Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Analytics › Google Analytics 4

Cleaning Up an Inherited GA4 Property

How to clean up a GA4 property someone else built: list its settings, find what relies on each one, fix key events and definitions, and log every change.
Tutorialintermediate10 min readUpdated

What an inherited property hides

The configuration of a property accumulates. Agencies mark the events their reports needed, developers add interface rules to patch a release, a campaign manager builds audiences for a launch and moves on. Few of those decisions are written down, and the GA4 hub describes a model in which every one of them shapes what the reports say. The common GA4 mistakes page diagnoses the symptoms one at a time; an inherited property usually has several at once, and they interact.

Two things make a clean-up risky. The first is that other systems read the property. Key events imported into Google Ads become conversion actions that bidding can optimise towards, as the Google Ads integration tutorial explains; audiences are shared to Google Ads as targeting lists; dashboards in Looker Studio read custom dimensions by name through their data sources. The second is that most changes in GA4 act only from the moment they are made. The guides on this site state the rules that matter here:

ObjectWhat the site's guides say about changing itWho may depend on it
Key eventUnmarking applies going forward and history is not recalculated; the counting method applies from the change (key events guide)Google Ads conversion actions, reports, dashboards
Custom definitionRegistration is not retroactive for the interface; a wrong scope cannot be edited, so the dimension is archived and registered again (events guide, common mistakes)Reports, explorations, dashboard fields
Modify or create event ruleActs from the moment it is saved; a create rule leaves the original event in placeKey events, audiences, reports
AudienceNot retroactive, so a list removed and rebuilt starts empty (audiences guide)Google Ads campaigns, reports
Data filterApplies from activation, and an active filter drops data permanently (setup tutorial)Every report

This is the situation Chesterton's fence describes: nothing is removed until someone has found out why it was put there and who still relies on it. What archiving, deleting or changing each object does to history and limits is documented in the Google Analytics help centre and revised over time, so read the current page before each action rather than relying on memory. The container that sends the events is a separate clean-up, covered by the audit routine in container governance, and Google Ads conversion actions have their own in the conversion action clean-up.

Before you start

  • The handover finished. The business holds administrator rights and owns the linked accounts, as the handover checklist requires. If the property arrived with a whole programme, the tutorial on taking over marketing comes first. If ownership cannot be secured at all, a clean-up is the wrong job: the tutorial on moving to a new GA4 property replaces the property instead.
  • Access to every reader: the linked Google Ads account, the Looker Studio reports and data sources, the BigQuery project if the export runs, or a person in the room who has it.
  • The tracking plan, or whatever stands in for it. The tracking plan guide describes what a complete one holds.
  • The business's own records of orders, bookings or leads for a recent closed period.
  • The measurement change log, open at a new entry, and a spreadsheet for the register with a row per object.
  • A quiet period away from sales, launches and campaign changes.

The steps

  1. Freeze changes and confirm ownership and administrator access

    Tell the agency, the developers and the marketing team that the property's settings are frozen until the clean-up ends, and that any urgent change goes through you. Confirm that the business, not an individual or an agency, owns the account and the linked products, and that you hold administrator rights. Note the property's time zone and currency now: every comparison you make later depends on where a day begins and how revenue is counted. A freeze matters because a change made by someone else halfway through is the one nobody can explain afterwards.

  2. Export the configuration into a register

    List every Data Stream and its tag settings, every key event with its counting method and default value, every custom definition with its scope and parameter name, every modify and create event rule, every audience with its membership duration and where it is shared, every data filter and its state, the internal traffic rules, the unwanted referrals and the cross-domain list, any custom channel groups, the retention setting and every product link. Add columns for what each object is believed to do, its owner, its readers and a proposed action, all blank for now. Resist fixing anything while building the register; it is the single place the clean-up is decided in.

  3. Compare the events actually arriving with the tracking plan

    Read the event names the property received over a recent period and set them beside the plan, as the reports against plan check in the tracking plan guide describes. Names in the data but not in the plan are usually a test event, a misspelling or an automatic event nobody noticed; names in the plan but absent from the data are broken or were never built. Check the enhanced measurement options on the stream as the setup tutorial describes them, because form interactions firing on a search box are a common source of nonsense events. If any parameter carries Personal Data, such as an email address in a page address or a name in a search term, stop it at the source the same day using the guide to keeping personal data out of analytics, and record what was found.

  4. Find what reads each key event, definition and audience in Google Ads, reports and the export

    For every key event, check whether Google Ads imports it and whether the imported action is a primary goal. For every audience, check whether it is shared to Google Ads and which campaigns use it. For every Custom Dimension, search the Looker Studio data sources and shared explorations that use it. The BigQuery export basics guide notes that the export holds every parameter whatever is registered, so saved queries filter by event and parameter names rather than by definitions; list the ones that name events you may change. Then ask the people who read the outputs. Write every reader into the register. A key event that a primary Google Ads goal depends on is the most dangerous object in the property and is handled first.

  5. Decide each key event and move Google Ads goals before unmarking anything

    Apply the three tests in the key events guide: is the event a business outcome, does it fire once per real occurrence, and is the set small. Page views of a pricing page, scrolls and form starts fail the first test and go back to being ordinary events for funnel analysis. Where an event to be unmarked is imported into Google Ads as a primary goal, make its replacement observable first and switch the goal using the conversion goal change checklist, then unmark it in the property. Set the counting method deliberately, once per session for most leads, and write the date down, because neither change restates history. If a create rule and the original event are both marked, mark only one.

