KPI template for a team: choosing measures you can collect every week

A spreadsheet with eleven KPIs, last updated five weeks ago, sitting in a shared drive. Three columns have never had a number in them. One measure needs an export from a system that only one person had access to, and that person has moved on. The single row anybody still looks at is the one somebody types in from memory on the morning of the review.

That is the normal end state for a downloaded KPI template. The template was not the problem. The measures were chosen because they sounded important, not because anyone checked that the number could be collected on a Monday morning, in under ten minutes, by whoever happens to be doing the collecting that week.

A KPI template that survives gets built backwards. Start from what can be counted reliably and cheaply. Then work out which of those countable things actually moves the outcome the team is accountable for. Most published templates run the other way, starting from an impressive list of indicators, and that is why they go stale.

Why the sheet stops being updated

Three causes account for nearly every abandoned KPI sheet, and they are all mechanical rather than a failure of discipline.

Collection costs more than the number is worth. A measure that needs two exports, a lookup against a second list and a manual deduplication pass will take twenty minutes. Twenty minutes is affordable once. It is not affordable every week for the next year, especially when the person doing it also has delivery work. The cost is paid weekly, the benefit is paid at the review, and the two are never compared.

Nobody decides anything from the number. If a measure has been red for six weeks and nothing changed, the team has learned that the sheet does not drive decisions. After that, filling it in is theatre. The fastest way to kill a KPI is to watch it without ever acting on it.

The measure moves for reasons outside the team. Revenue, headcount, website sessions and the number of inbound leads all swing on things a delivery team cannot touch. These are worth reporting, but treating them as the team's performance indicator teaches people that the sheet is noise, because good weeks and bad weeks look the same regardless of what anyone did.

There is a fourth cause that is harder to see. The sheet lives somewhere other than the work. When the numbers sit in a document and the tasks sit on a board, in a tracker or in a chat thread, updating the sheet is a separate errand. Errands get skipped. Measures pulled from the place where work already happens, such as counts of cards in a column or items that crossed a date, get updated because the act of working updates them.

The test a candidate measure has to pass

Before a measure goes on the sheet, put it through four questions. A measure that fails any of them is not a KPI, whatever else it might be.

Can it be collected in five minutes, every week, by two different people, and give the same answer? If the answer depends on who ran it, the number will be argued about instead of used. Write the collection method down next to the measure, not in someone's head.

Can the team change it inside one cycle? A measure the team can only influence over six months is a goal, not a weekly indicator. Keep it, review it quarterly, and put something upstream of it on the weekly sheet.

Would a bad reading force a specific conversation? Name the conversation in advance. If a bad reading on "team velocity" leads to nothing but a shrug, it does not belong. If a bad reading on "items sitting in review for more than three days" leads directly to asking who is blocked, it does.

Is it hard to improve by gaming it? Every measure gets optimised once it is watched. Count of tasks closed rewards splitting work into slivers. Hours logged rewards slow work. Pair anything countable with a quality or outcome measure so the easy cheat shows up somewhere else on the same sheet.

Four to six measures that pass all four beat fifteen that half pass. The useful upper limit for a team of five to thirty is roughly six, because the review has to fit in the time available and each measure needs a named owner.

When nothing countable exists yet

Some of the things a team cares about most have no number attached. Handover quality, whether a client feels informed, whether the work is enjoyable to do. The temptation is to invent a score out of five and put it on the sheet, which produces a column of fours that never moves. A better move is to find the nearest countable proxy and accept that it is a proxy: the number of handovers that came back with questions, the number of promised updates that went out late, the number of weeks somebody worked past a Friday. Each of those is a count of an event rather than an opinion, so two people collecting it will agree, and the proxy can be argued about openly rather than hiding inside a rating.

Pairing the outcome with something upstream

The standard advice to mix lagging and leading indicators is right, but it is usually left abstract. In practice, the pairing is a mechanical exercise: for each outcome the team is judged on, name one thing earlier in the chain that the team controls and that can be counted this week.

