KPIs & Reporting for SMBs: From Anecdote to Signal

Start here

Five numbers, one per critical function, reviewed at the same time every week — each with an owner, a target and a response agreed before it ever goes red. That is the whole system. This guide shows you how to choose the five, install them in 30 days, and keep them honest for the quarter it takes them to become real signal.

Who this is for. CEOs and business owners with a leadership team of at least two people besides themselves. If you are the only person who can tell how the business is doing, this guide is about changing that.

What you will have at the end of 30 days:

• Five KPIs that each pass the decision test
• A written definition, target and red / amber / green threshold for each
• One named owner per number, and a pre-agreed response for every red
• A weekly leadership meeting that opens with the scoreboard and spends its time only on the reds
• A spreadsheet scoreboard that takes minutes, not hours, to update each week

Every worked example uses illustrative numbers. Swap in your own — the method is the asset, not the figures.

1. The anecdote problem

Most leadership teams don’t lack data — they lack an agreed handful of numbers that trigger action. Without them, the business is run on whoever spoke last and loudest.

You will recognise the symptoms. One unhappy customer email reshapes the week. A salesperson’s confidence about “a big one closing soon” stands in for a pipeline. Cash surprises arrive at month-end. And the CEO becomes the human dashboard: the only person holding enough context to know whether things are fine.

The three reporting setups that don’t work

The 40-tab dashboard. Someone built a report with every metric the software could produce. Nobody owns any of it, so nobody acts on it. Within a few months it is opened only before board meetings.

The monthly P&L as the only report. Accurate, but late — it describes a month that ended weeks ago. By the time it shows a problem, the causes are two months old.

The gut. Experienced CEOs often have excellent instincts. But instinct doesn’t delegate, doesn’t scale past one person, and can’t be challenged by the team.

Anecdote versus signal

Signal has a precise meaning here: a number that, when it moves, tells you something has changed and tells someone what to do about it. A number that only describes is noise, however accurate it is.

The goal is not more reporting. It is less reporting, with more consequence.

2. The five-number scoreboard

One number for each of the five functions that can quietly kill a growing business: sales, pipeline, delivery, cash and team. Five is enough to see the whole business on one screen and few enough that every number gets discussed when it goes red.

The rules

• One number per function. Not three sales numbers. The scoreboard is a smoke alarm, not a diagnosis.
• One owner per number. A named person, never a department or “the team”.
• A written target and thresholds for green, amber and red.
• A response agreed in advance for when it goes red.
• Reviewed at the same time every week, in the same meeting, in the same order.
• Leading indicators where possible. You want numbers that warn you, not numbers that confirm what already happened.
• Unchanged for at least a quarter. Swapping numbers every month destroys the trend line, and the trend line is where signal lives.

Two layers, not one

The five numbers sit on top. Underneath, each owner keeps their own working dashboard of three to eight diagnostic metrics — conversion rates, ticket backlogs, debtor days — that explain why their number moved.

The leadership team looks at the top layer every week. It goes into the second layer only when a number is red. This is what keeps the weekly meeting short and the owners accountable: the owner is expected to arrive already knowing why.

If you find yourself wanting a sixth number, ask which of the five it would replace. Usually the answer is that it belongs in an owner’s second layer.

3. Choosing the right KPIs

The right KPI is the one that changes a decision. Everything else is reporting theatre — numbers that look diligent and change nothing.

The decision test

For every candidate number, answer three questions in writing:

1. Who owns it? One name.
2. What is the target? A number, not “improve” or “grow”.
3. What will we do differently when it misses? A specific action, not “look into it”.

If you cannot answer the third question, drop the number. It may still be interesting. It does not belong on the scoreboard.

Then apply four practical filters: it can be measured weekly; the owner can influence it within four to eight weeks; collecting it is cheap, minutes not a day; and it is hard to game, because hitting it requires doing the real work.

