Why data operations is an operations job
Most marketing data problems are not discovered by the people who cause them. A specialist names a campaign in the ad platform the way that makes sense to them; the analyst finds out a month later that it does not match the name in the analytics property or the CRM. A list is imported for an event without de-duplication; the email team finds out when contacts complain about receiving the same message twice. The cost lands on someone other than the person who made the choice, which is exactly the situation in which an individual habit will never fix the problem.
That makes data consistency a process problem with a system attached, and process is what operations owns. The analytics team designs the tracking and diagnoses data-quality faults; the marketing data quality guide is where that diagnosis lives. The automation team designs the CRM structure and lifecycle; the CRM fundamentals guide covers that. Operations sits between them and the people who create data every day, and owns three things: the shared taxonomy, the routines that enforce it at the point of creation, and the cadence of checks that catches what gets through. Without that middle layer, the analysts clean the same faults every month and the cause never changes.
A campaign taxonomy across platforms
A campaign taxonomy is the agreed list of attributes that describe a campaign, the allowed values for each, and how those values are written in every system that holds campaign data. It is the shared language that lets a report join spend from the ad platforms, sessions from the analytics property (GA4Google's event-based analytics platform.) and pipeline from the CRM without manual matching.
| Attribute | Example allowed values | Where it appears |
|---|---|---|
| Objective | awareness, consideration, leads, sales, retention | Campaign name in every platform; tracking campaign parameter |
| Product or line | Controlled list maintained by operations | Campaign name; CRM campaign field |
| Market | Country or region codes from a fixed list | Campaign name; tracking parameter; CRM field |
| Audience | prospecting, remarketing, customers, partners | Ad set or audience naming; tracking content parameter |
| Channel and source | Fixed list aligned with analytics channel groupings | Tracking source and medium parameters |
| Launch period | Year and month in a fixed format | Campaign name |
| Creative or variant | Short identifier from the brief | Ad name; tracking content parameter |
The detailed rules for tracking parameters, including case, separators and which parameter carries which attribute, belong to the analytics root, and the UTM and naming conventions guide sets them out. The taxonomy above is the operational layer above that convention: it decides which attributes exist and what their values are, so that the UTMURL tags identifying traffic source. convention, the ad platform naming and the CRM fields all express the same thing. Keep the list of attributes short. Each attribute added is one more value to get right on every campaign, and attributes that no report ever groups by should be removed.
The taxonomy needs a home: one document with the attributes, their allowed values, the format in each platform, an owner and a change log. New values, such as a new product line or market, are added by the owner on request, not invented by whoever launches first. Where finance books campaign spend to its own codes, the tutorial on mapping marketing spend to finance codes shows how to make each finance code point at the same campaign the taxonomy describes, so cost and results can be joined.
Naming governance: who enforces it and how
A naming convention that relies on people remembering it is followed most of the time, and "most of the time" is not enough for reporting. Governance is the set of mechanisms that make the right name the easy name and the wrong name hard to create. They work best in layers, from the point of creation outwards.
Build names, do not type them. The intake form or a naming builder assembles the campaign name and tracking parameters from dropdowns of allowed values. The specialist copies the result into the platform. This removes most variation at the source and is the single most effective control.
Carry the name through the work system. The name is created once at intake and stored on the work item, so that every later stage uses the same string. The work management guide shows how intake forms and templates hold it.
Check at launch. The launch check compares the live names and parameters with the work item. This catches the cases where the specialist edited the name in the platform.
Audit after launch. A scheduled check lists every new campaign name in each platform and flags those that do not parse against the taxonomy. Faults go back to the campaign owner with a fix-by date, and repeated faults by the same team point to a gap in training or in the builder.
Enforcement needs an owner with the authority to ask for a fix, and the backing of the marketing lead when the answer is "we are too busy". Tag container changes (GTMTag management system.) that affect how names and parameters are captured should follow the container's own change rules rather than being made ad hoc.
CRM and audience hygiene routines
Contact and audience data degrade faster than campaign names, because they are fed by many sources: forms, event lists, imports, integrations and sales activity. Hygiene is a set of small routines run on a cadence, not a one-off clean-up. A big clean-up without routines simply resets the clock. The import check below is written out step by step in the tutorial on importing contact lists into the CRM.
