CRM Hygiene Is A Marketing Job

A business cannot report on what it cannot query, and most businesses cannot query their own pipeline reliably.

The usual symptom is an argument. Marketing says it generated two hundred leads, sales says it received eighty, and both are reading the same system. The gap is almost always data quality rather than dishonesty.

Where the data goes wrong

Duplicates. The same person entered three times from three sources, with the activity split across all three. Every count is wrong and the one with the real history is not the one anybody is working.

Missing source. A lead with no record of where it came from cannot be attributed, which means the channel that produced it gets no credit and the budget moves away from it.

Stages that do not mean anything. Eleven pipeline stages, four of which are used, two of which mean the same thing to different people. Conversion rates between stages become meaningless.

Deals that never close or die. Opportunities sitting in a stage for eight months because nobody marks a loss. Pipeline value becomes fiction and forecasting becomes guessing.

Free text where there should be a field. Industry typed by hand produces forty variations of the same industry, and no segment can be built from it.

Contacts with no company, which makes account-level reporting impossible.

Why it is marketing's problem

Because marketing is the function that needs the data to make decisions, and sales is the function that enters it under time pressure during a call.

A salesperson entering a record is optimising for getting back to selling. That is correct behaviour. Which means the system has to be built so the right data is captured without relying on discipline, and designing that is a marketing operations job.

The fixes that hold

Capture source automatically, at form submission, with the campaign data written to the record rather than typed later. A source field filled in by a human is a source field that is empty most of the time.

Required fields, but only three or four. Every required field is friction and every unnecessary one gets filled with junk to get past it.

Picklists instead of free text on anything that will be segmented or reported.

Deduplication rules at entry, not a quarterly merge. Catching a duplicate at creation costs nothing, merging a year of split history costs hours.

Stages with an exit criterion written down. A stage nobody can define is a stage that will be used inconsistently. Five stages with clear definitions beat eleven with vague ones.

An automatic ageing rule, so a deal with no activity for a set period is flagged or closed. Pipeline that cannot be cleaned by a human will be cleaned by a rule or not at all.

The audit worth running first

Four counts, which take an afternoon and usually settle the argument.

How many contact records share an email or a phone number. What proportion of records created in the last quarter have a source. What proportion of open opportunities have had no activity in sixty days. How many distinct values exist in the fields that should be picklists.

Those four numbers tell you whether any report built on this system can be trusted, and they are usually worse than anybody expects.

Doing the cleanup without losing history

Merge rather than delete, because the activity on a duplicate is real history and the record with the most recent activity is not always the one with the most complete data.

Archive rather than purge, so a record that turns out to matter can be recovered.

And fix the entry path before cleaning the data, in that order. A cleaned database with an unchanged entry path is dirty again within two quarters, and the second cleanup never gets funded.

What good looks like

Any question about the pipeline answerable in one query, by one person, in under a minute.

Where it takes a spreadsheet export and two hours of manual work, the reporting will be done rarely and trusted less, and every budget decision downstream of it is being made on a guess.