How do you evaluate a marketing channel?
Give every channel the same five scores, built from your own customer records rather than each platform's dashboard. Return decides. The other four tell you how far to trust that number and whether it's too early to judge.
| Score | The question | Where the answer comes from |
|---|---|---|
| Return | What does one paying customer cost here, against what a customer is worth? | Your spending and your customer list |
| Intent | Was the person already looking for what you sell? | The kind of channel |
| Traceability | Can you tell which customers came from it? | Tagged links, call tracking, asking customers |
| Time to judge | Has it run long enough to read? | Each platform's own guidance |
| Room to grow | Would more money bring more customers at a similar cost? | A step-up test |
Score each from 1 (weak) to 3 (strong), except time, which is "ready" or "too early". Keep the scores side by side instead of adding them up: a strong return with weak traceability needs a better count, not a cut.
Step 1: List every channel and what it really costs
Write down every place customers might find you, paid or not: search ads, pay-per-lead ads, social ads, your Your free business listing on Google Search and Maps: hours, photos, reviews, services and contact details. More, your website and Search engine optimization: work that helps your pages show up higher in the unpaid results on Google and Bing. More, lead sites such as Angi or Thumbtack, email to past customers, referrals, AI assistants that mention you, print and events.
Next to each, put its full monthly cost: ad spend or lead fees, agency and software fees, and your own hours at what your time is worth. Posting on social media is unpaid, not free.
Step 2: Count paying customers per channel, not leads
Take the last three to six months and match every new paying customer to where they came from. Use all three methods; each sees what the others miss.
- Tag your links. Add UTM tags (short labels on a link, such as utm_source and utm_medium) to links in ads, emails and profiles. Google Analytics says that with tags in place, "the parameter values are visible in the Traffic acquisition report" (Google Analytics Help: Collect campaign data with custom URLs). Set up GA4 key events for calls, forms and bookings.
- Track calls. Analytics can't see a phone call that never touched your website. Give each channel its own tracking number, or ask.
- Ask every new customer "How did you hear about us?" and write the answer next to their invoice. It's the only way to count referrals and word of mouth.
Why customers and not leads: at an accounting firm we run marketing for, across Google, Meta and ChatGPT ads the firm spent about $6,100 from October 1 to November 4, 2026. Those ads brought about 13 booked calls and $0 in new-client revenue. Every client who paid in that period came from ChatGPT answers or a Reddit comment, at no ad cost. Counted by bookings, the ads looked like they were working. Counted by customers, they brought none.
Step 3: Score return against your break-even
Two numbers per channel:
- Cost per customer (customer acquisition cost, or CAC): the channel's cost ÷ new paying customers from it.
- Break-even: the gross profit from one new customer, meaning price minus materials and labor. If customers come back, use profit over a year or two (their lifetime value).
| Return score | Rule |
|---|---|
| 3 | Cost per customer is under half your break-even |
| 2 | Between half and all of your break-even |
| 1 | Above break-even, or no paying customer yet |
Set the line from your own margins, not a dollar figure from someone else's business. A cost per customer that is cheap for a roofer can sink a dog groomer.
Google Ads shows return on ad spend (ROAS) in its "Conv. value / cost" column (Google Ads Help: About attribution models), but only for conversions it saw. Check it against your list, or send booked jobs back.
Step 4: Score intent: was the buyer already looking?
High-intent channels reach people who are already shopping. Low-intent channels reach people doing something else. Google draws the same line between its own campaign types:
"It's great for driving sales, leads, or traffic to your website, as you can show your ads to people actively searching for your products and services." (Google Ads Help: Choose the right campaign type, on Search campaigns)
Display campaigns, by contrast, are "a great way to expand your reach and stay top of mind with an audience beyond just Google Search" (same page).
| Intent score | Typical channels | What to expect |
|---|---|---|
| 3 | Search ads, pay-per-lead ads, your Business Profile on Maps, lead sites, referrals, past customers | Fewer people, closer to buying |
| 2 | Ads to past site visitors, email to people who asked for a quote | Warm, but some would come back anyway |
| 1 | Social feed ads, display, video, print, radio | Reaches people before they need you; costs more per customer and takes longer |
Low intent isn't a verdict. Feed and video ads can create demand that shows up later as searches for your business name (Search Console shows these). Give them a longer clock and a smaller share of the budget.
Step 5: Score traceability: can you trust the platform's count?
Each ad platform counts customers by its own rules, so the same customer can appear in two reports:
- Google Ads counts a conversion up to 30 days after a click, unless you change the window (Google Ads Help: About conversion windows).
- Meta counts a website action within 1 or 7 days after a link click, or within 1 day after someone only saw the ad (Meta Business Help: About attribution models and attribution settings).
Someone who sees your Facebook ad on Monday and clicks your Google ad on Tuesday can be claimed by both. Meta even warns against comparing its own results counted under different attribution models, and adds: "If using external analytics tools, evaluate performance in those tools." (Meta Business Help: About attribution models and attribution settings). For a small business, that tool is your customer list.
