Read Google Ads results without fooling yourself

A Google Ads account will give you a number for anything you ask. The hard part is knowing which number describes what actually happened to the business. These are the practices I use to read an account, from which conversion column to trust to how I check the ad platform against the calendar, written so an owner can check an agency's reading and a marketer can apply them in the weekly review.

Know which conversion column you are reading

The Conversions column counts the primary conversion actions in the goals a campaign bids toward. All conversions counts those plus every secondary action, and it can include other conversion sources besides your lead action. The default reports show Conversions, so that is the number most people quote.

The real lead count sits in All conversions when the lead action was secondary for part of the period, or when a campaign bids toward a goal that does not include it. This happens most often around a launch, when tracking gets fixed while the ads are already running. In the accounts where I have promoted an action to primary, the change applied from that day forward. Leads recorded before it stayed out of the Conversions column and were in All conversions the whole time. Check this in your own account before comparing periods across a change.

Never quote All conversions as a total. Segment it by conversion action and read the row for the action that means a lead.
When the real count comes from All conversions, say so in the report: the count, the action, and the column.
After fixing the goal settings, confirm a few days later that new leads land in Conversions. The conversion goals guide covers the settings.

Give recent days time to fill in

Google Ads reports a conversion against the date of the ad click, not the date the lead arrived. Someone who clicks on Monday and books on Thursday adds a conversion to Monday, and it only appears on Monday once Thursday has happened. Depending on the conversion window, conversions can keep arriving for up to 90 days after the click, and some conversions take a day or more to show up in reports at all.

The result is that the most recent days of any report look worse than they will once the data settles, and a report pulled on the first of the month undercounts the end of the month before. When I compared two pulls of the same date range taken a day apart, spend and clicks were identical and the conversion count had gone up.

Write the pull date on every report, and expect conversions for recent days to rise.
Leave the most recent days out of any verdict. How many depends on how long this business's customers take to act after clicking.
When you need conversions by the day they happened, add the "by conv. time" versions of the conversion columns. Confirm the column names on your screen, because Google renames them.
When last month's numbers have moved, restate them in this month's report and say why.

Do not judge an automated strategy while it is learning

Google marks an automated bid strategy as Learning when it is new or reactivated, when its settings change, when campaigns or keywords are added to or removed from it, and when conversion actions are added or removed. Google's article on the duration of the learning period says calibration can take up to around 50 conversion events or three conversion cycles, faster with more data, and that the algorithms keep learning after the label disappears.

A new account on manual bids shows no Learning label, but it has the same problem in another form: nobody yet knows which searches, hours, and devices produce leads. The first weeks of any launch are a measurement period, whatever the bid strategy.

Before launch, agree with the owner that the first weeks are for measuring, and put the date of the first real verdict on the calendar.
Batch changes. Five small changes on five different days can send an automated strategy back into learning each time, and they make it impossible to tell which change mattered.
Keep a dated change log, and read it before blaming the market for a bad week.

Respect small samples

Most local accounts produce a handful of leads a week, and a handful cannot settle whether a keyword, an ad, or a campaign works. Chance alone produces long dry spells.

Here is the arithmetic with a made-up rate. Say a landing page turns about 1 in 20 ad clicks into a lead. The odds that a keyword gets 20 clicks and no lead purely by luck are about 36 percent, so that keyword has told you almost nothing. At 40 clicks with no lead, luck still explains it about 13 percent of the time. Only around 60 clicks, about 5 percent, does a keyword with no leads start to look like a real problem.

Clicks with no lead	Chance of that by luck alone, if 1 in 20 clicks converts
20	About 36 percent
40	About 13 percent
60	About 5 percent

Set pause rules before launch, in clicks or spend, based on the conversion rate you expect. A threshold chosen after you have seen the numbers is easy to move.
Judge a week against the trailing two to four weeks, not against the week before it alone.
Give each campaign a stop-loss: if it spends a set amount without a lead, pull its budget back and read the search terms before spending more.

