Best Marketing Attribution Metrics: The 5 That Matter

Marketing attribution metrics that actually change decisions are narrower than most dashboards suggest. For a US service business — HVAC, law, dental, home remodeling, med spa — the five that matter are cost per qualified lead (CPQL), lead-to-sale conversion rate by source, customer acquisition cost payback period, incremental return on ad spend (iROAS), and contribution margin per channel. Everything else — impressions, clicks, last-click ROAS, "engagement" — is diagnostic at best. If you track only one, track cost per booked job by source, because it survives contact with your P&L. The rest of this page covers how to calculate each, the benchmarks, and the specific situations where attribution measurement is a waste of money.

The Five Metrics That Move Money

1. Cost per qualified lead (CPQL). Total channel spend ÷ leads that met your qualification bar. Not form fills. A roofing contractor spending $9,000/month on Google Ads generating 180 form fills looks like $50/lead. If 62 of those are in-service-area, insured, and have a real project, the real number is $145. Across US home services, raw CPL typically lands between $45 and $120 while CPQL runs 1.8x to 3.2x higher.

2. Lead-to-sale conversion rate, segmented by source. Google Local Services Ads leads convert to booked jobs at roughly 25–35% for most home service categories. Facebook lead-form leads convert at 4–9%. Organic search and referral leads often clear 40%. A blended 18% conversion rate tells you nothing; the segmented view tells you where to move $3,000.

3. CAC payback period. How many months until a customer's gross profit repays acquisition cost. A dental practice with a $420 CAC and $310 in first-visit gross profit has a payback near 1.4 visits — fine. A med spa with a $680 CAC on a $199 intro offer has a 5–7 month payback and needs a membership program, not more ad spend. Under 3 months is healthy for services; over 12 months means you're financing growth from cash reserves.

4. Incremental ROAS. What a channel produced *beyond what you'd have gotten anyway*. Branded search is the classic offender: pause it for two weeks and measure the delta in total booked jobs. Published geo-holdout tests across retail and services have found branded search incrementality between 10% and 50% — meaning a reported 12x ROAS may be a true 1.5x.

5. Contribution margin per channel. Revenue minus COGS minus channel spend minus the sales labor to close those leads. A channel producing $40,000 revenue at a 34% gross margin and $9,500 spend contributes $4,100 — and if it took 26 hours of estimator time at $38/hour to close, it contributed $3,112.

Attribution is not a reporting problem. It is a budget-reallocation problem. If a metric would not cause you to move at least $1,000 next month, stop tracking it.

The Metric Almost Nobody Tracks: Time-to-First-Contact by Source

Here is the number missing from every attribution dashboard we audit: median seconds from lead creation to first human or AI contact, broken out by source. It belongs in your attribution stack because it is a confound. Response-time research going back to the Harvard Business Review lead-response study found contact rates drop roughly 10x between a 5-minute and a 30-minute response, and roughly 21x by 30 minutes versus 5.

That means a channel's "conversion rate" is partly measuring your ops, not the channel. Facebook leads arrive at 9:40 PM. LSA calls arrive during business hours. If your median response is 4 minutes on calls and 6 hours on forms, you will conclude Facebook is garbage when you have actually measured your own after-hours staffing. Segment conversion rate by *both* source and response-time bucket (under 5 min, 5–30 min, 30+ min) before you cut a channel. We've watched businesses kill a $4,000/month channel that was converting at 6% under a 4-hour response and would have converted near 15% under a 5-minute one. Fixing speed-to-lead with automated and AI-assisted follow-up is usually cheaper than replacing the channel.

How to Build the Measurement Stack (Cheapest to Most Expensive)

Most service businesses under $10M revenue should live at Tier 1–2 permanently. Our ROI calculator will get you a defensible CAC and payback figure without any of this tooling.

When Attribution Isn't Worth It — And Who Should Not Buy This

This is the part that costs us business, so we'll be direct.

If you spend under $3,000/month on marketing, don't buy an attribution system. The statistics don't work. At a $95 CPQL, $3,000 buys about 31 qualified leads. Detecting a real 20% difference between two channels at that volume requires roughly 5–6 months of data — by which point seasonality, a price change, and a competitor's new location have all corrupted the comparison. Ask every caller "how did you hear about us," log it in a spreadsheet, and spend the software budget on ads instead.

If your close rate is under 20% and you don't know why, fix sales first. Attribution tells you which faucet to open. It's useless if the bucket has a hole. A plumbing company we reviewed had beautiful multi-touch reporting and a 32% unanswered-call rate — 47 missed calls a month at roughly $1,100 average ticket. That's $17,000/month in leaked revenue that no dashboard would have surfaced. No attribution metric fixes that.

Multi-touch attribution is oversold for service businesses. The 6–14 touchpoint journeys that justify MTA are B2B SaaS patterns. An emergency AC repair has a 90-minute consideration window and one or two touches. Paying $1,500/month for algorithmic MTA on a 2-touch journey is buying precision you can't act on. Even for considered purchases — cosmetic dentistry, roof replacement, personal injury — the honest ceiling is a first-touch/last-touch split plus a self-reported source field.

Failure modes to expect, plainly:

Do not hire an attribution consultant if you have fewer than 40 leads/month, no CRM, or one dominant channel producing 70%+ of revenue. In that last case, you don't have an attribution problem — you have a concentration risk problem, and the money is better spent testing a second channel.

A Reporting Cadence That Actually Gets Used

The trap is monthly dashboards nobody reads. If a report hasn't changed a spending decision in three consecutive months, delete it.

What Good Numbers Look Like

Rough US service-business benchmarks to calibrate against — your category and market will shift these by 30–40%:

If you want a second set of eyes on which of these you're actually hitting, the free marketing audit covers exactly these five numbers against your real data, and our case studies show the before-and-after CPQL figures from businesses that made these changes.

Pick the five metrics at the top. Track them monthly. Ignore the rest until you're spending enough that the math earns its keep.