What happens when a page requests an ad
The chain starts in the publisher's ad server. A person opens a page; the publisher's page code asks the publisher's ad server, typically Google Ad Manager, which line item should fill each slot. If the winning line item is a campaign the advertiser serves through CM360, the publisher's ad server does not return a creative. It returns the advertiser's placement tag, the small piece of code CM360 generated when the placement was trafficked and the publisher pasted into their line item.
The browser runs that tag, which calls CM360. CM360 identifies the placement from the tag, looks at the ads assigned to it, discards any outside their dates or excluded by targeting or a frequency cap, applies the rotation rule among those that remain and chooses one ad and one of its creatives. It then returns the creative, either directly or as a redirect to where the asset is hosted, and the browser renders it. When the creative begins to render, CM360 records an impression. If the person clicks, the click goes first to CM360's click redirect, which records it and then forwards the browser to the landing page set on the ad.
In a programmatic buy the first half of the chain changes and the second half does not. The demand-side platform (DSPSoftware advertisers use to buy programmatic inventory.) wins the auction and returns, as its creative markup, the same CM360 tag; the browser calls CM360 exactly as it would from a direct placement. That symmetry is why one ad server can count direct and programmatic buys side by side, which the programmatic ecosystem guide places in the wider picture of exchanges, supply-side platforms and buyers. The cost of the chain is latency: each hop is a network round trip, and a slow redirect can mean the person has scrolled past before the creative draws.
Ad tags, redirects and counting
CM360 documents several tag formats, each a different trade-off between capability and what the publisher's page or the buying platform can run. The trafficker chooses the format per placement when the tags are exported.
| Tag format | What it is | What it supports |
|---|---|---|
| Standard tag | The element-based tag CM360 recommends for display; loads CM360's script on the page | Every display creative type, including rich media and HTML5; impression and click counting |
| JavaScript tag | A script tag that calls CM360 directly | Most creative types; the older form of the standard tag |
| Iframe and JavaScript tag | Wraps the ad in an iframe, with script inside it | Isolation from the publisher page; most creative types with some rich media limits |
| Image tag | A plain image request wrapped in a click link | Image creatives only; impression and click counting; works where no script may run |
| Internal redirect | A creative that points CM360 at an asset hosted on another server | Counting an asset the advertiser does not upload |
| Click tracker | A link that records a click and forwards to the landing page | Click counting on text links, email and placements where no creative is served |
| Tracking ad | An impression pixel without a creative | Counting impressions that another system serves |
| VAST tag | A URL that returns VAST markup to a video player | In-stream video: the player fetches the creative and reports video events |
Two pieces of the tag deserve attention. The cache buster is a parameter the publisher's ad server fills with a random value on every request, so that the browser never serves a cached copy of the ad and every request reaches CM360 to be counted. The click macro is a placeholder the publisher's ad server replaces with its own click-tracking URL, so that both systems count the same click. A tag pasted with the placeholders left unfilled still serves, and its impressions and clicks are quietly undercounted.
Counting rules are the ad server's definition of truth. Google documents that CM360 counts a display impression when the creative begins to render, in line with the industry guidelines from the Media Rating Council, rather than when the request was answered. A click is counted when the browser reaches CM360's redirect. A video impression and the quartile events are reported by the player through the VAST response. Click rate (CTRShare of impressions that result in a click.) in CM360 is therefore clicks on the ad server's redirect divided by impressions the ad server saw begin to render, which is not the same ratio the publisher calculates from its own counts.
Why publisher and advertiser counts differ
They always differ. The question is whether the gap is normal and stable or large and growing. Both the CM360 and the Ad Manager help centres document a set of causes, and the Ad Manager documentation names a range of discrepancy it considers acceptable; the documented threshold is the point to check against, rather than a figure remembered from a previous campaign.
Different counting points. The publisher's ad server usually counts when it answers the request. CM360 counts when the creative begins to render. Every person who leaves the page between those two moments is an impression for the publisher and nothing for the advertiser.
Latency and abandonment. Each redirect in the chain takes time. Slow pages, slow networks and heavy creatives widen the window in which people leave.
Blockers and script restrictions. Browser extensions and privacy settings that block ad scripts stop CM360's tag from running after the publisher has already counted the request; an image tag survives more of these than a script tag does.
Time zones and reporting days. The two systems close their reporting day at different hours unless both are set to the same zone, so daily totals shift across the boundary even when the period total matches.
Invalid traffic filtering. Each system removes traffic it classifies as invalid using its own method and at its own stage, so the removed amounts differ.
Implementation errors. A missing cache buster lets browsers reuse a cached ad, a wrong tag type on a placement drops creatives that cannot run in it, and tags pasted into the wrong line item serve to inventory the plan never bought.
Investigation follows the same order each time: confirm the time zones match, compare day by day rather than by period total to see whether the gap is constant or started on a date, break the comparison down by placement to find whether one tag is responsible, and then inspect that tag's implementation on the live page. The reporting guide covers the report that gives the placement-level figures for this.