Treat the search terms report as the truth about what you paid for

Keywords are what you asked for. The search terms report shows what people actually typed before your ad showed, and it regularly includes searches you never bid on, matched as close variants. It is the only place to see whether the money reached buyers or people looking for something else, such as jobs, free information, parts, or a service you do not offer.

Know its two limits. Google leaves low-volume searches out of the report for privacy, so the terms listed will not add up to the campaign's total clicks and cost. And the report says nothing about where the searcher was. Google's default location option shows ads to people who have shown interest in your area as well as people in it, so read the locations report alongside the search terms.

Read the report every week while an account is new, and at least every two weeks after it settles. Sort by cost, not by clicks.
Add wrong-intent searches as negative keywords the same day you find them.
Promote searches that produced leads into keywords of their own, so they get their own bids and ads.
For a business that serves a local area, set the location option to people in or regularly in your locations, and confirm in the locations report that clicks come from there.

Reconcile the ad platform against the calendar or CRM

Expect Google Ads, Google Analytics, and the business's own records to disagree, and know why each one counts what it counts.

Source	What it counts	Why it can differ
Google Ads	Conversions it can match to an ad click, in the column and goals you are reading	Misses leads when cookies are blocked or cleared, consent is declined, or someone comes back later on another device. Moves with primary and secondary settings. Dated by click.
Google Analytics	Visits that reached the confirmation step and were measured, credited by its own attribution rules	Can be higher or lower than Google Ads for the same traffic. Dated by the event. Includes your own test submissions unless you filter them.
Calendar or CRM	The appointments and leads that actually exist	The truth about volume, but it can only say which leads came from ads if a source is recorded on each one.

Ask the owner for the list of appointments or leads for the period. Name and date is enough to match most of them to ad activity.
Mark every test lead you create, in the form or in the CRM, so it can be excluded from every count.
When the counts disagree, report the one you can confirm as the headline, show the others as a possible range, and name the source that settles it. Usually that is the owner's own calendar.
Record a lead source on every CRM record, so next month's reconciliation takes minutes instead of an afternoon. How attribution works explains how the source gets there.

Expect the API and the interface to differ at the edges

Once you report on several accounts, pulling data through the Google Ads API is faster than reading tables in the interface, and it repeats exactly. It is fair to ask whether the two agree. When I checked an API pull against a hand pull from the interface for the same account and date range, spend, clicks, and impressions matched exactly. Every difference traced back to one of three causes, and they are worth knowing before you put any two tools side by side.

Timing

Conversions keep arriving for recent click dates, so a pull taken a day later can show more conversions for the same range.

Current values

Some fields describe the account as it is right now, such as Quality Score, ad strength, optimization score, and first-page bid estimates. They change from day to day and are snapshots, not history.

Row shape

The two can group the same data differently. The interface's search terms table can show a row for each combination of term, match type, and ad group, while an API query can return a row per keyword, so row counts differ even when the totals agree.

Pull both on the same day, for the same date range.
Reconcile totals first: spend, clicks, impressions, and conversions. Compare rows only after the totals agree.
Set a tolerance in advance, such as money to the cent and counts exact, and investigate anything outside it.
Choose one source as the report of record, and name it in the report with the pull date. In API queries, remember that metrics.conversions and metrics.all_conversions are different numbers, just as the two columns are.

Report what the owner can act on

An owner does not need impression share or Quality Score. They need to know how many leads or booked calls the ads produced, what each one cost, which searches produced them, what is holding the account back, and what happens next and when.

Lead with the count you can stand behind, and say where it came from.
Give cost per lead or cost per booked call, and compare it with the break-even line agreed before launch.
Show the few searches that produced leads, in the words people typed.
Put caveats and reconciliation after the numbers, never in front of them.
End with the next change and its date, then any asks.

Present early results honestly covers how to write that report after a launch, and the monthly report guide covers the one that follows every month.