Marketing Analytics 101: How to Tell What Is Actually Working

Most owners are not short of marketing data. They are short of a way to decide anything with it. The platforms each report their own version of success, the totals never reconcile, and the monthly report arrives full of figures that go up without the bank balance following. This guide is about separating the numbers that change decisions from the numbers that merely fill a slide.

What is marketing analytics?

Marketing analytics is the part of marketing that connects what you spent to what you earned, closely enough to decide where the next dollar goes. That is the entire purpose. Every metric that cannot eventually be tied back to that question is decoration, however interesting it looks on a chart.

The distinction that makes the rest of this manageable is between activity and outcome. Impressions, followers, sessions and click-throughs are activity — they describe motion. Enquiries, booked jobs, revenue and retention are outcomes. Activity is worth watching as an early warning system, but no amount of it substitutes for an outcome, and a report built entirely of activity metrics is a report designed not to be judged.

The numbers that actually matter

For a local service business, five figures carry nearly all the decision-making weight. If you track nothing else, track these.

1. Cost per acquired customer. Not cost per click and not cost per enquiry — the total spend divided by the number of people who actually paid you. This is the number that says whether a channel is viable.
2. Customer lifetime value. What an average customer is worth across every job they will ever give you, including the ones they refer. Without this figure you cannot tell whether your acquisition cost is a bargain or a slow leak.
3. The ratio between the two. A business acquiring customers for a fifth of what they are worth can afford to be aggressive. One at half is fragile. One above one is losing money on every sale and making it up in volume.
4. Close rate by source. The share of enquiries from each channel that turn into work. A source producing plenty of enquiries at a terrible close rate is generating labor, not demand.
5. Speed to first response. Not strictly analytics, but it moves outcomes more than almost anything else you can measure, and it is entirely within your control.

Metrics that mislead

Some numbers are not wrong so much as routinely over-read. These are the ones we most often find driving decisions they should not be driving.

• Impressions. A count of times something was theoretically on a screen. Useful as a denominator, meaningless alone.
• Follower count. A vanity figure that correlates with revenue far more weakly than anyone expects, and can be purchased.
• Bounce rate. Frequently misunderstood. A visitor who read your hours, got the phone number and left has technically bounced and has also become a customer.
• Time on page. Ambiguous in both directions — it rises when content is engrossing and when it is confusing.
• Total traffic. The most commonly celebrated number and one of the least informative. A thousand visits from the wrong area are worth less than forty from the right one.

If a metric has never once caused you to do something differently, it does not belong on the report. Most dashboards would be more useful with two thirds of the tiles deleted.

Attribution, and why the totals never agree

Attribution is the problem of deciding which touch deserves credit for a sale, and it is genuinely hard rather than merely fiddly. A customer sees you three times over six weeks before calling. Each platform counts that customer as its own, so adding the platform reports together produces a number larger than the truth.

This is not a bug you can configure away. Some practical ways to live with it:

• Use one source of truth for outcomes — normally your own booking or customer record — and treat platform numbers as estimates rather than facts.
• Ask every caller how they heard about you and write the answer down. It is imperfect and self-reported, and it still catches things no tracking will.
• Use distinct phone numbers or tracked links per channel where the volume justifies it.
• Watch the aggregate. If total spend rose thirty percent and total booked work did not move, the detail of which platform claimed what is a secondary question.
• Run a holdout. Turning one channel off for a month is blunt, unpopular and the most honest measurement available.

Building a report you will actually read

The monthly report fails for a predictable reason: it is built to demonstrate effort rather than to prompt a decision. A useful one fits on a page and answers three questions — what did we spend, what did we get, what are we changing because of it.

Include the things that went badly. A report with no bad news is not a report, it is marketing collateral, and it means somebody is filtering what reaches you. This is one of the questions worth asking before you sign a marketing contract, because the answer tells you a lot about how the relationship will run.

How long before the numbers mean anything

Small samples lie confidently. Twelve enquiries in a month is not enough to rank channels, and the temptation to act on a two-week dip is how good work gets cancelled just before it compounds. Give a new effort a quarter before judging it, and compare year over year rather than month over month wherever seasonality is a factor — which for most local trades it very much is.

If you want an outside read on which of your numbers are load-bearing and which are noise, that is the first thing we do in a proposal.