Decision metrics
What is a marketing KPI, and which ones matter?
Most marketing reports are full of numbers that cannot change anything. A KPI is the short list that can, agreed before the work starts rather than after it.
What is a marketing KPI and which ones actually matter?
A marketing KPI, or key performance indicator, is one of the small number of measures you have agreed in advance to judge marketing by, chosen because a change in it should change what you do next.
For most local, healthcare and service businesses the ones that survive that test are qualified enquiries, cost per qualified enquiry, the share of enquiries that become booked work, and revenue from new customers.
A number that cannot change a decision is a report line, not a KPI, and impressions, followers and keyword counts are usually report lines.
If the number moves, what do you do differently?
That one question separates a KPI from a statistic. Ask it of every line in your current report and most of them will not survive.
Take impressions. Impressions doubled last month. What will you now do differently? If the honest answer is nothing, impressions are not a KPI for your business. They may still be worth recording as context, because context explains movement in the numbers that do matter, but context belongs further down the page and it does not get discussed first.
Now take cost per qualified enquiry by campaign. That rose from a comfortable number to an uncomfortable one. What do you do? You look at the search terms, you pause what is wasting money, you move budget to the campaign that is still working. The number changed a decision, so it is a KPI.
Agreeing this list in advance is what makes reporting honest. When nobody agreed beforehand, every month becomes an exercise in finding whichever number went up. That is the mechanism behind the complaint people make about agencies going quiet or reporting vanity metrics: not dishonesty so much as the absence of a definition anybody committed to.
Vanity metrics and decision metrics, side by side
The left column is what usually gets reported. The middle column is what would actually tell you something. The right column is the decision it changes, which is the reason the middle column exists.
Open the full comparison: Vanity metrics and decision metrics, side by sideHide the full comparison: Vanity metrics and decision metrics, side by side
| Item | Commonly reported | Worth deciding on | The decision it changes |
|---|---|---|---|
| Organic search | Keyword rankings tracked | Non-branded organic sessions that reached a contact or booking page | Which pages get rewritten or merged next |
| Google Business Profile | Profile impressions | Calls, direction requests and messages from the profile | Whether categories, services and photos need work |
| Paid search | Clicks and impression share | Cost per qualified enquiry, by campaign | Which campaign gets the next budget increase and which gets paused |
| Website | Total sessions | Enquiry rate by landing page | Which page gets a clearer call to action first |
| Social | Follower count | Profile visits, saves and direct messages | Which format and subject to make more of |
| Open rate | Clicks, replies and bookings per send | Whether the list is still worth mailing and how often | |
| Reviews | Star rating on its own | New reviews in the last ninety days and response rate | Whether the request process is actually running |
Open rate deserves a specific warning. Mail privacy features that pre-load images have made opens unreliable as a measure of whether an individual read anything, so a rising open rate can mean nothing at all.
One number per stage, not fifteen per channel
A workable set covers the whole path, with a single measure at each point rather than a wall of them everywhere.
The value of one number per stage is that it localises the problem. If demand is up and arrivals are up but enquiries are flat, the site or the offer is the problem and no amount of extra traffic will fix it. If enquiries are up and bookings are flat, the problem is the phone, the follow-up or the schedule, and it is not a marketing problem at all.
This is also how you avoid the most expensive mistake in marketing spend, which is buying more of the thing that already works while a later stage quietly leaks. Doubling traffic into a page that converts poorly doubles the waste.
Five numbers is enough for a monthly review for most businesses. If your report has thirty, nobody reads any of them, and the thirty exist to look thorough rather than to be used.
Leading indicators move first, lagging indicators settle the argument
Revenue from new customers is a lagging indicator. It is the number that matters most and the number that tells you last, because it arrives after the enquiry, the consultation, the treatment plan and the invoice. In a business with a long decision cycle it can lag by months.
Published pages, reviews collected, non-branded impressions and enquiry volume are leading indicators. They move sooner, and they are how you tell whether work is on track before the lagging number has anything to say. They are also easier to fake, which is exactly why they cannot be the whole report.
Read the full breakdown: Leading indicators move first, lagging indicators settle the argumentHide the full breakdown: Leading indicators move first, lagging indicators settle the argument
A defensible set uses both. Leading indicators for the monthly conversation, so problems surface while they are still cheap. Lagging indicators for the quarterly one, so the programme is ultimately judged on money rather than activity. If a report contains only leading indicators, that is worth asking about.
Set a baseline before anything starts, and write it down. Without a recorded starting point, every later discussion becomes an argument about what the situation was like before, and memory always flatters whoever is speaking.
How to choose yours in an afternoon
This works for a single location practice and for a multi-site business. The questions are the same, only the numbers get bigger.
Write down the outcome in money
Not leads, not traffic. New patients, signed cases, booked installations, subscriptions. Whatever your business actually sells, and whatever the average one is worth to you over a year. If nobody knows that figure, finding it is the highest value hour available this month.
Find the last step before the money
For most service businesses it is a booked appointment or a returned call. That step is your primary KPI, because it is the closest thing to revenue that marketing can genuinely influence and measure within a reasonable window.
Pick one measure per earlier stage
One for demand, one for arrivals, one for enquiries. Resist adding a second. The purpose is to locate a problem quickly, and more numbers make that slower rather than faster.