Splitting credit across every touch sounds fairer, but the options have shrunk:
"The first click, linear, time decay, and position-based attribution models are no longer supported by Google." (Google Ads Help: About attribution models)
Google Analytics dropped the same four in November 2023 (Google Analytics Help: Select attribution settings). In Google Ads, what's left is data-driven attribution, the default for most conversion actions, which splits credit using your account's data, and last click (Google Ads Help: About attribution models). Both credit only touches that were tracked, so a call from your truck sign or a neighbor's recommendation gets nothing.
| Traceability score | Rule |
|---|---|
| 3 | Matched to paying customers on your list |
| 2 | Traced to leads (forms, calls), not to who paid |
| 1 | Known only by asking customers |
Referrals usually score 1 here. That calls for asking every customer, not cutting them.
Step 6: Wait until each channel can be judged
Don't score a channel before it has had a fair run. Mark it "too early" instead.
- Meta ads: an ad set usually leaves its learning phase "after about 50 results in the week after the ad set's last significant edit", and while learning, ad sets "usually have a higher CPA". Ad sets that can't get there show "Learning limited" (Meta Business Help: About the learning phase).
- SEO: Google says "Some changes might take effect in a few hours, others could take several months" (Google Search Central: SEO Starter Guide). See how long SEO takes.
- A brand-new channel: Gabriel Weinberg, co-author of the startup book Traction, says a small test should show "a rough idea of a channel's effectiveness" with at most $1,000 and a month (The Bullseye Framework for Getting Traction, 2015).
Step 7: Score room to grow
A channel can look cheap because it's small: search ads reach only the people searching for your service in your area. Weinberg's test asks the same: what a customer costs there, how many customers the channel can reach, and whether they're the ones you want.
Score 3 if spend went up in steps and cost per customer held, 2 if you haven't tried, and 1 if cost per customer jumped or there's no more volume to buy.
What to do with the scores
| Pattern | Decision |
|---|---|
| Return 3, traceability 2 or 3 | Keep. Raise spend in steps while cost per customer holds |
| Return 1, intent 3 | Fix before cutting. The buyers are there; check the landing page, how fast you answer, and follow-up (ads not getting calls, speed to lead) |
| Return 1, intent 1, past its window | Cut, or shrink to a small test |
| Too early | Leave it alone until its window ends |
| Traceability 1 | Ask every customer for a month before deciding |
Update the numbers monthly. Move money quarterly, so each channel gets a full window between decisions.
Example
Illustrative; numbers are invented.
A plumber's quarter (US dollars)
The average first job clears $300 in gross profit, so $300 is break-even and $150 is the line for a 3.
| Channel | 3-month cost | New paying customers | Cost per customer | Return | Intent | Traceability | Decision |
|---|---|---|---|---|---|---|---|
| Google Search ads | $4,200 | 30 | $140 | 3 | 3 | 3 | Keep; raise spend in steps |
| Lead site | $1,200 | 3 | $400 | 1 | 3 | 3 | Fix: answer leads within minutes |
| Facebook ads | $1,800 | 3 | $600 | 1 | 1 | 2 | Shrink to a test |
| Referral thank-you cards | $150 | 12 | $13 | 3 | 3 | 1 | Keep; ask every customer |
| New service pages (month 2) | $1,500 | 1 | n/a | Too early | 3 | 2 | Wait |
A bookkeeper in Canada with no tracking (Canadian dollars)
She starts asking every new client. After three months: 14 clients, 8 from referrals, 4 from her Google Business Profile, 2 from LinkedIn posts that took 5 hours a week. Valued at C$60 an hour, LinkedIn cost about C$3,900, or C$1,950 per client, against a first-year profit of C$1,500 per client. Return 1, intent 1: she cuts LinkedIn to an hour a week and asks happy clients for referrals instead.
FAQ
What does a marketing channel evaluation framework include?
The same few scores, applied to every channel from the same records each time. Which scores you pick matters less than keeping them fixed.
What is the Bullseye Framework?
A three-step method from the book Traction: brainstorm all 19 channels, test the most promising cheaply, then focus on the one that works.
Is the 70-20-10 budget rule worth following?
Only as a rule of thumb: 70% to proven channels, 20% to promising, 10% to experiments. Let your scores decide what counts as proven.
What is incremental ROAS?
Return counted only on sales the ad caused, leaving out buyers who'd have come anyway. Meta's incremental attribution model predicts which conversions an ad caused (Meta Business Help: About attribution models and attribution settings).
Different for you?
- You haven't set a budget yet: small business marketing budget.
- Your agency reports clicks and leads, not customers: the report to ask for.
- You're choosing between SEO and ads: SEO vs Google Ads.
- You pay lead sites: are Angi and Thumbtack leads worth it.
- You email or text past customers in Canada: CASL consent.
- You'd rather have ad spend moved for you: Paid Ads Agent.