Why the numbers need care
An ad report is a long list of figures that look precise. Precision is not the same as meaning. Some numbers describe what the system did, some describe what people did, and some describe what the system believes people did. Knowing which is which helps you avoid decisions based on the wrong figure.
Exact names vary from service to service, so check each tool’s own definitions. The ideas below are common to most.
Delivery numbers: what was shown
- Impressions: the number of times the ad was displayed. One person can account for several.
- Reach: the number of different people who saw it at least once.
- Frequency: average impressions per person reached.
These tell you about exposure, not about interest. An ad can be shown a great deal and still be ignored. How these figures differ for unpaid content is covered in the discussion of organic reach. One caution: an ad counted as shown does not guarantee the person looked at it. It may have appeared briefly on screen while they scrolled.
Interaction numbers: what people did on the ad
- Clicks: the number of times people clicked. Some tools count every click, including taps on the profile name or a “like,” while “link clicks” count only those that lead to your destination. Check which one is shown.
- Click-through rate (CTR): clicks divided by impressions, usually shown as a percentage.
- Engagement: a bundle that may include reactions, comments, shares and saves.
A hypothetical example of CTR: if an ad had 2,000 impressions and 50 link clicks, the CTR is 50 divided by 2,000, which is 0.025, or 2.5 percent. A calculator makes this easy to check for your own numbers.
There is no single good CTR. Typical values differ by industry, placement and objective, and an unusually high rate can come from a catchy but misleading ad that attracts curiosity clicks and few real buyers.
Cost numbers: what you paid
- Cost per click (CPC): total spend divided by clicks.
- Cost per thousand impressions (CPM): the cost of 1,000 displays.
- Cost per result: spend divided by the outcome you chose as the objective.
Take a hypothetical spend of $40 that produced 50 link clicks. The CPC is $40 divided by 50, which is $0.80. Cheap clicks are not automatically good. If those clicks come from people who leave immediately, a low CPC just means you are paying little for little. Judging an ad on cost per result tied to your actual aim is usually more informative than judging on cost per click.
Outcome numbers: what happened afterward
- Conversions: actions you define as valuable, such as a purchase, a sign-up, a form submission or a message sent.
- Conversion rate: conversions divided by clicks (or visitors).
- Cost per conversion: spend divided by conversions.
Continuing the hypothetical: if 50 people clicked and 2 signed up, the conversion rate is 2 divided by 50, or 4 percent. If the spend was $40, the cost per sign-up is $20 (40 divided by 2). With only 2 sign-ups, this figure could swing a lot by chance, so treat it as a rough indicator and not a stable fact. The caution about small samples is explored in the article on split testing.
Numbers that are easy to misread
Reach versus audience size. Reaching 1,000 people does not mean 1,000 customers or even 1,000 readers.
Clicks versus visits. Some visitors click but leave before the page loads, so the number of clicks reported by the ad tool may be higher than the visits recorded on your own site. Small mismatches are normal; large ones are worth investigating.
Attribution. A conversion is credited to an ad based on a rule, such as “within seven days of a click.” Different tools use different windows and methods, and several sources can claim credit for the same sale. Be wary of adding up conversions from different reports. Measurement is also affected by privacy settings that limit tracking, so figures can be incomplete.
Engagement counts. Likes are easy to give and weakly connected to buying. High engagement can be nice without producing any business result.
Averages. An overall average can hide the fact that one ad did well and another did poorly. Look at ads individually before drawing conclusions.
Time. Early numbers fluctuate. A single day is rarely enough to judge.
Reading a report in a sensible order
- Start with your goal and the single result that represents it.
- Check spend and the dates covered.
- Look at delivery: did the ad reach a reasonable number of people?
- Look at interaction: did those people respond?
- Look at outcomes: did responses lead to what you wanted?
- Compare ads against one another, not against an imagined benchmark.
- Note anything unusual, such as a holiday or a policy change.
Following the funnel from exposure to outcome helps you locate where the weakness is. If many people see the ad but few click, the message or image may need work. If many click but few convert, the page or offer may be the issue.
Beyond the dashboard
Ad numbers describe ad behavior, not the whole business. Customers may see an ad and buy later through another route, or hear about you from a friend. Combining ad reports with your own records, as discussed in the overview of social analytics, gives a fuller picture.
Reading reports calmly, with definitions in hand and a healthy doubt about small numbers, is a skill that improves with practice and keeps expectations realistic.