned it to a partner who then contacted you, and anybody who searched your name later on a different device.

None of that appears in any platform report, and for many local businesses it is most of the response.

This is not a measurement problem to be solved with better tools. It is a structural limit, and the response is to measure the whole business against the whole spend rather than trying to attribute each part.

Rule out a broken setup first

Before attributing a discrepancy to any of the above, confirm the measurement is actually working, because a broken tag produces exactly the same symptom.

The common faults are dull. A conversion event firing on a page nobody reaches, so it never records. The same event firing twice, on the confirmation page and again on a reload, so every sale counts as two. A tag removed during a site change and never replaced.

Each of these produces a figure that is wrong rather than differently defined, and no amount of understanding attribution windows will reconcile it.

The check is to complete a real transaction yourself, from a phone on mobile data, and confirm it appears once in each system within the expected time.

Do this after every site change, since a rebuild is the most reliable way to lose tracking, and the loss is silent.

The counter-case

Ignoring platform reporting entirely goes too far the other way.

Within a single platform, comparisons are meaningful, because the counting is consistent. If one campaign shows a cost per conversion three times another's, that is real, whatever the absolute figures.

Trends over time are also meaningful for the same reason, provided nothing about the setup changed in between.

What is not meaningful is comparing across platforms, adding them together, or treating any of them as a count of your business.

Use them for the decisions they can support, which are mostly about which campaign to favour, and use your own records for the decision about whether to be advertising at all.

Making it manageable

1. Write down which number is the truth.
2. Record spend and sales monthly in one place.
3. Add a how did you hear about us question everywhere.
4. Use platform figures for relative comparison only.
5. Note when attribution settings change, with the date.
6. Stop reconciling figures that were never comparable.

Step five prevents next quarter's investigation into a decline that was a settings change.

Measuring a call rather than a click is covered in measuring an ad that produces a phone call.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.