Reconciling a homepage takeover
A publisher invoices 2,400,000 impressions for a three-day homepage takeover; CM360 reports 2,196,000, a gap of 8.5%. Day by day the gap is 6% on day one, 7% on day two and 13% on day three, so it is not a constant counting-point difference. By placement, the mobile web placement is 19% short while desktop is under 5%. The trafficker opens the mobile page and finds the publisher pasted the desktop standard tag into the mobile line item, where a heavier creative loads slowly over mobile networks and a share of people scroll away before it begins to render. The publisher swaps in the correct tag with the lighter creative; the day-four gap on mobile falls to 6%. The invoice is settled on the ad server count for the three days, as the insertion order specified.
Rotation, weighting and targeting rules
An ad in CM360 is a set of rules that decides whether and what to serve into a placement. Each has start and end dates, a priority, a rotation setting for its creatives, optional frequency caps and optional targeting. Every placement must also carry a default ad, which serves whenever no other ad qualifies; it is the safety net that stops a slot going blank, and a placement where only the default is serving is the classic sign of a trafficking error.
Priority is evaluated first. Ads with a higher priority are served ahead of those with a lower one; rotation only applies among ads of equal priority. Rotation then chooses among the creatives assigned to an ad. CM360 documents even rotation, which serves each creative in turn; weighted rotation, where the trafficker sets a weight per creative; sequential rotation, which shows creatives in a set order to the same browser; and optimised rotation, where CM360 shifts delivery towards the creatives performing best on the metric chosen, such as click rate or conversions. Frequency caps limit how many times an ad is served to one browser over a period, counted on the ad server's own cookie.
Targeting restricts when an ad is eligible: geography, browser and operating system, language, day and time, keywords passed on the placement, and audience lists built from Floodlight activity or shared from other Google products. Targeting in the ad server works on the request after the placement has already been bought; it cannot buy different inventory, only choose which creative to show, or decline to show one, on inventory the plan already pays for. That distinction separates it from targeting in a demand-side platform, where the targeting decides whether to bid at all.
The settings are evaluated on every request, so a change made in the interface takes effect on the next call without the publisher re-trafficking anything. That central control is the operational reason an advertiser accepts the extra redirect.
Third-party tracking and verification pixels
An advertiser often needs a second system to count alongside the ad server: a verification vendor measuring viewability, brand safety or invalid trafficAd impressions and clicks not caused by genuine human interest: bots, fraud and accidents.; a measurement partner running a brand study; or a second ad server during a migration. CM360 supports this through third-party tracking URLs on creatives and ads. An impression tracker fires when the creative serves, a click tracker fires through the redirect chain, and in-stream video creatives can carry trackers for the documented video events such as start, quartiles and completion.
Each added tracker is another request from the browser, so each is subject to the same abandonment and blocking that separate advertiser and publisher counts. The vendor's number will not match CM360's for the same reasons the publisher's does not, and the expected gap is of the same order. CM360's own Verification features cover much of the ground without a vendor: Active View measures viewability on impressions the ad server counted, and content and geography verification report where the ads ran. A vendor is added when an independent count is required by the contract or when the vendor measures something CM360 does not.
Trackers should be added at the level where they belong. One placed on the creative fires wherever that creative serves; one placed on the ad fires only for that ad's placements. Placing a campaign-specific tracker on a creative reused across campaigns is a common way to spend a vendor's budget measuring the wrong flight.
One ad server for all buys
The argument for routing every buy through one ad server is reconciliation. Direct insertion orders, programmatic deals, open auction, search clicks and anything else that can carry a tag are counted under a single methodology, with one invalid-traffic treatment and one reporting time zone. Frequency can be managed across them because the ad server sees the same browser whichever route served the impression, and Floodlight attributes a conversion across all of them rather than letting each platform claim it. Creative hosting sits apart from buying, so a creative change is made once and takes effect everywhere, and a change of buying platform does not mean rebuilding every asset.
The cost is a serving fee, charged on a cost per thousand (CPMCost per 1,000 impressions.) basis at the rate on the advertiser's contract, plus the latency of the extra hop and the operational discipline of trafficking. Whether the trade is worth it depends on how many buying routes there are and how much the conversion record matters; the programmatic versus direct buying comparison sets out when a plan has more than one route, and the Campaign Manager 360 hub shows how DV360 and SA360 plug into the same server.
Readers unsure where the ad server ends and the buying platform begins can use the CM360 versus DV360 comparison, which sets out the division of labour and when an advertiser needs one, both or neither.
Common mistakes
- Treating the publisher's count as wrong. It measures a different moment. Compare against the documented range, then investigate by day and by placement.
- Shipping tags with placeholders unfilled. No cache buster means cached impressions; no click macro means the publisher cannot see clicks.
- Choosing a tag type for the publisher's convenience. An image tag on a placement meant for rich media silently drops every creative that needs script.
- Letting the default ad do the work. A default serving at scale means the standard ads are mis-dated, mistargeted or unassigned.
- Weighted rotation set once and forgotten. The weights that made sense in week one are rarely right in week six; optimised rotation or a review date fixes it.
- Trackers on the wrong object. A campaign-specific pixel on a shared creative counts other campaigns' impressions against this one's vendor budget.
- Reading click rate from two systems as one number. CM360's rate divides its own clicks by its own impressions; the publisher's does the same with different counts, and neither is a conversion.
