When You Inherit a Mess
Years of links tagged with no list, and now they're yours. The order of work: catalog every value, map it to a target taxonomy, stop new mess at the source, then read the history through the map.
On this page
You’ve joined a team where several people tagged links for years with no list. Maybe your predecessor left no notes. You open analytics and find dozens of sources, most of them spellings of the same few platforms, and campaign names in every shape anyone ever typed.
Nothing here is new. Each step leans on a chapter that teaches the idea; this is the order to do them in.
1. Catalog before you fix anything
Export every distinct source, medium and campaign, unfiltered, for as far back as the reports you still read. For each value, note three things:
- what it probably means:
li,linkedinandLinkedInall mean LinkedIn - how much traffic it carries, so the big values get mapped first
- whether it’s still being created, or only history
The values still being created come first. Somebody makes those links today, and every day adds more rows.
2. Write the target taxonomy, and map every old value to it
Start from the values and the campaign pattern in Chapter 3, and choose a structure with Chapter 4. Then give every old value its new one:
| Slot | Values in use | Approved value |
|---|---|---|
| source | linked-in, LinkedIn, li, linkedin | linkedin |
| source | Instagram, ig, instagram | instagram |
| medium | e-mail, Email, email | email |
The mapping is the document you’ll use for months. Keep it, and add to it whenever an audit finds a new spelling. The Cheat Sheet’s mapping table template adds a column for how the old data reads, for the values that can’t simply be remapped.
A value maps only if it meant one thing. When one spelling held two meanings, like the mailchimp source in Chapter 1 that was written on newsletters and promotions alike, no row in the map can be right for both. Leave it as it is, and label it.
3. Stop new mess at the source
Fix where links get made before you touch the history, or the map grows every week. Each place links get made outside the rules adds rows to the map: an agency’s spreadsheet, a contractor typing into an ad platform, a developer coding a homepage banner. Chapter 9 maps what each tool refuses, from a shared sheet to a platform with approvals.
4. Read the history through the map
Recorded values never change. The map reads them back in your reports instead: Chapter 1’s remap, applied to every old spelling. Where it runs depends on where you read your numbers:
- in your analytics tool, with its own grouping (in GA4, custom channel groups for channels)
- in a dashboard, as a lookup table between the raw value and the one you report
- in a data warehouse, as a model every report reads; For Data Teams has its mapping table
5. Run the change without breaking the history
Campaigns are running while you do this. Which ones keep their old values until they end, when new campaigns switch, and how long old and new run side by side is Chapter 10’s subject, along with who owns the list from now on.
6. Keep looking
Run Chapter 8’s audit after every change and on a schedule between: every distinct value, unfiltered, held against the list and against the links you made. Its two numbers, unapproved values per slot and the share of campaign names that match the pattern, tell you whether the cleanup is holding.
You won’t get perfect continuity across the cutover (Chapter 10). Set it, get clean data flowing from it, and read everything before it through the map.