How do you optimize marketing spend?
Move money toward channels that win paying customers for less than a customer is worth, and away from the rest, on a fixed schedule and by rules you write down first.
| Step | What you do | Done when |
|---|---|---|
| 1. Measure | Work out the cost of one paying customer per channel, and your break-even | Every channel has one number you trust |
| 2. Protect | Fund the channels under break-even first | Proven channels are paid for before anything new |
| 3. Test | Put a capped amount on one new idea at a time | Each test has a question, a cap, an end date and a pass mark |
| 4. Move | Shift money once a month, by rule | Extra money only goes where it can buy more customers |
The U.S. Small Business Administration sums up why the measuring comes first:
"Measuring ROI will help you know which part of the plan is working and which part needs to be updated." (SBA: Manage your business, Marketing and sales)
How much to spend in total is a separate question: small business marketing budget. This page is about where the money goes once you have a number.
Step 1: which metrics show where to move money?
Cost per paying customer, set against what one customer earns you. Every other number is a clue.
| Metric | Formula | Use it for | Catch |
|---|---|---|---|
| Cost per lead | Spend ÷ calls, forms and bookings | Spotting a channel that went quiet | A lead isn't a sale |
| Customer acquisition cost (CAC) | Ad spend plus fees ÷ new paying customers | Comparing channels | Count only customers your records tie to that channel |
| Break-even CAC | Gross profit from one new customer, repeat work included | The line each channel must stay under | Profit, not price |
| Return on ad spend (ROAS) | Revenue from ads ÷ ad spend | Businesses that record a sale value | Needs real revenue; a good number depends on margin |
Google shows ROAS as plain division: "$5 USD in sales ÷ $1 USD in ad spend x 100% = 500% target ROAS" (Google Ads Help: About Target ROAS bidding).
No ROAS is good in itself. Break-even ROAS is 1 ÷ your gross margin. At a 25% margin, 4 to 1 only breaks even; at 50%, anything above 2 to 1 is profit. A rule like "add budget above 4 to 1" pours money in at break-even for the first business and holds the second one back.
Two traps make platform reports look better than your bank account:
- Platforms count actions, not customers. A call or a form counts the moment it happens; the platform can't see who paid unless you tell it (the report to ask for).
- Two platforms can claim one customer. Meta's standard view-through setting counts actions that happen "within 1-day after an impression of your ad" (Meta Business Help: About attribution models and attribution settings). Google records conversions up to 30 days after an ad click by default (Google Ads Help: About conversion windows). Someone who saw a Meta ad, then clicked a Google ad, can show up in both. Compare the platforms' total with your own count of new customers.
Your own records settle it: the booking form, the call log, invoices, and a "How did you hear about us?" question. To feed paying customers back into Google Ads: tracking a lead to a booked job.
Step 2: protect the channels that already pay
Rank every channel by CAC, including ones that cost time rather than ad money: referrals, your Your free business listing on Google Search and Maps: hours, photos, reviews, services and contact details. More, Search engine optimization: work that helps your pages show up higher in the unpaid results on Google and Bing. More, email to past customers. Fund the ones under break-even first, and don't raid them to pay for tests.
Judge each channel on its own clock: search ads after a few weeks of leads, SEO after months (how long SEO takes), email to past customers after each send (what to send). Ads or SEO first: SEO vs Google Ads.
Step 3: how much should go to experiments?
No platform or regulator sets a share. The 10% or 15% you see quoted are rules of thumb. Size each test by what it needs to give a clear answer. Meta puts it this way:
"Your A/B test should have a budget that will produce enough results to confidently determine a winning strategy." (Meta Business Help: Best practices for A/B tests in Meta Ads Manager)
Work it backward: results needed × your cost per result. For scale, Meta's delivery usually settles "after about 50 results in the week after the ad set's last significant edit," and until then ad sets "usually have a higher CPA" (Meta Business Help: About the learning phase). At $40 a lead, 50 leads a week is $2,000 a week. Few small budgets get there, so ask fewer, bigger questions and run them longer. If a test would starve a proven channel, the test waits.
How do I test a new channel or idea safely?
Write the test down before spending: the question, the one change, the budget cap, the end date and the pass mark. Then:
- Change one thing. Meta's results are clearest when ad sets are "identical except for the variable that you're testing" (Meta Business Help: Best practices for A/B tests in Meta Ads Manager).
- Use the platform's split tool. Meta advises against testing by "turning ad sets or campaigns on and off manually" (Meta Business Help: About A/B testing for Meta Ads). In Google Ads, a custom experiment runs a changed copy of a Search, Display, Video or Hotel campaign beside the original, sharing its traffic and budget. Editing either side mid-test "may make it harder to interpret your results" (Google Ads Help: Set up a custom experiment).
