Quick answer
An ad server is the system that decides which ad fills each slot on a page or app, delivers it and counts impressions and clicks. Publishers run one to manage inventory; advertisers use a third-party one to verify delivery across publishers.
Formula: CTR = clicks ÷ impressions; fill rate = filled impressions ÷ ad requests; vCPM = CPM ÷ viewability rate
What is an ad server?
An ad server is the system that decides which ad goes into which slot on a publisher's page or app, delivers that ad, and counts what happened. Microsoft's Xandr documentation describes ad serving as the process of determining which advertisement goes in which ad slot and then delivering it; the computer or group of computers that does this is the ad server. Most large ad servers also record how many ads were served and on which pages, record performance data such as clicks, and feed that data back into the choice of the next ad.
An ad server is not a marketplace. It holds the campaigns, the creatives and the rules (targeting, frequency limits, dates, priorities) and answers one question many millions of times a day: for this request, which ad? Where several buyers compete for the same impression, an ad exchange runs the auction; the ad server is where the winning ad is chosen, delivered and counted.
Publisher ad server vs advertiser ad server
There are two kinds, and the same ad is often counted by both.
| Publisher ad server | Advertiser (third-party) ad server | |
|---|---|---|
| Run by | The site or app owner | The advertiser or its agency |
| Decides | Which advertiser gets each impression | Which creative the advertiser puts in the slot it was given |
| Typical settings | Priorities, guaranteed volume, price floors, frequency limits | Creative rotation, click trackers, cross-publisher reporting |
| Counts | Impressions and clicks on that publisher's inventory | Impressions and clicks for the campaign across every publisher |
| Used to | Invoice advertisers and manage inventory | Verify delivery and compare publishers on one yardstick |
Xandr's documentation adds a third case, an intermediary ad server that tries to maximize the benefit to both sides.
Ad server vs ad network vs ad exchange
| Ad server | Ad network | Ad exchange | |
|---|---|---|---|
| What it is | Software that picks, delivers and counts ads | A broker that aggregates inventory from many publishers into one buy | A marketplace that auctions each impression |
| Does it sell inventory? | No | Yes | Yes, via auction |
| Who uses it | Publishers, networks and advertisers | Advertisers who want a consolidated buy | DSPs and SSPs on either side |
| Example question it answers | Which ad do I show here? | Which package fits my budget? | Who pays the most for this impression? |
See ad network and programmatic advertising for the buying side.
The metrics an ad server reports
| Metric | Definition | Formula |
|---|---|---|
| Ad request | A request from a page or app for an ad to fill a slot | counted per slot per load |
| Fill rate | Share of ad requests that returned an ad | filled impressions ÷ ad requests |
| Impression | A creative served to a single user at a single point in time | counted by the server |
| Click | A user clicking the creative, recorded at the redirect | counted by the server |
| CTR | Click-through rate | clicks ÷ impressions |
| Viewable impression | An impression that met the viewability standard (see below) | measured by script |
| Viewability rate | Share of measured impressions that were viewable | viewable ÷ measured |
| CPM, CPC, CPA | Price per 1,000 impressions, per click, per acquisition | spend ÷ the denominator |
Xandr's glossary defines an impression as a creative served to a single user at a single point in time. Viewability asks whether the ad could have been seen. For standard display ads, the Media Rating Council (MRC) guideline counts an impression as viewable when at least 50% of the ad's pixels are on an in-focus browser tab in the viewable space of the page for at least one continuous second after render; video needs 2 continuous seconds at the same pixel threshold. For display ads of 242,500 pixels or more, the guideline allows a 30% pixel threshold.
A worked example: one campaign report
Sample data: one display campaign, one week, from the advertiser's ad server.
| Item | Value | Calculation |
|---|---|---|
| Ad requests (publisher side) | 1,250,000 | |
| Impressions served | 1,000,000 | fill rate = 1,000,000 ÷ 1,250,000 = 80.0% |
| Clicks | 2,200 | CTR = 2,200 ÷ 1,000,000 = 0.22% |
| Measured impressions | 900,000 | |
| Viewable impressions | 486,000 | viewability = 486,000 ÷ 900,000 = 54.0% |
| CPM paid | $5.00 | |
| vCPM (cost per 1,000 viewable impressions) | $9.26 | $5.00 ÷ 0.54 |
| CTR on viewable impressions | 0.45% | 2,200 ÷ 486,000 |
Two readings. First, at a 54% viewability rate, the real price of an ad someone could actually see is $9.26 per thousand, not $5.00. Second, the CTR on impressions that were viewable is about double the headline CTR, because ads nobody could see cannot be clicked. Compare publishers on vCPM, not CPM, when viewability differs.
Why ad server clicks and site sessions do not match
Sample data, continued: the same campaign sent clicks to a landing page that was tagged with UTM parameters. The site's analytics recorded 1,850 sessions from that campaign.
Click-to-session rate = 1,850 ÷ 2,200 = 84.1%, so the gap is 350 clicks, or 15.9%.
That is normal. Google's Ad Manager documentation says that, with third-party ad servers, campaign variances of up to 20% are common and lists the causes: different impression counting and invalid-traffic filtering, users leaving before the page or the creative loads, brief network failures, ad blockers that stop delivery after a click or an impression has been counted, broken click-tracker setup, and passbacks between servers (each additional server increases the expected variance). A separate Google page on comparing Ad Manager with Analytics treats a gap above 10% as significant and adds mismatched time zones, filters and consent settings to the list.
A rule to apply before arguing with a vendor: first compare the same date range and the same time zone; then check the click tracker on the creative; then expect up to 10% as ordinary noise and investigate above that. Small campaigns amplify the percentage: Google notes that a small absolute difference looks large on low-volume line items.
Reconciling your ad server in TapCub
TapCub does not serve, rotate or count ads. It measures the visit that follows a click, so you can reconcile with the ad server's click count.
- Put UTM parameters on every ad's destination URL (source, medium, campaign), following name UTM parameters.
- Open Web analytics › Sources and use the UTM tab to read sessions for the campaign. Compare that to the ad server's clicks: in the sample, 1,850 ÷ 2,200 = 84.1%.
- Check the Channels tab for the channel the traffic landed in. Visits with ad click IDs such as
gclidormsclkidare classed as paid search by default; rules you add for the site take priority. If the campaign's visits show up as direct, a redirect may have dropped the UTM tags. - Import the campaign's spend on the Costs tab of All-platform analytics › Business › Channel ROI. TapCub's cost per click is spend ÷ sessions, so it reads higher than the ad server's cost per click whenever some clicks never become sessions. In the sample, spend is 1,000,000 ÷ 1,000 × $5.00 = $5,000: the ad server shows $5,000 ÷ 2,200 = $2.27 per click, while sessions give $5,000 ÷ 1,850 = $2.70. Use one definition when comparing channels.
- Read conversions per campaign on the Channels tab, or conversions and revenue on Channel ROI. That is the number the ad server cannot give you.
Sources
- Microsoft Xandr, Introduction to ad serving (updated October 22, 2025) and Online advertising and ad tech glossary.
- Media Rating Council, Viewable Ad Impression Measurement Guidelines, Version 2.0 (August 18, 2015).
- Google Ad Manager Help, Investigate report discrepancies and Google Analytics and Ad Manager discrepancies.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Experience Specialist
Explores usability, user journeys, and feature adoption through practical examples of onboarding and everyday product use.