How do you measure a run-of-site campaign?
Ask five questions in order, each with its own number: did the site deliver what you paid for, could people see the ad, how often, did they act, and did it pay. Clicks answer only part of one.
| Question | Number to check | Where it comes from |
|---|---|---|
| Did you get what you paid for? | Impressions delivered against the order | The site's report |
| Could people see it? | Viewable rate | The site's ad server, or Google's Active View |
| How often did each person see it? | Reach (unique people) and average frequency | The ad server |
| Did people act? | Clicks, and visits carrying your tracking tag | The site's report and Google Analytics |
| Did it pay? | Cost per lead and cost per booked job (or return on ad spend, if you sell online) | Your own records |
A served ad isn't a seen ad. Google measures viewability to the industry standard set by the Media Rating Council (MRC):
"A display ad is counted as viewable when at least 50% of its area is visible on the screen for at least one second." (Google Ads Help: Understanding viewability and Active View reporting metrics)
For video, it's 2 seconds of play. Even then, Google adds, Active View "can't guarantee that a user is looking at the screen at that time."
Expect few clicks. LocaliQ's 2026 display benchmarks, medians from 7,074 of its customers' campaigns run April 2025 to June 2026, put click-through rate between 0.12% (restaurants) and 0.40% (education). At 0.2%, 50,000 The number of times your ad or listing was shown, whether or not anyone clicked. bring about 100 clicks. That's why the last row, cost per booked job, is the one to manage.
What report should the website give you?
A weekly report, by day, split by ad and by page or section, with impressions, clicks and cost, plus viewable rate. Get the promise in writing before you pay.
All but viewable rate is the default in the standard terms US agencies use for online media buys:
"Reports will be broken out by day and summarized by creative execution, content area (Ad placement), impressions, clicks, spend/cost, and other variables as may be defined on the IO" (4A's/IAB Standard Terms and Conditions, Version 3.0)
Those terms say they don't cover "direct advertiser buys with publishers". If you buy straight from the site, you get what your insertion order (the order form, or IO) and the site's own terms say. In Canada, IAB Canada's 2013 terms are written for advertisers as well as agencies, with the same weekly report.
Put these in the order:
| Term | Why it matters |
|---|---|
| Impressions guaranteed, not just estimated | Under IAB Canada's terms, buys without guaranteed delivery, such as flat-fee or time-block buys, get no makegoods |
| Weekly report by day, ad and page or section | Shows which parts of the site bring leads |
| Viewable rate, and who measures it | Tells you whether the slots are seen |
| A frequency cap per person | Keeps one reader from seeing your ad all day |
| A makegood if delivery falls short | Under both standards, extra ads or a credit for the shortfall |
Sites can cap frequency in their ad server. Google Ad Manager, for one, can cap "the number of times a single viewer can be exposed" to one line of an order, per minute, hour, day, week, month or lifetime (Google Ad Manager API: FrequencyCap).
How do you track leads from a run-of-site ad?
Give the site a tagged link, count the actions you care about, and catch the people who never click.
- Tag the link. UTM parameters are short labels in a link that Google Analytics reads. Google says "you should always use utm_source, utm_medium, and utm_campaign", and suggests utm_content "to differentiate creatives" (Analytics Help: URL builders). For example: yoursite.com/spring-tune-up/?utm_source=dailyherald&utm_medium=banner&utm_campaign=spring_tuneup&utm_content=photo_a. Visits then show in the Traffic acquisition report. Stick to lowercase: Google treats "Meta" and "meta" as different values.
- Count leads, not visits. Mark calls, forms and bookings as key events in Google Analytics (how).
- Catch the non-clickers. Nearly everyone who sees a banner scrolls past; some search your name later. Watch three signals: searches for your business name in Google's free tool that shows which searches bring people to your site and flags pages Google can't index. More (how), a phone number or offer code used only in this ad, and a "How did you hear about us?" question on forms and calls.
- Follow leads to booked jobs. Note which leads booked and what they paid. That gives cost per booked job.
If you reach the site through a Google Ads Display campaign instead, Google reports unique users, average frequency and view-through conversions. Those, Google says, "happen when a customer sees an image or rich media ad, then later completes a conversion on your site without interacting with (for example, clicking on) another ad" (Google Ads API: Metrics reference). Treat them as a hint: some of those people would have bought anyway.