Worked example: testing five popular candidates

Candidate menu by function

Write a definition card for each number

Most scoreboards fail on definitions, not choices. “Qualified” means three different things to three people. Before week one, write one card per KPI: the name; the exact definition; the source report or sheet; the owner and who enters it; the target and thresholds; the direction; and the red response.

If two people can calculate the number from the card and get different answers, the definition isn’t finished.

4. Leading indicators, targets and thresholds

A lagging number tells you what happened; a leading number tells you what is about to happen while you can still change it. Put the most leading number you can honestly measure on the scoreboard, and keep the lagging outcome for the monthly review.

Follow the chain back

Every outcome sits at the end of a chain. For a typical B2B sale: conversations held → proposals sent → deals won → revenue → cash collected.

Cash collected is the outcome you care about most, and the last to move. Proposals sent moves first. If proposals drop this week, revenue drops in a month or two — and you can act today. Choose the earliest link that you can define tightly and that the owner can influence.

Then check the link is real. After 8–12 weeks of data, look at whether the leading number moved before the lagging one. If it didn’t, you’ve picked activity rather than a true leading indicator. Swap it at the next quarterly review.

Setting the target: work backwards from the plan

The simplest reliable method converts an annual outcome into a weekly leading target. Illustrative example: a business needs £2.4M of new revenue this year. Its average deal is £30,000 and it wins one proposal in four.

• Deals needed: £2.4M ÷ £30,000 = 80
• Proposals needed: 80 ÷ 25% = 320
• Proposals per week: 320 ÷ 50 working weeks = 6.4, so the target is 7

That single number — seven proposals a week — now carries the annual revenue plan. If the business sends four, everyone knows what it means.

Two other methods are useful where there’s no plan to work back from. From baseline: take the median of the last 8–13 weeks and set the target modestly above it — good for delivery and team numbers. From capacity: what the team can deliver at full, sustainable load — good for utilisation and throughput.

Red, amber, green

Define the bands before the first meeting, not during it.

For the proposals example: green is 7 or more, amber is 5–6, red is 4 or fewer. For “lower is better” numbers such as overdue receivables or defect rate, reverse the direction and write it on the definition card.

Set the tolerance band from normal variation. If the number routinely swings 15% week to week, a 5% amber band will turn it red every other week and the team will stop listening. One data point is noise; act on a red, or on three consecutive weeks moving the same way.

5. Owners and the red response

A number without an owner is a weather report. A red without a pre-agreed response is a debate. Settle both before the scoreboard goes live, and the weekly meeting stops being an argument and starts being a review.

What the owner actually does

Ownership does not mean the owner personally produces the result. It means the owner makes sure the number is entered correctly before the meeting; reads it out in one line — number, target, colour, direction; arrives already knowing why a red is red; brings the first response, already started where possible; and raises it as an issue when the fix needs someone else’s time or money.

Ownership is about visibility, not blame. If a red number gets someone shouted at, the numbers will soon start arriving green — and wrong. The CEO’s reaction to the first few reds sets the culture for the whole system.

The escalation ladder

Week 1 red — owner response. The owner diagnoses the cause within 48 hours and runs the agreed first response. It is noted in the meeting, not solved there.

Week 2 red — leadership issue. The number goes to the top of the issues list. The team agrees a fix, an owner for the fix and a date.

Week 4 red — structural priority. One month of red means something structural is wrong. It becomes a quarterly priority with resources, or the target is formally reset with reasons.

The ladder turns “we should look at that” into a fixed timetable: nothing stays red for a month without becoming someone’s resourced priority.

A red playbook for the five numbers

Write your own version before week one. The specific actions matter less than the fact that they were agreed calmly, in advance, rather than invented under pressure in the room.

6. Wiring the numbers into the rhythm

A scoreboard nobody reviews on schedule decays within weeks. The numbers only become signal when they are looked at at three distances — weekly for points, monthly for trends, quarterly for whether they’re still the right numbers.

Weekly: points. The scoreboard opens the meeting. Two to five minutes of green, amber, red — then the meeting’s time goes only to the reds.