See the remaining steps: How to choose yours in an afternoonHide the remaining steps: How to choose yours in an afternoon
Check each one can be counted reliably
If a number depends on somebody remembering to tick a box in a busy front office, it will be wrong within a fortnight. Prefer measures that a system records by itself, and accept a slightly worse measure that is captured consistently over a better one that is not.
Record the baseline and the target window
Note where each number stands today and when you would expect to see movement. Organic and local search compound over weeks and months rather than days, and a target set for four weeks out invites the wrong conclusion at exactly the wrong time.
Fix the review cadence, then leave them alone
Monthly for the leading indicators, quarterly for the money. Changing the KPI set every month is how a programme avoids ever being judged, and it is worth noticing when it happens.
The output is a single page anybody in the business can read. If it needs explaining every month, it is the wrong set.
Signs your reporting has become theatre
None of these prove bad work. All of them are worth asking a direct question about.
- The report leads with impressions, reach or ranking counts and gets to enquiries on the last page, if at all.
- The set of metrics quietly changes whenever a number goes the wrong way.
- Nobody can say what the baseline was before the work started.
- Enquiries are counted but nobody knows how many became customers.
- Every channel is described as performing well while the phone rings less than last year.
- The report is beautiful and nobody in the business has ever made a decision from it.
- Branded and non-branded search are merged, which lets existing demand disguise itself as new demand.
That last one is worth checking yourself. Search Console will split queries containing your business name from everything else, and the two behave completely differently.
What companies are actually spending
The CMO Survey, run by Duke University's Fuqua School of Business, surveyed 308 US marketing leaders in January 2026 and reported where marketing budgets currently sit.
SourceThe CMO Survey, February 2026 (Duke Fuqua, n=308 US marketing leaders)
Treat that as orientation rather than a rule. It is an average across company sizes and sectors, and a new practice buying its first patients and an established one defending a full book are not the same problem.
Cost per outcome, by channel, every month
Once you can divide what a channel cost by what it produced, most budget debates answer themselves.
Include the fee, not just the media spend. A channel that looks cheap on ad spend alone can be the most expensive thing you do once management is counted, and a channel that looks expensive can be the cheapest per booked job.
Give slower channels a fair window. Organic search, content and reviews build over months, so judging them on a thirty day cost per outcome will always flatter advertising and always under-rate the work that compounds.
And keep one honest caveat in view: attribution is imperfect. Consent choices, shared devices and recommendations that arrive by word of mouth all blur the picture. Use cost per outcome to compare channels against their own history rather than to declare a winner to two decimal places.
Where this leads next
Three questions that follow directly from choosing what to measure.
How do I track where my leads come from?
Your KPIs are only as good as the source data behind them. This is the plumbing underneath the report.
How do I track where my leads come from?What is a good conversion rate?
Why the honest answer is a range that depends on intent, and what to compare yourself against instead.
What is a good conversion rate?How much should I spend on marketing?
Budget set against what a customer is worth, rather than against a percentage somebody read online.
How much should I spend on marketing?How to choose a marketing agency
The reporting, ownership and exit questions worth asking before you sign anything.
How to choose a marketing agencyWant a report you would actually use?
We will look at what you are being shown today and tell you which lines are decisions and which are decoration.
Straight answers.
How many KPIs should a small business have?
Four or five. One for demand, one for arrivals, one for enquiries, one for booked work, and a cost figure that ties them to money.
Any more and nobody holds them in their head, which means nobody notices when one moves.
Are rankings a KPI?
Rarely. Rankings vary by device, location and personalisation, so the number you see is not the number your customer sees. They are useful as a diagnostic when something changes, and poor as a headline measure.
Non-branded impressions and clicks from Search Console carry more information and are harder to argue with.
What is the difference between a KPI and a metric?
Every KPI is a metric. Most metrics are not KPIs. The distinction is commitment: a KPI is the handful you agreed to be judged on before the work started, and everything else is context that explains their movement.
How often should KPIs be reviewed?
Monthly for the leading indicators, quarterly for revenue. Weekly reviews of a slow channel mostly generate anxiety, because normal week to week variation on a small local site looks exactly like a trend until you have a few months of it.
Should a marketing agency set my KPIs?
They should propose them and you should agree them, in writing, before work starts. A set of measures chosen after the first quarter tends to be a set of measures chosen to look good.
If the proposed KPIs contain no cost per outcome and no booked work, ask why not.
What if I cannot measure booked work?
Then measure the closest reliable step and be explicit that it is a proxy. A qualified enquiry logged by whoever answers the phone is imperfect, and it is still far better than counting form fills that include recruiters and suppliers.
Improving what you can capture is usually a better investment than arguing about which metric is theoretically superior.
Where this comes from.
Primary documentation and published research behind the guidance on this page.
- Google Ads Help: About conversion measurement (opens in a new tab)
- Google Search Console Help: About Search Console (opens in a new tab)
- Google Analytics Help: Engagement rate and bounce rate in GA4 (opens in a new tab)GA4 defines bounce rate as the inverse of engagement rate, not as a single-page visit.
- FTC: Advertising and marketing basics for business (opens in a new tab)
Talk to the team
A short call, a look at how the business currently shows up, and a straight answer on what we would do first.