The figures, exactly

Every table in the product computes these the same way, from the same two sources: clicks, and conversions. This is what each one counts.

The three worth reading before the rest are conversions, because it counts approved ones only; clicks, because a click-out from a landing page is not one; and the date range, because a click and its conversion can fall in different ones.

#Counts

#Clicks

A visitor who lands on a page and presses the button produces two rows: the arrival, and the click-out. Only the first is a click.

That is what keeps a landing page from doubling your traffic. It is also why LP clicks is a separate column rather than part of the total — the two describe the same visitor at two moments, and adding them together would count that person twice.

Bots are excluded from Clicks and counted in their own column. Nothing is silently dropped: a click judged automated is stored, with the reason, and can be read in the clicks log.

#Conversions

Approved only. A conversion the network reported but has not approved sits in Pending and contributes nothing to Conversions, Revenue, CR, EPC, ROI or Profit.

This is the single most common surprise, and it is deliberate. A network that reports a lead and rejects it a week later has not paid you, and a tracker that books it as revenue on arrival tells you a campaign is profitable during exactly the days you are deciding to scale it.

It also means a network sending a status word UTMCAP does not recognise holds the conversion as pending rather than guessing. See postbacks.

#Money

Cost arrives on the click, from the traffic source's cost macro — so a source with no cost macro configured records every click at zero, and the campaign reads as pure margin. See traffic sources.

Bot clicks cost nothing here. They are excluded from the cost sum as well as from the click count, so a bot wave does not inflate spend. Whether the platform charged you for them is between you and the platform.

EPC is per click, not per unique click and not per LP click.

#Rates

CR divides by clicks, not by unique clicks. On a campaign with a high repeat rate the two differ, and this is the one the product reports.

LP CTR is how many of the people who arrived pressed the button. On a campaign with no landing page — the offer direct — it is zero, and that is correct rather than missing.

#Where a divide by zero goes

Every rate and average is 0 when its denominator is 0, not blank and not infinite.

That is worth knowing because a row reading ROI 0% means one of two different things: a campaign that made exactly what it spent, or a campaign with no cost recorded at all. Cost in the same row separates them, which is why the two columns are worth reading together.

#Which date a row falls under

A click is counted on the day the click happened. A conversion is counted on the day the conversion arrived.

They are usually the same day and sometimes not: a network that approves on a weekly cycle reports Monday's sale on Friday. Over a range of a month this is invisible. Over a range of one day it is not, and it has a specific look — conversions with no clicks to explain them, or a CR that seems impossibly high or low.

If a narrow range looks wrong, widen it before looking for a bug. If it is still wrong widened, the conversions log shows each conversion with the click it attached to.

#What a report can be grouped by

Sixteen groupings in the report builder, and no others:

• Campaign
• Country
• Device, Operating system, Browser
• ISP
• sub1 – sub10 — whatever your traffic source put there

The sub slots are the useful ones once a campaign is running, because they carry the platform's own identifiers — ad set, creative, placement, keyword. What is in each is decided by the source's macros.

Landing page, offer and traffic source are not in that list. They are on the dashboard instead, as four small tables shown at once — because the question they answer is "where did the money come from", which you ask by comparing them rather than by pivoting on one.

Region, city and device brand are groupings nowhere, though a routing rule can test all three and a token can pass them on. They are recorded on every click and are columns in the clicks log — just not something a summary can be pivoted on.

Rows come back best-first on profit, capped at 1000 per report. A grouping that would exceed that — sub1 on a large account — is telling you to narrow the range or filter to one campaign first.

#Two numbers that look wrong and are not

Revenue below what the network's dashboard says. Almost always pending conversions. Compare Conversions + Pending against their count, not Conversions alone.

Clicks below what the platform reports. Three ordinary causes, in order of likelihood: bots excluded here and not there; the platform counting an impression-click that never reached the page; and clicks the platform recorded after your daily budget stopped serving. The clicks log is where the difference becomes visible.