Monthly: trends. Individual weeks are noisy; four or more weeks show direction. Once a month, put the 13-week trend for each number on screen. Look for slow drift: a number that has been green but falling for six weeks deserves attention before it crosses into amber. This is also where lagging outcomes are checked against the leading numbers that were supposed to predict them.

Quarterly: the right numbers? Ask of each KPI: did it change any decisions? Did it warn us early? Is it still tied to this quarter’s priorities? A business that has fixed its pipeline problem may need its pipeline number replaced by a delivery or capacity number. Change numbers here, and only here.

Put it in the calendar now. Book the weekly meeting at the same day and time for the next 13 weeks. Book the monthly review after the books close. Book the quarterly review in the last two weeks of the quarter. A rhythm that depends on someone remembering to schedule it will not survive the first busy month.

7. Running the weekly scoreboard meeting

The meeting has one job: find what is off track and assign it, fast. Sixty to ninety minutes, same day, same time, same order, starting on time whoever is missing.

If you already run EOS, this is the shape of the Level 10 meeting, and the five numbers are your scorecard. The method works the same way in any operating system — what matters is the discipline, not the brand.

The read-out script

Each owner says four things in one breath: number, target, colour, and — only if it isn’t green — what happens next.

“Qualified pipeline created: £410K against £450K. Amber. Second week amber, so I’m putting it on the issues list.”

“On-time delivery: 96% against 95%. Green.”

Anything longer belongs in the issues segment. The facilitator’s job is to stop discussion during the read-out, politely and every time: “Let’s take that to issues.”

Rules that keep it honest

• Numbers are entered before the meeting, by a fixed time. No updating live in the room.
• No number, no excuse. A missing number shows as red. It means either the work or the reporting failed, and both matter.
• Reds are solved in issues, not in the read-out. Name the cause, decide, assign an owner and a date.
• “The number is wrong” becomes its own issue, with an owner to fix the definition or the data source by next week.
• Actions are written down and read back at the end, then reviewed next week.

Expect the first three or four meetings to run long and feel awkward. People defend their numbers, definitions get questioned, and some reds turn out to be data errors. That’s the system working: it is surfacing disagreements that were always there but hidden. By week six, a well-run scoreboard segment should take under five minutes.

8. Dashboards without the BI project

You don’t need a data warehouse. A disciplined spreadsheet — or the dashboard module of the project system you already run — beats an abandoned BI tool every time. Build the simplest thing that gets five accurate numbers in front of the team every week.

The four-tab spreadsheet

1. Definitions — one row per KPI, holding the definition card: name, definition, source, owner, target, amber floor, red line, direction.
2. Entry — one row per week, one column per KPI. Owners type their number here and nowhere else.
3. Scoreboard — the last 13 weeks, laid out for the meeting, with automatic colouring.
4. Trends — a line chart per KPI with the target drawn across it, for the monthly review.

Automatic colouring

Give each KPI’s cell a status formula that reads its thresholds from the Definitions tab. For a “higher is better” number, with the target in D2 and the amber floor in E2:

=IF(B5>=D2,"Green",IF(B5>=E2,"Amber","Red"))

For a “lower is better” number, flip the comparisons:

=IF(B5<=D2,"Green",IF(B5<=E2,"Amber","Red"))

Then add conditional formatting so the cells colour themselves. Keep the thresholds in the Definitions tab, never typed inside formulas — so the quarterly review can change a target in one place.

Automate two or three numbers first

Don’t try to automate everything. Start with the numbers that already live in a system and are tedious to copy: cash from your accounting software or bank feed; sales and pipeline from a saved CRM report; delivery from the project or job management system you already use. Check what your specific tools support before building anything. Manual entry for the remaining numbers is fine at this scale.

If your team already lives in a work-management platform, build the scoreboard there instead. A number that sits where people already work gets updated; a number in a separate tool gets forgotten.

Consider a BI tool only when the spreadsheet starts to strain: several entities or currencies, more than a handful of people entering data, or second-layer dashboards needing daily refreshes. Even then, the five-number scoreboard stays the front page.

