Why an inherited Floodlight configuration needs understanding before cleaning
The Campaign Manager 360 hub places Floodlight at the advertiser rather than in any campaign,
which is why its mistakes outlive every campaign and agency that made them. A Floodlight configuration is shared infrastructure. The Floodlight guide describes one configuration per advertiser,
identified in every tag by its src value, with activity groups and activities beneath it, and it notes that linked
DV360 and SA360 advertisers read the same activities so that one tag serves display, video and search. The
DV360 audiences guide adds that audience lists are built from them. An activity
that looks dead in the configuration may still be the goal of an automated strategy described in the
DV360 bidding and pacing guide, the conversion goal of a portfolio in the
SA360 bid strategies guide, or a column in a report somebody reads every
Monday.
That is the situation Chesterton's fence describes: something should not be removed until someone has found out why it was put there and who still relies on it. Two Floodlight rules make the caution sharper. The Floodlight guide warns that a group's type is fixed at creation and that changing group or activity tag strings after tags are live breaks every deployed tag at once. So the cleanup works by understanding first, then building correct objects beside the wrong ones, moving dependents across, and only then retiring what is left.
The Google Ads conversion action cleanup tackles a similar mess in a different object, a Google Ads Conversion ActionA Google Ads definition of one action to count, such as a purchase or lead, with its source, category, counting method, value and conversion window., with its own rules about primary and secondary actions; if the business has both, clean them separately and record both in the same change log. If the reason for opening the configuration is an activity that has stopped counting, the diagnosis of Floodlight not recording conversions separates a silent tag from a missing link or an unexplained inherited group before this cleanup begins.
Before you start
- Access to the CM360 advertiser with permission to view and edit the Floodlight configuration.
- Access to every linked DV360 advertiser and SA360 account, or a person in the room who has it.
- Access to the Tag Manager container that deploys the tags, or to whoever controls the site's tags.
- The measurement plan, or the business's own statement of which outcomes count and which carry revenue.
- The business's own records of orders, bookings or leads for a recent closed period, to check counts against.
- The measurement change log, open at a new entry.
- A spreadsheet for the register, with a row per activity and per custom variable.
- A quiet period in the trading calendar, away from sales and launches.
The steps
Export the configuration into a register
List every activity group and activity with its group type, tag strings, counting method, click-through and view-through windows, expected URL, default and dynamic tags, and status, and every custom variable with its index, name and type. Add columns for what the activity is believed to count, its owner, its dependents and a proposed action, all blank for now. The register becomes the single place the cleanup is decided in, so resist fixing anything while building it; a change made from memory halfway through is the hardest one to explain later.
Find when each activity last received a tag
The Floodlight guide notes that the configuration lists each activity's status and when it last received a tag. Record that date for every activity, then run a Floodlight report over a recent period from the CM360 reporting guide to see the counts behind it. Sort the register into activities that fire normally, ones that fire rarely and ones that have not fired for a long time. Silence does not prove an activity is unwanted: the page may have moved, the tag may be broken, or the action may simply be seasonal. Mark each silent activity as a question to answer, not yet as a candidate for retirement.
Find what depends on each activity in DV360, SA360 and audience lists
For every activity, look for its readers. In DV360, check automated bid strategies and goals on insertion orders and line items, and the Floodlight-based audience lists. In SA360, check conversion columns and the goals of bid strategies, which the SA360 Floodlight and reporting guide describes as built on the same activities. Then check scheduled reports, dashboards and data exports that filter on activity names, and ask the people who read them. Write every dependent into the register. An activity with a live bid strategy depending on it is the most dangerous object in the configuration and is handled before anything else. The academy exercise on untangling inherited Floodlight activities for a furniture rental firm practises tracing what relies on each activity before deciding to keep, fix, replace or retire it.
Check each counting method and window against the business question
For each activity that will stay, ask what the business wants to count and compare it with what the settings do. The Floodlight guide sets out the counter methods (standard, unique and per session) and the sales methods (transactions and items sold), and explains that a sales tag uses
ordfor the order identifier so that DeduplicationRemoving double counts when one conversion is reported by more than one source. stops a refreshed confirmation page becoming a second order. Look for a lead counted every time a form reloads, a purchase counted in a counter group with no revenue, two activities counting the same action, and windows nobody chose. Long view-through windowsA conversion credited to an ad that was shown but not clicked, within a set view window. deserve a deliberate decision, because the Floodlight guide names them as the biggest source of inflated counts. Compare the counts with the business's own records for the same period. Where the counts run above the records and the register shows no duplicate activity to blame, the diagnosis of Floodlight conversions counted twice reads the excess as repeated firings, a tag fired twice, long windows or shared credit.Compare the tags on the site with the register
Open the Tag Manager container and match each Floodlight tag to a register row by its tag strings, noting its trigger, its counting fields and the Data LayerA structured object on a web page that passes values to a tag manager for tags to use. variables it reads, as the tags, triggers and variables guide explains them. Then walk the real journeys in preview mode, which shows which tags fired on each event and with which values, and use Tag Assistant to read the request leaving the browser. Look for tags that point at activities the register calls retired, activities with no tag at all, tags hard-coded on pages outside the container, and values scraped from page text. Test with consent granted and refused, as the Floodlight guide advises.
