What Floodlight measures
A Floodlight configuration is created per advertiser in CM360 and is identified in every tag by its
configuration identifier, the src parameter. Inside the configuration sit activity groups, and inside
each group the activities. An activity is the thing counted: a confirmation page reached, a form submitted,
a button pressed, an order completed. When the activity's tag fires in a browser, CM360 records a Floodlight
event with the activity's identifiers, the time, any custom variables the tag carried and, for sales
activities, quantity and revenue.
The ad server then does what no site analytics tool can: it looks for impressions and clicks it served to that browser, across every placement and every linked buying platform, within the activity's conversion windows, and attributes the conversion to them. That is the AttributionAssigning conversion credit to touchpoints. step, and it is why Floodlight numbers are expressed as click-through conversions and view-through conversionsA conversion credited to an ad that was shown but not clicked, within a set view window. rather than as a single count. The same activities feed audience lists, which the DV360 audiences guide covers, and they are read by linked DV360 and SA360 advertisers so that one tag serves display, video and search.
Floodlight is a browser-side mechanism. Google documents an offline conversions upload through the Campaign Manager 360 API for actions that happen away from the browser, which plays a role similar to a server-side channel such as Meta's Conversions APIServer-side event sharing with Meta.; it is a complement, not a replacement, and needs the identifiers from the original impression or click to attribute anything.
Activity groups and activities
Groups and activities are the two levels of the Floodlight hierarchy, and each has a documented job.
| Object | What it is | Settings that matter | Identifier in the tag |
|---|---|---|---|
| Floodlight configuration | One per advertiser; the container for everything below and the owner of custom variable definitions | Custom variable definitions, natural search tracking, consent and privacy settings | src |
| Activity group | A set of related activities of one type: counter or sales | Type, which cannot change after creation; name | Group tag string, the type parameter |
| Activity | A single trackable action | Tag format, counting method, conversion windows, expected URL, default tags, dynamic tags | Activity tag string, the cat parameter |
| Custom Floodlight variable | A named field defined at configuration level and filled by the tag | Name, type (string or number), reporting availability | u1, u2 and so on up to the documented limit |
Design the groups around questions reporting will ask. A group named for a funnel stage, such as "Leads" or "Purchases", lets a Floodlight report sum sensibly at group level. Activities inside it should be one action each, named so that a reader of a report six months later knows what fired without opening the configuration. The group tag string and activity tag string are set once and appear in every tag, in every Tag Manager template and in every report, so a naming convention agreed before creation saves a great deal of later confusion. Changing them after tags are live breaks every deployed tag at once.
An activity also carries an expected URL and optional default tags, which are publisher or vendor pixels that CM360 fires alongside the activity, and dynamic tags, which fire the pixel only when a given ad or placement is in the path. Both are documented ways to feed a third party without adding its tag to the site.
Counter vs sales tags and counting methods
The group type decides what an activity can record. A counter activity counts events. A sales activity counts events and records a quantity and a revenue value with each, which is what makes revenue and return columns possible in reports and what a sales-based bid strategy in a linked platform optimises towards.
Within each type, CM360 documents counting methods that decide how repeat firings are treated. For counter
activities: standard counts every firing; unique counts one firing per browser per day; per
session counts one per session as the ad server defines it. For sales activities: transactions counts
every firing and sums the quantity and revenue it carries; items sold counts the quantity rather than
the number of firings. The method is chosen on the activity and determines what the ord parameter must
hold. A standard counter tag uses ord as a random cache buster. A unique counter tag fixes ord and
carries the random value in num. A sales tag uses ord for the order or transaction identifier, which is
how CM360 deduplicates repeat loads of a confirmation page, and carries the value in cost and the count in
qty.