  6. Re-register the custom definitions reports need and archive the rest

    Keep a definition when a report, an exploration or a dashboard uses it, and when its scope matches how the site sends the value: user properties for things true of the person, event parameters for things true of the moment, item parameters for things true of a product line, as the scope table in the reports and dimensions guide sets out. A dimension registered at the wrong scope cannot be edited; archive it and register a new one at the right scope, then remap the dashboard fields that read the old name, knowing that the new one starts empty because registration is not retroactive. Archive definitions nobody reads, and leave high-cardinality values such as timestamps and full addresses in the export rather than in the interface. The documented limits on definitions are in the help centre.

  7. Replace interface rules with tag fixes, then remove them

    Modify and create event rules are useful patches and poor foundations: they act from the moment they are saved, they sit where developers and the next analyst will not look, and a create rule leaves the original event in place, which double-counts when both are marked. For each rule, ask what it corrects and make that correction in the tag or the data layer through a normal container release. Run both for a short overlap, confirm in the reports that the tag now produces what the rule used to, then remove the rule and record its removal in the tracking plan with the reason it existed.

  8. Retire unused audiences only after checking where they are shared

    Audiences are not retroactive, so a list removed today and rebuilt next month starts again from nobody. Keep every audience a live campaign uses, and every audience used as a comparison in reports, and write down its purpose and owner. For the rest, confirm in Google Ads that no campaign or exclusion uses them, check the help centre for what removal does, and retire them in a dated batch. Where an audience is worth keeping but badly defined, build the corrected version beside it and let it fill before moving campaigns across, as the audiences guide advises for any new list.

  9. Check filters, referrals, channel groups, retention and links against the setup tutorial

    Walk the setup tutorial's settings against the register. An internal traffic filter left in testing labels traffic without removing it; test the rule against the offices and agencies that actually visit, then activate it on purpose, because an active filter drops data permanently. Add missing payment and booking providers to the unwanted referrals. Keep one or two well-named custom channel groups at most, as the attribution settings guide recommends. Choose the longer retention option unless a policy forbids it. Confirm that each product link points at an account and a project the business owns.

  10. Record every change in the change log and the tracking plan

    Write one dated change log entry for every change that alters a count, a key event or a dimension, and mark those dates on the trend charts that cross them. Update the tracking plan with the key events, definitions and remaining interface rules, each with its purpose and owner, and keep the final register with it. Then book the full implementation audit to score the result. The register is also the documentation the next handover checklist asks for, so the next owner starts where this clean-up ended rather than where it began.

A kayak hire company sorts out its property (illustrative)

Illustrative example — not a real result

Quabbrel Kayak Hire cleans up an inherited GA4 property

Quabbrel Kayak Hire, an invented kayak and paddleboard hire company, takes its property back from an agency after 4 years. The register lists 14 key events, 23 custom definitions, 6 interface rules and 19 audiences. The analyst spends 5 working days on the clean-up.

FindingAction
9 of the 14 key events were page views, scrolls or form startsThe micro-actions were unmarked after the Google Ads goal moved; 3 key events remain: booking, gift voucher purchase and group enquiry
A create rule made booking_complete from the confirmation page view while the tag also sent purchase; both were markedpurchase kept; the rule removed after a fortnight's overlap showed 212 purchases against 214 bookings in the booking system
trip_type was registered at user scope, so each customer showed only their latest tripArchived and registered again at event scope; the 2 dashboard fields that read it were remapped
11 definitions were read by no report, exploration or dashboardArchived in 1 dated batch
12 audiences were used by no campaignRetired after checking Google Ads; the 7 in use were given owners
The internal traffic filter had been in testing for 8 monthsRule corrected for the new office, then activated

The change log gains 8 dated entries, and the next month's audit scores the key event gate as solid.

Common mistakes

  • Deleting what looks dead. A key event, definition or audience may still feed a bid strategy or a dashboard.
  • Unmarking a key event before moving the Google Ads goal. Bidding loses its signal on the day of the change.
  • Editing instead of replacing. A wrong scope cannot be edited; archive and register again, then remap readers.
  • Leaving interface rules as the fix. They hide the problem from the next analyst; fix the tag and remove them.
  • Rebuilding an audience to tidy its definition. The new list starts empty; build beside the old one and wait.
  • Activating a filter without testing it. An over-broad rule removes real visitors for good.
  • Cleaning up without a record. The next owner inherits the same mystery, only tidier.

Frequently asked questions

Can I just delete the key events and custom dimensions that look wrong?

Not as a first move. Something may still read them: a Google Ads conversion action imported from the property, a dashboard field or a colleague's exploration. Find the readers first and move them. Unmarking a key event applies from that day and does not recalculate history, and a dimension registered at the wrong scope cannot be edited, so it is archived and registered again at the right scope with the date recorded. Read the help centre's current description of each action before taking it.

Will cleaning up the property change our historical reports?

Mostly not. Changes to key events, counting methods, definitions, filters and interface rules apply from the moment they are made, so earlier reports keep the old behaviour and anything that reads across the date shows a step. That is why every change goes into the change log with its date and the step is marked on trend charts. Raw event history sits in the BigQuery export only if the export was running at the time.

Should the clean-up come before or after an implementation audit?

After the handover, and around the audit rather than instead of it. A quick pass through the audit's gate rows shows where the property is unsafe to use, the clean-up fixes what it can in a safe order, and a full audit afterwards scores the result. Scoring first and fixing second both work, as long as the fixes follow the order in this tutorial rather than the order the findings were written in.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Google Analytics Help — Google Analytics Help
  2. [GA4] About events — Google Analytics Help
  3. [GA4] Audiences — Google Analytics Help
  4. [GA4] Set up BigQuery Export — Google Analytics Help

Related