Replace misplaced activities without renaming live tag strings
Where an activity is in the wrong group type or counts the wrong thing, create a correct activity in a group of the right type and leave the old one firing. Deploy the new tag beside the old one, check that both record the action, and compare the counts with the business's records. Then move each dependent from the register across, one at a time: audience lists, SA360 columns, and last of all bid strategy goals, using the conversion goal change checklist, which adds the new event as observed before it carries bidding and avoids changing several things in the same week. Never change the tag strings of a live activity to make it look tidy; that breaks every deployed copy at once.
Fix custom variables and remove any personal data
For each custom variable, write down its name, type, source in the data layer, allowed values and the reports that use it, as the Floodlight guide recommends. A variable that nobody can name should be traced through the tags before anything is decided, because reusing it for a new purpose breaks its history. Check what the variables actually carry in preview mode. If any carries Personal DataAny information relating to an identified or identifiable living person, such as a name, email address, location data or an online identifier., such as an email address, a phone number or a name, stop sending it at once in the container: the Floodlight guide notes that Google's policies prohibit personal data in Floodlight as in every other tag. Record what was found and when it stopped, and tell whoever is responsible for data protection in the business.
Retire activities in stages with a dated record
Retire only what has no dependents left. First remove the tag from the container and watch the activity fall silent while every report and strategy that used to read it carries on unaffected. Only then change the activity's status in the configuration. The Campaign Manager 360 Help Centre is the reference for the options available and what each does to reporting history; check it before acting rather than assuming, and prefer the option that keeps history readable. Retire in batches rather than all at once, dating each batch in the register, so that if something breaks the cause is a short list of recent changes.
Record the clean configuration in the change log
Write the result where the next person will find it: the register in its final state, each activity's purpose and owner, the variable definitions, the retirements with their dates, and the dependents that were moved. Add an entry to the measurement change log for each change that alters a count or a goal, and mark those dates as breaks in trend reports. The programmatic advertising audit scores Floodlight as a gating criterion; a register kept up to date is what lets the next audit score it as solid without repeating this work.
A language school untangles its activities (illustrative)
Mirquoth Language School cleans up an inherited Floodlight configuration
Mirquoth Language School, an invented school selling online and in-person courses, changes agency and inherits a configuration with 46 activities in 7 groups, 11 of them named only "Activity" with a number. The new analyst spends four working days on the cleanup.
| Finding | Action |
|---|---|
| 15 activities had received no tag for more than six months | 12 had no dependents and were retired in two batches a fortnight apart; 3 were seasonal summer-course enquiries and kept |
| 2 activities both counted the same enrolment form, one standard and one unique | The unique one kept; the standard one's tag removed after its 2 audience lists were rebuilt on the unique activity |
| Course bookings sat in a counter group, so reports showed no revenue | A new sales activity "Course booked" created with transactions counting and the booking number in ord; ran beside the old one for 3 weeks, counting 412 bookings against 418 in the school's records |
| An old "Brochure download" activity was the goal of an SA360 bid strategy | The strategy's goal moved to the new booking activity first, using the conversion goal change checklist, before the old activity was touched |
| Custom variable u4 carried the enquirer's email address | Removed from the container the same day; data protection lead told |
| View-through window on all activities left at the default | Set deliberately per activity after discussion with the marketing manager |
After the cleanup the configuration holds 22 activities, each with a named purpose and owner in the register, and the change log records 9 dated entries. The first monthly report after the change shows booking revenue in CM360 for the first time.
Common mistakes
- Deleting activities that look dead. Silence may be a broken tag or a season, and something may still read it.
- Renaming tag strings to tidy up. Every deployed tag that uses them breaks at once.
- Moving a bid strategy's goal in the same week as other changes. The learning that follows cannot be attributed.
- Checking only CM360 for dependents. DV360 strategies, SA360 columns, audience lists and dashboards read the same activities.
- Retyping instead of replacing. A group's type is fixed; create a correct activity beside the wrong one.
- Reusing an unexplained custom variable. Its history becomes meaningless; trace it first.
- Leaving personal data flowing while the cleanup is planned. Stop it the day it is found.
- Cleaning up without a record. The next person inherits the same mystery, only tidier.