The choice follows the business question. A lead form is a standard or unique counter depending on whether
a person submitting twice is two leads or one. A purchase is a sales activity with transactions counting
and the order identifier in ord, so that a refreshed thank-you page cannot become a second order. A
newsletter sign-up is a counter, not a sale with a zero value, because a sales group pulls it into revenue
columns where it does not belong. Cost per acquisition (CPACost of each conversion.) in reports depends on these
choices being right, since the denominator is the activity count under its counting method. The academy exercise on
untangling inherited Floodlight activities for a furniture rental firm
practises deciding what to do with a rental order recorded in a counter group with no revenue.
Deploying through Tag Manager
Google documents Floodlight tags in three formats: the Google tag with an event snippet, an iframe tag and an image tag. The Google tag format is the one Tag Manager's templates produce and the one that supports the current consent and conversion-linking features; the iframe and image forms exist for pages and environments that cannot run script. Tag Manager (GTMTag management system.) offers two built-in tag types, Floodlight Counter and Floodlight Sales, each taking the advertiser identifier, group tag string, activity tag string, counting method and custom variables as fields. The tags, triggers and variables guide explains the container mechanics these templates rely on.
The firing sequence is what the diagram below animates. A trigger in the container fires when the page or the data layer says the action happened. The Floodlight tag builds its request from its fields and from variables that read the data layer, such as the order identifier, value, quantity and any custom variable values. The browser sends the request to CM360, where the activity is identified by its tag strings, the counting method decides whether it is a new conversion, and the ad server matches the browser against served impressions and clicks within the windows. A Conversion Linker tag, deployed on every page, stores the click identifiers that arrive on landing pages so that later activity firings can be matched in browsers that restrict cookies set by the ad server directly.
Tag firing flow
Deploy the base and linker on all pages, one Floodlight tag per activity on its own trigger, values read from data layer variables rather than scraped from page text, and every tag gated behind the consent trigger the site's consent platform raises. Google documents that Floodlight tags deployed through the Google tag respect consent mode, and it documents the privacy parameters the tag can carry; follow the help centre's current description of what fires in each consent state rather than a remembered one. The Google Ads conversion tracking tutorial follows the same container pattern for a different tag, and the two are usually deployed together.
Custom variables and values
Custom Floodlight variables are defined once at configuration level with a friendly name and a type, and
then filled by any activity's tag through the u parameters. They carry detail that the activity itself
cannot: product category, lead type, customer status, plan tier, store identifier, a campaign label passed
from the landing page. In reports they appear as dimensions or, for number-typed variables, as metrics that
can be summed, and they can be used to build audience lists. Google documents the limit on how many a
configuration may define; the point is to spend them on fields that reporting or audiences will actually
use.
Values travel with the tag as string or number, and the type set at definition decides how CM360 treats
them. A variable meant to be summed must be defined as a number and sent without currency symbols or
thousands separators. Personal data must not be sent in any variable; Google's policies on this are
explicit and apply to Floodlight as to every other tag. Revenue and quantity do not go in custom variables
at all. They belong in cost and qty on a sales tag, where the reporting columns expect them.
A disciplined pattern is to document each variable alongside the activity list: index, name, type, source
in the data layer, allowed values, and which reports use it. Without that document, the next trafficker
finds a variable called u7 holding something nobody can name and either ignores it or reuses it for a new
purpose, breaking the history. Where that has already happened in a configuration someone else built, the
tutorial on cleaning up an inherited Floodlight configuration traces each activity and variable before anything is changed.
Attribution and conversion windows in the ad server
Each activity carries a click-through window and a view-through window. CM360 attributes a conversion to a click when the click happened within the click-through window before the activity fired, and to an impression when the impression happened within the view-through window and no eligible click exists. The help centre documents the default lengths and the range each window can be set to; set them deliberately per activity rather than accepting the default, because a long view-through window on a cheap display placement is the single biggest source of inflated conversion counts in a CM360 account.
Attribution models are applied in reporting, not in the activity. The default model credits the last interaction, and CM360 documents alternative models, including first interaction, linear, time decay, position-based and data-driven, plus custom models that can be built in the attribution tools and applied to path reports. The model changes which placement or channel receives the credit; it does not change whether the conversion was counted. The theory behind those models and when each is defensible belongs to analytics rather than to the ad server; the CM360 hub describes where the reports that apply them live.