9. Starting sets by industry

These are starting points, not answers. Run each one through the decision test before adopting it — the right five depend on where your business is constrained, not on your sector label.

Adjust for your constraint. The five functions are a default, not a law. If one function is clearly your bottleneck, it can earn a second number at the expense of the strongest function. Make that choice deliberately, at a quarterly review, and write down why.

Watch the ratios. Some of the most useful numbers are ratios: utilisation, labour cost as a share of sales, on-time percentage. They’re harder to game than raw counts because they cannot be improved simply by adding effort or headcount.

10. The ten ways scoreboards fail

Scoreboards rarely fail because the numbers were badly chosen. They fail because the behaviour around them erodes. Check this list at every monthly review.

Failures 7, 8 and 10 are the dangerous ones, because the spreadsheet looks perfectly healthy while they happen. They are leadership behaviours, not reporting problems — which is why fixing them usually needs someone with the standing to say so in the room.

11. The 30-day install plan

Four weeks from blank page to a live scoreboard, then one quarter before you trust the trend lines.

Week 1 — Choose

• Book a two-hour session with your leadership team
• List every number anyone currently reports or asks for
• Run each through the decision test; cut anything that fails question three
• Choose one number per function: sales, pipeline, delivery, cash, team
• Assign one named owner to each

Week 2 — Define and baseline

• Each owner writes the definition card for their number
• Each owner pulls the last 8–13 weeks of history, where it exists
• Set targets: worked back from the plan, or from the baseline
• Set amber and red thresholds from normal week-to-week variation
• Draft the red response for each number

Week 3 — Build and rehearse

• Build the four-tab spreadsheet, or the equivalent in your existing system
• Load the historical weeks so the trend lines start full
• Set the weekly entry deadline and put it in every owner’s calendar
• Book the weekly meeting for the next 13 weeks, plus monthly and quarterly reviews
• Run one rehearsal of the scoreboard segment and time it

Week 4 — Go live

• Run the first live weekly meeting with the scoreboard first on the agenda
• Send every red to the issues list with an owner and a date
• Log every definition dispute as an issue and fix it by the next week
• After the meeting, note what took too long and adjust the format

Day 90 — First quarterly review

• Check whether each leading number moved before the outcome it was meant to predict
• Ask whether each number changed at least one decision this quarter
• Keep, swap or redefine — then freeze for the next quarter

The common shortcut is to skip week 2. Don’t. Definitions and thresholds agreed in advance are what stop the first live meeting from turning into an argument about whose spreadsheet is right.

12. Health check — and when to bring in help

Run this at day 90 and every quarter after. Score one point for each honest yes.

1. We have exactly five scoreboard numbers, one per critical function.
2. Every number has one named owner, and that owner can explain its definition.
3. Every number has a written target and red, amber and green thresholds.
4. At least two of the five are leading indicators.
5. The numbers are entered before the meeting, every week, without chasing.
6. The scoreboard segment takes five minutes or less.
7. Every red in the last month produced a named action with a date.
8. No number has been red for more than four weeks without becoming a priority or a reset target.
9. People raise reds early, without fear of how the CEO will react.
10. At least one significant decision this quarter was made because of a number, not an opinion.

Where this gets hard

Building the spreadsheet is the easy part. The hard part is behaviour: owners who don’t want to own a red, a leadership team that argues with the numbers rather than acting on them, and a CEO who — with the best intentions — keeps stepping in to fix things rather than holding the system.

If your score is under eight after two honest attempts, the problem is rarely the KPIs. It is usually the operating rhythm around them, and that is very hard to change from inside, because the person who would have to change it is often the person running the meeting.

Working with Markinly

Markinly provides embedded COO leadership to business owners across the UK, Europe and the US. We install scoreboards like this one as part of a wider operating rhythm — and then sit in the room until the behaviour holds.

Book a 30-minute scoreboard review. Bring your current numbers, or your health-check score. You’ll leave with the two changes that would make the biggest difference — whether or not we work together.