The lead source field was being edited by hand, and everyone knew why.
At a logistics software company, commissions depended on lead source. Marketing-sourced and sales-sourced paid differently. So the field, which was supposed to be an attribution fact, had become a negotiation, and the data reflected whoever cared more about the outcome. Nobody was lying, exactly. They were correcting the record in their favor, which is what any field that pays someone will eventually become.
That's the principle, and it's a design rule. The fix was splitting the field: a Marketing Lead Source, populated automatically from the sync and locked to edits, and a Sales Lead Source, editable by reps and used for commissions. Marketing got attribution it could trust. Sales got a field it could argue over without touching attribution. The argument still happens. It just doesn't corrupt the data anymore.
The same failure with different numbers showed up at a fintech: MQL counts were inflating because a workflow stamped the MQL date on any contact-us submission, including contacts created by support and users invited into the platform by existing customers. Nobody gamed it. The incentive was on the MQL number, so the definition drifted toward whatever produced more of them. Incentive design is data governance whether you treat it that way or not.
Here's the second half, which I didn't expect to write until the last month of client work. Lead source breaks without anyone touching it.
HubSpot's Salesforce sync will overwrite a blank Salesforce lead source with "Salesforce" unless a workflow preserves the original. So every contact that came in through a form and synced across arrives in Salesforce looking like Salesforce created it. An events app pushed numeric IDs into the original-source drill-down field, so a whole event's worth of leads reported a number as their source until it was remapped through a UTM field. A "form submission page URL" workflow trigger fired unreliably; "last page seen" was the property that actually worked. And one sync couldn't carry lead source at all until the field types were rebuilt on both sides.
None of that is a rep's fault. It's what happens when nobody owns lead source as a system: a definition, a locked source of truth, a preservation workflow across every sync, and a monthly check that the values still mean what they meant.
To steelman the single field: it's simpler, and in a small team with no commission tied to it, one lead source is fine. The moment money attaches, or a second system syncs into it, one field can't serve both truth and incentive.
Open your lead source report and look at the top value. If it's the name of one of your systems, you've found the sync overwrite. If it's suspiciously favorable to whoever gets paid on it, you've found the other thing. Which one did you find?
This came out of a client project. The story behind it, and what to check in your own portal, goes to subscribers on Thursdays. Get Thursday's note.