If you don't know which channels generate which customers, you'll over-invest in the worst ones. Attribution doesn't have to be sophisticated, it has to exist.
The minimum viable system. A single field on every new customer record: "How did you hear about us?" Filled in by whoever takes the call or processes the signup. Simple values: Google search, GBP, referral (from whom), door hanger (which neighborhood), Facebook, Nextdoor, drive-by, other.
Better: source-tagged phone numbers. Use a call-tracking service (CallRail, etc.) to assign distinct phone numbers to each major channel (your GBP, your website, your door hangers). Calls automatically attribute to source. Cost: $30–$100/month, easily worth it past 3 trucks.
Better still: UTM parameters and Google Analytics. For digital channels, every link should carry UTM parameters identifying campaign source. Combined with conversion tracking on your website's signup form, this gives you channel-level CAC and LTV.
What to compute monthly.
- New accounts by source (count and % of total)
- Spend by source (where you can attribute)
- CAC by source (spend / accounts)
- Average ticket by source
- Estimated LTV by source (average ticket × estimated tenure)
Decisions this drives.
- Channels with CAC over 12 months of revenue: reduce or kill
- Channels with CAC under 6 months of revenue: scale aggressively
- Channels with CAC between: hold steady, optimize
Common mistakes.
- Asking "how did you hear about us?" but never aggregating the answers
- Confusing first-touch (saw a billboard) with last-touch (Googled and clicked), most customers have multiple touches; pick a model and stick with it
- Stopping channels prematurely without enough data (need 30+ acquired customers from a channel before judging it)
- Ignoring referrals because they "happen automatically", the channel does have a CAC (your service quality and review investment) and a measurable LTV
Attribution doesn't have to be perfect. Directionally right and consistently tracked beats "we'll figure it out later" forever.