TL;DR
Self-reported attribution beats pixels, call tracking, and platform-reported conversions for US service businesses. Here's the ranked tier list and why.
→ See how this applies to your business (free 30-min call)Marketing attribution sources fall into four tiers, and for a US service business the ranking is not close. Self-reported attribution ("How did you hear about us?" on the booking form) is the single most valuable source — it catches the 40-60% of influence that no pixel can see. Second: call tracking with dynamic number insertion, since 30-50% of service-business leads arrive by phone. Third: CRM-side revenue data (closed-won dollars, not form fills). Fourth, and least useful for businesses under roughly $5M in revenue: platform-reported conversions from Google Ads, Meta, and GA4, which systematically over-credit themselves.
Most attribution advice is written for e-commerce companies with 10,000 monthly transactions. A plumbing company doing 180 jobs a month cannot run a multi-touch attribution model. The statistics do not work. Here is what does.
Why Statistical Attribution Models Fail Below ~500 Conversions/Month
This is the part almost nobody says out loud. Data-driven attribution in Google Ads requires a minimum of 300 conversions and 3,000 ad interactions in a 30-day window before Google will even enable the model. Meta's attribution reporting needs comparable volume before its numbers stop bouncing.
A typical US home-services business books 60-250 jobs a month. At 150 conversions spread across 6 channels, you have 25 conversions per channel — and a 95% confidence interval on a 25-event sample is roughly ±40%. You cannot distinguish a channel producing $18 cost-per-lead from one producing $25. The model output looks precise. It is noise with decimal places.
If your monthly conversion count is under 300, every dollar you spend on attribution software instead of on asking customers where they came from is a dollar spent buying false precision.
The practical threshold: below 300 monthly conversions, use self-reported attribution plus call tracking and stop there. Between 300 and 1,000, add CRM revenue matching. Above 1,000, multi-touch modeling starts to earn its keep.
The Four Attribution Sources, Ranked by Actual Usefulness
1. Self-reported attribution (post-conversion survey). One required field on the booking form or the intake script: "How did you first hear about us?" with 6-8 options and an open text box. Cost: roughly $0 to implement, about 20 minutes of form work. It is the only source that captures the neighbor's recommendation, the truck wrap, the church bulletin, and the Nextdoor thread.
The known weakness: recall accuracy is poor. People say "Google" when they mean "I saw your Facebook ad three weeks ago, then searched your name." Self-reported data over-credits the last channel by a wide margin and under-credits display, video, and social. Treat it as directional, not precise.
2. Call tracking with dynamic number insertion (DNI). For service businesses, 35-60% of qualified leads come by phone, and a phone call is invisible to a website pixel unless you instrument it. DNI swaps the displayed number based on traffic source, so a call from a Google Ads click reports as a Google Ads call. Cost runs $30-$150/month for a small pool of numbers on platforms like CallRail or WhatConverts — often $45/month for the tier most 1-3 truck operations need.
The trap: DNI only works if you retire every hardcoded phone number on the site, in your Google Business Profile, on Yelp, and on directory listings. Half-implemented DNI produces attribution data that is worse than none, because it looks authoritative.
3. CRM revenue data. Form fills are not the business. A channel producing 40 leads at $22 each looks better than one producing 12 leads at $65 — until you find the cheap channel closes at 8% on $400 tickets and the expensive one closes at 45% on $6,200 tickets. That is $1,280 in revenue versus $33,480 from the "expensive" channel. Push closed-won amount and job type back from your CRM (Jobber, ServiceTitan, Housecall Pro, GoHighLevel) into your ad platforms via offline conversion import.
4. Platform-reported conversions. Google Ads, Meta Ads Manager, and GA4 each measure their own homework. Run all three and add up the conversions: you will routinely find 130-180% of your actual lead count, because two or three platforms claim the same lead. Use platform data for in-platform optimization decisions — which keyword, which creative — and never for cross-channel budget allocation.
The Instrumentation Stack That Actually Fits a Service Business
Our ROI calculator will show you what a 10-point shift in close rate does to channel economics before you spend anything on tooling.
When Attribution Is Not Worth It — Read This Before You Buy Anything
We sell attribution work. Here is when you should not buy it from us or anyone else.
The failure mode we see most often: a business spends $6,000 on an attribution build, learns that Google Ads drives 62% of revenue, and does exactly what it was already doing. That is $6,000 for a confirmation. Before you start, write down the specific decision you will make differently depending on the result. If you cannot name one, do not start.
Also worth naming plainly: attribution can be actively harmful when it kills upper-funnel spend. Brand search, YouTube, and referral-generating activity look terrible in last-click and mediocre in most multi-touch models, because they influence demand that later shows up as a branded search. Businesses that cut every "unprofitable" channel by the dashboard often watch branded search volume decline 15-30% over the following two quarters and never connect the two.
What the Numbers Look Like When It Works
A realistic before/after for a 12-tech HVAC company spending $18,000/month:
That reallocation is the entire value of attribution. Not the dashboard.
How to Start This Week for Under $100
Add a required "How did you hear about us?" field to every form and script it into phone intake today. Zero dollars.
Sign up for call tracking at the entry tier — about $45/month — and put one DNI number on the site. Two hours of work.
Export 90 days of closed-won jobs from your CRM into a spreadsheet with source and job value. Sort by revenue per source, not lead count.
Wait 30 days. Compare self-reported source against platform-reported source. The gap between them is your measurement error, and it is usually 30-50%.
That sequence costs about $45 and a weekend, and it answers the budget question for most businesses under $10,000/month in ad spend. Everything above it — server-side tagging, MMM, incrementality testing — is a real discipline, and it is the right investment somewhere north of $25,000/month in media. Our services page covers where that line sits for each type of operation, and the case studies show the reallocation math on real accounts.
The best attribution source is the one that changes what you do on Monday. For most US service businesses, that is a free form field and a $45 phone number — not a platform.
Frequently Asked Questions
What is the most accurate marketing attribution source?
For service businesses, self-reported attribution — asking "How did you hear about us?" on the booking form — is the most accurate single source. It captures word-of-mouth, offline exposure, and dark-social influence that pixels and cookies never see, which accounts for roughly 40-60% of the influence on a booking decision.
Why do Google Ads and Meta over-report conversions?
Each platform only sees its own touchpoints and claims credit whenever one appears in the path, so a lead touched by both is counted twice. Google Ads and Meta also use view-through windows and modeled conversions. Summing platform-reported conversions typically exceeds the actual number of leads in your CRM.
Do I need call tracking for attribution?
If you take leads by phone, yes. Roughly 30-50% of service-business leads arrive as calls, and those are invisible in web analytics without call tracking. Dynamic number insertion swaps the displayed phone number by traffic source, tying each call back to the campaign, keyword, or channel that produced it.
How many conversions do you need for multi-touch attribution?
Multi-touch models need thousands of monthly conversions to separate signal from noise across paths. A business doing 180 jobs a month splits those across channels and touch sequences until each path has a handful of data points — far too few to be statistically meaningful. Use self-reported attribution and CRM revenue data instead.
Free Weekly Briefing
One AI Marketing Tactic.
Every Tuesday. Free.
What's actually working across our client accounts right now — ROAS moves, follow-up sequences, creative angles. The stuff that isn't in any blog post yet.
No spam. Unsubscribe anytime. 1,200+ business owners already in.