| Routine | What it does | Cadence | Owner |
|---|---|---|---|
| Import check | Validates format, source, consent basis and duplicates before a list enters the CRM | Every import | Operations or CRM administrator |
| Duplicate review | Merges duplicate contacts and companies by agreed rules | Regular, by volume | CRM administrator |
| Suppression sync | Confirms that unsubscribes and objections are applied in every sending system | Continuous where integrated; checked on a cadence | Operations |
| Audience refresh | Reviews saved audiences and lists for stale criteria or expired sources | Before major campaigns and on a cadence | Channel leads |
| Field completeness | Reports on missing values in fields used for segmentation and reporting | Monthly | Operations |
| Archive and deletion | Removes or archives data past its retention period | On the retention schedule | Data owner with legal input |
Suppression deserves special attention because it is where data hygiene meets legal obligation. A person who has objected to direct marketing must not be contacted again through any channel the objection covers, and if suppression lists live separately in the email tool, the ad platform's customer lists and a customer data platform (CDPUnified first-party customer profiles.), they will drift apart. Operations should own the check that they match. The legal position is a matter for your advisers and the regulators' guidance, and the approvals and compliance guide covers how consent checks fit into review. When an export is headed for an AI tool rather than a sending system, the tutorial on minimising and redacting data before prompting applies the same discipline field by field. The archive and deletion routine needs a schedule to follow, and the guide to data retention across marketing tools sets one per purpose, keeping suppression entries while the rest of a person's data is removed from every copy.
Ownership, cadence and audits
Data operations works when every data domain has a named owner and a scheduled routine. Without both, the routines above happen once, after a visible failure, and then lapse. A simple ownership table, kept with the taxonomy, names the owner of each domain: campaign names and taxonomy, contact data, audience lists, suppression, tracking parameters, and each platform's account structure. Owners are roles, not people.
The cadence turns ownership into practice. Continuous controls run at the point of entry: naming builders, import checks, form validation. Periodic checks run on a calendar: a monthly naming audit, a regular duplicate review, a quarterly access review of who can change what in each platform. Each check produces a short record: what was checked, what was found, what was fixed and what is outstanding. That record is what lets a new operations lead understand the state of the data without starting from scratch.
Audits should also look for drift in the taxonomy itself: values that are never used, attributes that no report groups by, and new informal values that teams have started using because the official list lacked something. The fix for the last is usually to add the value properly, not to ban it. The CRM's fields drift in the same way, and the tutorial on auditing and documenting CRM fields applies the audit to them, ending in a field dictionary.
Hand-off to analytics and BI
Operations prepares the data; analytics turns it into answers. The hand-off works best when it is explicit. The analytics team should know the taxonomy, have read access to its change log, and be told before new attributes or values appear. In return, operations should know which fields reporting depends on, so that those fields are protected by the routines above.
The measurement plan framework is the natural meeting point. It lists the metrics the business tracks, their definitions and their owners. Each KPIA metric tied directly to an objective. in the plan depends on certain data being consistent, and the operations routines should be prioritised by that dependency: if pipeline by campaign is a headline measure, the CRM campaign field and the campaign taxonomy come first. When a report breaks, the analytics team diagnoses the fault, and if the cause is a routine that failed or a gap in the taxonomy, operations fixes the routine.
On the analytics side, the taxonomy feeds the campaign dimension and campaign mapping table described in the guide to designing a marketing data model, which is why a value that stops parsing shows up there first.
A software reseller rebuilds its taxonomy
Latticeperch Systems, a software reseller selling in six countries, found that its quarterly report needed three days of manual matching because campaigns were named in at least four styles. An audit of one quarter's campaigns found 212 distinct campaign names that mapped to 61 real campaigns. Operations defined a taxonomy with six attributes, built a naming builder into the intake form, added a name comparison to the launch check and scheduled a monthly naming audit. It also found that suppression lists in the email tool and the ad platforms had drifted by several hundred contacts and set up a weekly sync check. Two quarters later, the monthly audit was flagging fewer than five names per month, and the quarterly report took half a day.
Taxonomies hold only when people follow them day to day, and the UTM link request SOP shows how link requests are routed, checked and logged against the agreed naming.
When the drift has already reached the reports, the diagnosis of inconsistent campaign names and UTM tags counts the variants and traces each pattern to the routine that failed.
Common mistakes
- Treating data quality as an analytics problem only. Analysts can find faults; only operations routines stop them recurring.
- A taxonomy with too many attributes. Every attribute is a chance to get the name wrong; keep only those that reporting groups by.
- Typing names by hand. Free text guarantees variation. Build names from controlled values.
- One big clean-up with no routine. The data degrades back to where it was within months.
- Suppression lists kept separately in each tool. They drift apart, and the consequences are legal as well as reputational.
- No owner per data domain. Routines lapse after the first visible failure has been forgotten.
- Changing the taxonomy without telling analytics. Reports break silently when new values appear.
- Banning informal values without asking why they appeared. They usually signal a gap in the official list. The systems hub shows how data operations fits with the other operational systems.