Cross-platform attribution also depends on the links between products. A DV360 advertiser linked to the CM360 advertiser attributes its own line items against the same activities; an SA360 advertiser does the same for search. Each product reports under its own rules, so the same activity can show different attributed counts in three interfaces while the raw activity count in CM360 stays the one source.
Designing Floodlight for a subscription business
A meal-kit company sells three plan tiers and runs display through DV360, paid search through SA360 and a
small direct buy with a recipe site. The measurement plan produces two activity groups. A counter group
"Leads" holds two activities: "Quiz completed", standard counting, and "Account created", unique counting
so that a person who registers twice in a day counts once. A sales group "Subscriptions" holds one activity,
"First box ordered", transactions counting, with ord set to the order number, cost to the first-box
value and qty to one. Custom variables carry plan tier (u1, string), promo code (u2, string) and
delivery postcode area (u3, string). Windows are set to 30 days click-through and 7 days view-through for
the sale, and 30 days click-through with 1 day view-through for the leads. In the first month the sales
activity records 1,842 first boxes with revenue of £112,360; the DV360 interface attributes 410 of them to
its line items and SA360 attributes 1,120, while CM360's path report shows 260 orders where a display
impression preceded the search click.
Testing and troubleshooting
Test before any campaign depends on the numbers. In Tag Manager, preview mode shows whether the Floodlight tag fired on the right trigger and with which values. Google's Tag Assistant shows the request leaving the browser with its parameters, so the tag strings, counting method parameters and custom variables can be read before CM360 sees them. In CM360, the Floodlight configuration lists each activity's status and when it last received a tag, and a Floodlight report over the test period shows whether the firing became a counted conversion. Walk the whole journey as a real person would, with consent granted and then with it refused, and check both branches. When an activity that used to count reads zero, the diagnosis of Floodlight not recording conversions decides first whether the activity's own firings or one product's attributed share is missing before tracing the tag. When the count reads too high rather than too low, the diagnosis of Floodlight conversions counted twice separates repeated firings, duplicate activities, a tag fired twice, long windows, shared credit and a doubled bid goal.
| Symptom | Check first | Likely cause |
|---|---|---|
| Activity never shows a firing | Tag Assistant on the live page | Tag not published, wrong trigger, or consent blocking it |
| Firings arrive but no conversions are attributed | Floodlight report split by click-through and view-through | No served media reached that browser inside the windows, or the linker is missing |
| Orders counted more than once | The ord value in the request | Random value on a sales tag, or the tag firing on every load of the confirmation page |
| Revenue absent or wrong | The cost and qty values | Currency symbols, separators, or a counter tag used for a sale |
| Custom variable empty in reports | The data layer variable in preview mode | Value not pushed before the tag fires, or the index mismatched to the definition |
| Counts vanish after a site release | Preview mode on the changed template | Data layer renamed or the confirmation page URL changed |
For a DV360 programme judged on orders, the DV360 setup for online sales builds on a sales activity with transactions counting and reads post-view orders apart from post-click ones.
Common mistakes
- Creating activities before the measurement plan. Groups cannot change type later, and renaming tag strings breaks live tags.
- One activity for everything. A single "conversion" activity makes the lead and the purchase indistinguishable in every report and every bid strategy.
- Sales tags with a random
ord. Every refresh of the thank-you page becomes another order. - Default windows accepted without thought. The view-through window decides how much display gets credit; it should be a choice.
- Values scraped from page text. They survive until the next design change and then silently send nothing.
- Personal data in custom variables. Prohibited by policy and a liability the moment it is noticed.
- Testing only in the happy path. The consent-refused branch and the repeat visit are where tags misbehave.
- Trusting one interface's attributed count as the total. The CM360 activity count is the source; attributed counts in DV360 or SA360 are each product's share of it.