- Run it at least 7 days. Meta says "tests shorter than 7 days may produce inconclusive results," and Ads Manager allows up to 30. Go longer if customers take more than a week to decide.
- Decide on the end date, by the pass mark you wrote. Paying customers if you get enough, booked leads if you don't.
A brand-new channel with no split tool (a directory, a sponsorship, a new ad platform) gets the same treatment: a hard cap, a fixed end date, and a pass mark in cost per paying customer. It gets more money only after it passes.
Step 4: when should I move budget, and how fast?
Once a month, on a set day, using full months of data. Between reviews, only stop plain waste (search terms report). Leave the latest days out of the verdict: with Google's default 30-day window, a click from last week can still turn into a recorded conversion (About conversion windows).
| Signal | Move |
|---|---|
| CAC under break-even, and the campaign loses searches to budget | Add money |
| CAC under break-even, but it loses searches to rank | Fix ads, keywords or the page first; money won't help much |
| CAC over break-even for two reviews, with enough customers to judge | Cut it; move the money to the best channel or the test budget |
| Too few customers to judge | Hold steady; don't move on noise |
Google Ads has a column that separates budget from rank:
"Search lost IS (budget): The percentage of time that your ads weren't shown on the Search Network due to insufficient budget." (Google Ads Help: Get impression share data)
Its partner, Search lost IS (rank), counts searches lost "due to poor Ad Rank in the auction." When little is lost to budget, extra money mostly buys pricier clicks. An accounting firm we run marketing for lost about 80% of eligible searches to rank and none to budget. Doubling its Google daily budget from $180 to $360 raised spend 62% and clicks only 24%.
How fast to move depends on the platform. Google says budget updates "take effect in the next auction, but never trigger 'Learning' status," and that Smart Bidding performance "typically converges to your new goal within 1 to 2 conversion cycles" (Google Ads Help: About Smart Bidding learning). A conversion cycle is the time from click to conversion. Meta is different: a small change rarely matters, a large one can restart learning.
"if you change your budget from $100 to $1000, one or more ad sets may reenter the learning phase." (Meta Business Help: Significant edits and learning phase)
So on Meta, raise in steps and leave each step alone for a week or so. Bigger isn't always cheaper: when the same firm doubled a Meta budget from $50 to $100 a day, its cost per 1,000 The number of times your ad or listing was shown, whether or not anyone clicked. went from $26 to $56–58 for the same audience.
Do I need software for this?
No. A spreadsheet with one row per channel per month (spend, leads, paying customers, revenue, CAC) covers it, plus the free tools above: lost impression share, Google Ads experiments, Meta A/B tests. Automation such as Smart Bidding or Meta's Advantage+ campaign budget works inside one platform, toward whatever you told it to count. Choosing between Google and Meta, or ads and SEO, stays with you.
Example
Illustrative; round numbers.
A plumber with $4,000 a month, US dollars
Per month last quarter: Google Search $2,500 brought 10 paying jobs ($250 each); Meta $1,000 brought 2 ($500 each); a directory $500 brought 1. Break-even is $400 of gross profit per new customer. Search loses 30% of searches to budget. The move: Search to $3,000, Meta to $500 as an A/B test of one new ad, the directory paused, $500 held back. Next month shows whether the extra $500 still buys jobs under $400.
A dental practice testing a video ad
Question: does a video book new patients cheaper than the photo ad? Meta A/B test, $20 a day per version for 14 days, $560 in all. Pass mark: cost per booked visit 25% lower. Result: 4 bookings for the video, 3 for the photo. That meets the pass mark on paper, but one booking either way is noise, so the photo ad stays, and the test reruns next quarter with a bigger cap.
A bookkeeper who doubled the budget, Canadian dollars
Google Search at C$1,500 a month wins clients at C$300 each, against a C$600 break-even. Doubled to C$3,000, it barely adds clients: lost impression share was 2% to budget and 70% to rank, so the extra money bought pricier clicks. Budget goes back to C$1,500; the work goes into the ads and the service page.
FAQ
What percentage of revenue should go to marketing?
There's no official figure. The SBA says to compare marketing and sales costs with the revenue they bring; work back from one customer's worth (how).
How often should I move my marketing budget?
Monthly for most small businesses. On Meta, big or frequent changes can restart learning; on Google, budget changes apply at once, with no learning reset.
Should I cut the channel with the worst numbers?
Check your own records first. A channel customers name when asked "How did you hear about us?" may do more than its dashboard shows.
Different for you?
- You haven't set a total budget yet: small business marketing budget.
- Your work is seasonal: HVAC, roofing and plumbing budget by season.
- You email or text Canadian customers: CASL consent.
- Leads come in but few become customers: website not getting leads.
- You'd rather have the ads run and moved on rules like these: Paid Ads Agent.