Why does marketing attribution in ServiceTitan get worse the longer a customer has been with us?
Because the campaign or source field usually reflects the most recent booking path, not the original one, and repeat customers book through channels that have nothing to do with acquisition. A customer acquired through paid search two years ago who now calls the office directly will show up attributed to whatever source the office booking uses. Acquisition attribution has to be stored at the customer level and frozen.
Two different questions, one field
There are two legitimate attribution questions. Which channel acquired this customer, and which channel produced this specific job. A single campaign field on a job cannot answer both, and most reporting quietly assumes it answers the first when it is actually answering the second.
The distortion compounds over time. Your best channels are the ones producing customers who come back, and coming back is exactly what erases their credit. The channels that produce one-and-done customers keep their attribution intact forever, so they look relatively better every year purely as an artifact of how the field is maintained.
How the erosion actually happens
- Repeat calls land on the main number. The tracked number that produced the original call is not what the customer saved in their phone.
- Office staff select a default source. Under call volume, whatever is first in the dropdown wins.
- Membership and recurring maintenance jobs get a service source. Correct operationally, misleading for marketing.
- Records get merged. Duplicate cleanup often keeps one record's source and discards the other's.
The fix is a frozen first-touch record
Store acquisition source on the customer record the first time that customer is created, and never let it be overwritten. Keep the per-job source separately. Then you can report both: lifetime revenue by acquisition channel, and current-period revenue by booking channel. They will be very different numbers and both are useful.
This usually cannot be done inside the field service system alone, because the system is designed around the current state of a record rather than its history. It is done in the integration layer, which keeps its own customer table with an immutable first-touch field. That pattern is central to how we do AI systems integration.
A diagnostic you can run this week
Pull jobs for the last quarter, split them into first-ever jobs for that customer versus repeat jobs, and report source distribution for each group separately. If your repeat group is dominated by office, referral or unknown sources while your new-customer group is dominated by paid and organic channels, your blended attribution report has been crediting the wrong things for years.
The gap between those two distributions is roughly the size of the error in your cost-per-acquisition math. Correcting it usually moves budget rather than adding it, which is the useful kind of finding from marketing intelligence.
Topics: attribution · campaigns · repeat customers · data quality
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.