How do you know the ads caused the sales?
Switch them off and compare. Banners get credit for customers who would have come anyway, so a pause is the cheapest proof.
- On-off test: run a month, pause a month, run again. Compare tagged visits, leads, name searches and "heard about us" answers across the periods, and against the same weeks last year if your business is seasonal.
- Side-by-side test: run the site-wide buy and a targeted one, such as a single section, with the same ad, and compare Your marketing spend divided by the number of leads it brought in: calls, forms and bookings. More. TV ad platform Tatari ran this in 2021 for three advertisers' TV "run of schedule" buys, also called ROS. The broad buys delivered more impressions, and which buy won differed by company. Tatari suggests using the broad buy as a "control group" for the targeted one.
Which changes make run-of-site spend pay?
Fix the weakest step first, in this order: delivery, visibility, repetition, the ad, then the page.
| What you see | Likely cause | What to change |
|---|---|---|
| Fewer impressions than ordered | The site under-delivered | Ask for a makegood or a credit |
| Viewable rate well below the site's norm | Slots low on the page or in footers | Ask for in-article or top-of-page slots, or buy where you pay only for viewable impressions |
| Reach flat, frequency rising, clicks falling | The same readers see it too often | Cap frequency per person, or swap in a new ad |
| Low clicks in every section | The ad itself | Test two versions, one change at a time, each with its own utm_content |
| Clicks but no leads | The page or the offer | Send clicks to a page made for that offer (landing page) |
| One section brings most leads | Run of site found your readers | Buy that section next time |
| Cost per booked job above what a job earns, after a fair test | The channel | Stop, and move the money |
Google Ads already sells this way in Display campaigns: "If you'd like to pay only for ad impressions measured as viewable, you can with viewable CPM" (Google Ads Help: Bid on viewable impressions using viewable CPM). On a direct buy, ask the site whether it does.
Example
Illustrative. Numbers are invented, in US dollars unless marked.
A plumber on a local news site
$600 a month for a guaranteed 150,000 impressions. The month's report shows 120,000 delivered, 45% viewable, 210 clicks. Tagged visits: 160. Leads: 6 forms and 4 calls on the ad-only number, so $60 a lead; 3 booked, so $200 a booked job. The plumber asks for the missing 30,000 impressions as a makegood, placed in articles instead of the footer.
A dental office testing two ads
Each ad gets 50,000 impressions. Ad A, a smiling patient, draws 50 clicks and 1 booking. Ad B, "New patients seen this week", draws 110 clicks and 5 bookings, each counted by its own utm_content. B stays; a new challenger replaces A.
A Canadian bookkeeper on a business association site
C$400 a month, flat fee, no impression guarantee, so no makegood if delivery runs light. The bookkeeper gets the weekly report promised in writing, then runs two months on and one off. Name searches in Search Console fall by a third in the off month, with no seasonal dip the year before: the ads were likely reaching people who never clicked.
FAQ
What is a good click-through rate for run-of-site banner ads?
LocaliQ's 2026 display medians run 0.12% to 0.40% by industry. Few people click, so judge cost per lead, not click-through rate.
What viewable rate should I expect?
LocaliQ's medians on USA TODAY Network sites, July 2025 to June 2026: 62% high-impact display, 70% native, 71% video. Ask each site for its own.
What is a makegood?
Free extra ads, or a credit, when a site delivers fewer impressions than the order guaranteed. No guarantee, no makegood, so ask for guaranteed impressions.
How long should a run-of-site test run?
Long enough for each ad version to bring several leads. A week of clicks proves little.
Can I limit how often one person sees my ad?
Only if the site's ad server supports it and the site agrees. In Google Ads Display campaigns, you set it per day, week or month.
Different for you?
- Calls and forms aren't counted yet: GA4 key events for calls, forms and bookings.
- You want name searches as a signal: Search Console for a small business.
- Clicks arrive but nobody calls: what a landing page needs.
- Someone else runs your ads and reports to you: the report to ask for.
- You're deciding how much to spend at all: small business marketing budget.
- You'd rather have it run for you: Paid Ads Agent.