Your marketing is working if it brings in customers you can trace back to it, at a cost that pays back within the time your business can afford. Five numbers settle the question: leads per month by source, cost per lead, lead-to-customer rate, payback period, and the share of new customers who did not come from referrals. This guide shows where to find each one in about ten minutes, what the 2026 benchmarks say is normal, and the three patterns that mean the answer is no.
Why most owners cannot answer this question
Not because they are careless. Because the numbers live in five tools and nobody has the hour it takes to put them together. In a June 2025 survey of 2,500 small-business decision-makers in the US, UK, Canada and Australia, Constant Contact (2025) found that only 18% felt very confident in the results of their marketing, down from 27% a year earlier; the top frustration worldwide was "not knowing what's working", and 42% had less than one hour a day for marketing at all. Effort went up, confidence went down.
The fix is not a dashboard project. It is five numbers, checked once a month, in the same order.
What are the five numbers that show whether marketing is working?
Leads by source, cost per lead, lead-to-customer rate, payback period and non-referral share. Each one answers a question the one before it raises, which is why the order matters: volume first, then cost, then quality, then whether the economics close, then whether you have built something beyond word of mouth.
| Number | The question it answers | Where to find it in 10 minutes |
|---|---|---|
| Leads per month, by source | Is anything coming in, and from where? | Form submissions and calls in your CRM or inbox, split by source (Google, ads, social, referral, AI assistant) |
| Cost per lead (CPL) | What does each lead cost, by channel? | Monthly spend per channel, including tools and people, divided by that channel's leads |
| Lead-to-customer rate | Are these the right leads? | Customers won this quarter divided by leads received, by source |
| Payback period | How many months until a customer repays what it cost to win them? | Cost to win a customer divided by monthly gross profit per customer |
| Non-referral share | Have we built a source of customers we can turn up? | New customers from search, ads and content divided by all new customers |
Leads by source. Open your CRM, your form tool or your inbox and count the enquiries from the last 30 days. Tag each one with how it found you. If you cannot tell, that is the first fix: a "how did you hear about us?" field and consistent tracking links on every campaign. Our CAC, payback and LTV guide explains the tracking setup.
Cost per lead. Add up what each channel cost last month, including the share of a freelancer, an agency or a tool that serves it, and divide by its leads. A channel with spend and zero leads for two months running is the clearest signal in this whole exercise.
Lead-to-customer rate. This is where "lots of leads" turns into "the wrong leads". A channel that produces half the enquiries but none of the customers is costing you the hours spent answering it.
Payback period. Take what it cost to win a customer in a channel and divide by the gross profit that customer brings per month. If payback is longer than the time you can afford to wait, the channel is not working, however good its cost per lead looks.
Non-referral share. Referrals are the best customers most small companies have, and they are the one source you cannot turn up on demand. If nine out of ten new customers came from word of mouth, marketing is not failing; it has not started.
What does "normal" look like in 2026?
The most useful public benchmark for the first three numbers is the annual search-advertising study by WordStream by LocaliQ (May 2026), which analysed 13,474 US search campaigns across 23 industries between April 2025 and March 2026. Medians across industries: cost per click $5.42 (up from $5.26), click-through rate 6.64%, conversion rate 8.18%, and cost per lead $66.69, down from $70.11. LocaliQ called it the first overall drop in cost per lead since before 2020. Two caveats: these are paid-search leads only (a form, call, chat or email counts), and the spread between industries is wide, so compare yourself with your own category in the report, not with the median.
| Metric (US search ads, medians) | 2026 benchmark | What it means for you |
|---|---|---|
| Cost per click | $5.42 | Clicks got slightly dearer; the page that receives them has to convert |
| Click-through rate | 6.64% | Below this, the ad or the keyword match is the problem |
| Conversion rate | 8.18% | Below this, the landing page or the offer is the problem |
| Cost per lead | $66.69 | Above your industry's figure for three months, change the plan, not the budget |
For the last two numbers there is no honest universal benchmark; payback depends on your margin and non-referral share on your age and category. A practical rule used in our 90-day playbooks: payback inside the period you can fund from cash flow, and a non-referral share that rises every quarter.
Three patterns that mean it is not working
Activity without leads. Posts go out, the agency sends a report, the website looks fine, and the enquiry count has not moved in six months. In the survey above, this is the "not knowing what's working" frustration; in your numbers it is a flat first row. The cause is usually that no channel has been run properly long enough, because effort is spread across six of them.
Leads without customers. Enquiries arrive and die: wrong fit, slow follow-up, or a price mismatch nobody noticed because nobody read the lead-to-customer rate by source. Fix follow-up first (same day, every lead), then the targeting.
Customers without payback. The channel produces customers, but each one costs more than it returns within the time you can wait. Paid channels do this quietly when they are judged on cost per lead alone. Switch the measure to payback and the budget usually moves on its own.
The two sources most owners under-count
Referrals are usually under-recorded because nobody asks. Add the question to every form and every first call; the answer changes which channel gets next month's budget.
AI assistants are new and small, but they are the fastest-growing line in the source column, and they convert. Across 1,215 US enterprise domains, Conductor (2026) measured AI referrals at 1.08% of all website traffic between May and September 2025, 87.4% of it from ChatGPT; enterprise sites only, and the share has grown since. On the conversion side, Adobe Analytics data reported by TechCrunch (2026) showed AI-referred visitors to US retail sites converting 42% better than non-AI traffic in March 2026. Retail only; do not assume the same lift for a services firm, but do count the channel. In Google Analytics 4, create a channel group that collects chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai so the line exists before it matters.
Whether you show up in those answers at all depends less on your website than on how often other sites mention you: across 75,000 brands, Ahrefs (2025) found branded web mentions correlate 0.664 with appearing in Google's AI Overviews, against 0.218 for backlinks. Our GEM guide covers what to do about it.
What to do with the answer
If the five numbers look healthy: protect the measurement, raise the budget on the channel with the shortest payback, and review the five numbers monthly, in the same order, so a change shows up as a trend rather than a surprise.
If one number is missing: fix measurement before strategy. A "how did you hear about us?" field, tracking links on every campaign and one sheet with the five rows is a two-hour job that makes every later decision cheaper.
If the pattern is activity without leads: stop spreading effort. Pick the one channel your customers actually use to find businesses like yours, run it properly for 90 days with one metric, and judge it on that. The 90-day timeline shows what "properly" looks like by channel.
If you want someone to run this check for you: the Free Growth Assessment is this exercise applied to your business by our AI agents and reviewed by a senior strategist, returned as a 3-page report within 48 hours, with the recommendation included even when it is "keep doing it yourself". No credit card, no obligation.