Campaign Manager 360 Partners

Floodlights as shared infrastructure, not per-campaign tags

A Floodlight tag is measurement code on your site that records a conversion, and the mistake causing most damage is treating each campaign as a reason to create new ones. Floodlight activities should be modeled on your business rather than your media plan: one for a purchase, one for a qualified lead, one for a key engagement, each defined once and reused by every campaign that needs it. Those activities then feed conversion data to DV360 for optimization and can be imported into SA360 for bidding. A fresh set per campaign fragments the history and leaves buying platforms chasing inconsistent goals.

Ask a prospective partner to describe the Floodlight structure they would propose and how it maps to your existing analytics events. The answer should sound like a schema: a small number of activities, a defined counting method for each, and custom variables carrying detail such as order value, product category or lead type. It should also state which activities are exposed to bidding and which exist purely for reporting. That distinction matters, because an activity wired into automated bidding is no longer just a measurement artefact, and redefining it later moves real money.

Decisions that are painful to reverse

• Counting methodology is effectively permanent — whether an activity counts every conversion, one per session or one per visitor changes every historical number, so altering it later creates a break you keep explaining for years.
• Activity groups determine how conversions roll up — counter groups and sales groups behave differently in reporting and in the data passed to bidding platforms, and reorganising them afterwards is a migration.
• Custom Floodlight variables are a fixed, numbered set — assign them deliberately and document what each number holds, because reusing one for a different purpose silently corrupts the historical series.
• Lookback windows are set per activity — click and view windows shape both reported conversions and what bidding platforms see, and should come from your real purchase cycle rather than the default.
• The deployment method has governance consequences — Floodlights delivered through a tag manager are easier to change but depend on a system your marketing team controls, which is a question about authority.

Naming standards that make reporting possible

Every campaign, placement, creative and ad in CM360 carries a name, and those names are the only dimensions most reporting can group by. Typed freehand, they leave your quarterly report unable to separate channel from format from audience without somebody cleaning strings in a spreadsheet. A naming convention defines a fixed set of fields, in a fixed order, with a fixed delimiter, drawn from controlled vocabularies rather than invented per campaign. It is dull to agree and it is the highest-return hour in an ad server engagement, because retrofitting means renaming live objects.

Convention alone is not enough; somebody has to enforce it. Ask how names are validated before trafficking goes live, whether through a template that builds them from dropdowns, a script that checks structure, or a review step with a named owner. Ask what the convention does about things that change: a new market, a new format, a rebranded product line. A good standard has an extension process. A bad one is a document everyone abandoned in month three, and the version history tells you which you are being offered.

The QA pass that prevents next quarter's argument

• Click destinations tested live — every landing page opened and confirmed, with tracking parameters surviving each redirect, because a broken click URL spends budget and records nothing.
• Creative rendering checked per placement size — including fallback behavior when a rich media unit fails, which is where most silent under-delivery hides.
• Floodlight firing confirmed by a real conversion — a test transaction taken all the way through and verified in the reporting interface, not assumed from a tag debugger.
• Placement dates and caps reconciled to the media plan — mismatched flight dates are the most common cause of a campaign that appears to underspend in its first week.
• Duplicate tags hunted deliberately — the same Floodlight deployed twice through different routes doubles your reported conversions, and it is easier to find before launch than to explain afterwards.

Running both ad servers in parallel

Moving from one ad server to another is not a cutover you can validate afterwards. The only reliable method is a period where both systems run live on the same inventory simultaneously, counting the same impressions and clicks independently, so you can compare them directly. Four to six weeks is typical and it costs money, because you pay two serving fees and traffic everything twice. That cost buys the ability to explain next year's numbers. A clean switch at the start of a quarter optimizes for the implementer's effort, and the bill arrives later as an unresolvable argument.

During the parallel period, compare at the level you will eventually report at: by placement, by day, by creative. Aggregate totals can agree while individual placements are badly wrong in offsetting directions. Pay attention to click counts, which differ between ad servers because of redirect handling and bot filtering, and to conversions, where lookback windows and counting methods must be matched deliberately rather than assumed equivalent. Document each divergence with its cause as you find it; that document answers the year-on-year question later.

Agreeing the discrepancy threshold before you see the data

Agree the acceptable discrepancy before the data exists, in writing, with a named person empowered to declare the migration complete. Once numbers are on screen every threshold becomes negotiable and the conversation turns political. Reasonable starting points are a low single-digit percentage for impressions, a wider band for clicks, and a tighter one for conversions where counting methods have been matched. The specific figures matter less than having set them in advance, because the purpose of a threshold is to remove the temptation to rationalise whatever gap you observe.

Also agree what happens if the threshold is missed. The options are to extend the parallel period, to accept the gap with a documented explanation, or to stop and fix the cause, and each carries a cost somebody must approve. Include a decision about historical data: the old ad server's reporting becomes inaccessible at some point after its contract ends, so exporting the full historical detail rather than a summary should be a line item with a deadline attached.

Viewability, invalid traffic and geography as configured choices

CM360 can apply verification settings measuring whether an impression was viewable, whether it came from invalid traffic, and whether it served in the geography you intended. These arrive with defaults, and defaults are decisions somebody else made for a generic advertiser. Viewability has a standard definition that may not match what your business considers a seen advertisement, and video carries a different standard again. Whether you block on verification failure before serving or report on it afterwards is a genuine trade-off: blocking protects the buy and reduces deliverable inventory, reporting preserves scale.

Ask which settings are active, who set them and when they were last reviewed. Ask specifically about invalid traffic filtering, since general invalid traffic is filtered automatically while sophisticated invalid traffic detection usually needs a third-party vendor and a separate cost line. Geographic settings deserve attention too, because location is inferred and the inference is imperfect, so a strict restriction can quietly suppress delivery on an otherwise healthy campaign. An operator who cannot date the last review is telling you it has not happened.

Who owns the explanation when platforms disagree

Your CM360 click count will not match the clicks DV360 reports, neither will match the sessions GA4 records, and none will match what a search platform shows. These are not bugs. The systems count different events, at different points in the delivery chain, with different filtering, lookback windows and attribution. The failure is not the discrepancy; it is that nobody in the organization can explain it on request, so each quarter the same investigation is launched and abandoned. You are hiring for a named owner of that explanation and a document that survives staff turnover.

Make it a deliverable: a reconciliation note listing each pair of systems, the expected direction and rough magnitude of the gap, and the structural reason behind it, refreshed whenever a platform changes. Then decide which system is the source of truth for which decision — the ad server for delivery and pacing, your own systems for revenue, analytics for on-site behavior, the buying platforms for in-flight optimization. Written down, this ends most arguments before they begin. Left unwritten, it recurs every reporting cycle.

Programs rarely stop at one product. Buyers hiring for Campaign Manager 360 often pair it with BigQuery partners , Display & Video 360 partners or Google Analytics 4 partners , or review the whole Google Marketing Platform landscape before committing.