Outcome the team is judged on	Upstream measure that is countable weekly	How it is collected
Delivery dates met	Items still open past their due date	Filter the board by date, count
Client renewals	Promised updates actually sent on time	Count sent against count scheduled
Fewer escalations	Items sitting in review longer than three days	Age of cards in one column
Predictable capacity	Items started but not finished at week end	Count in progress at a fixed time
Quality of releases	Items reopened after being marked done	Count of reopenings

The right hand column is the part that decides whether the template lives. Every entry there should be something a person can get by looking at one screen. When each measure needs a different application, a different login and a different definition, the collection ritual becomes the reason the sheet dies. Boards, dates and the discussion around them sitting in one place is what makes a weekly count take minutes rather than an afternoon, and it is worth checking what the features of whatever tool the team already uses can count without an export.

The fields a KPI row actually needs

Templates published as spreadsheet downloads often carry twelve or more columns: measure, definition, owner, target, actual, variance, status, trend, category, frequency, data source, comments, and a red amber green flag. Most of those columns will be empty within a month, and a half filled tracker loses trust fast.

Six fields carry nearly all of the value.

Measure. Written as a count of something, with the unit in the name. "Open items past due" rather than "delivery health".

Definition. One sentence that settles the edge cases. Does an item on hold count as open. Does a weekend count towards the three days. Without this the number changes meaning quietly as people rotate through collecting it.

Owner. One person, not a team. The owner is not responsible for the number being good. The owner is responsible for the number being there.

Target or direction. Some measures have a real threshold. Many only have a direction, and writing "lower is better" is more honest than inventing a target nobody believes.

This week and last week. Two numbers, side by side. A single current value tells the reader almost nothing. The comparison is where the information is, and a four week trend is better still if the sheet has room.

What was decided. One line, added at the review. This is the field that most templates leave out, and it is the one that proves the sheet is being used. Six weeks of blank decision lines is the signal to cut a measure.

What to leave out

Percentage complete on a measure is usually invented. Variance columns are arithmetic the reader can do. Category and frequency matter when there are forty measures across a company, not when there are five on one team. A colour flag is fine, but only if the rule that sets the colour is written down, otherwise it becomes a mood.

Where the sheet should live and who reads it out

A KPI template is a meeting artefact before it is a data artefact. The decision that matters is not the file format, it is when the numbers get read aloud and by whom.

The pattern that holds up for small teams is a fixed slot, same day each week, fifteen minutes, with the owner of each measure reading their own number and last week's number. Reading it aloud does something a dashboard cannot: it puts a person next to the figure. A measure that has been stale for three weeks becomes audible.

The second decision is the audience. A sheet for the team and a sheet for a client or a board are different documents, even when they share numbers. The team sheet can carry ugly internal measures such as reopened items or blocked time. The external one carries dates, scope and what happened, and it goes stale faster because nobody outside the team notices. Keeping them separate stops the team sheet being sanitised, which is the usual way honest measures disappear.

The third decision is what happens to the sheet when the team changes shape. Measures are tied to a way of working. When a team changes its process, most of its indicators stop meaning what they meant. Reviewing the measure list once a quarter, and expecting to retire one or two, keeps the sheet from becoming a museum.

Automation is worth less than it looks

Wiring a dashboard to pull numbers automatically removes the collection cost, which sounds like the whole problem solved. It also removes the person who noticed that the number looked wrong. Automated KPI dashboards tend to drift silently when a field is renamed or a filter stops matching. A cheap compromise is automatic collection with a human sanity check at the review, where the owner is expected to say whether the figure looks plausible. Pulling counts straight from a board through an API or an assistant still leaves that check in place.

What to change first

Cut the sheet down to the four to six measures that someone can actually collect in five minutes, and delete the rest rather than leaving empty columns as a reproach. Put a name and a last week figure against each one, and add a single line for what was decided. If the numbers currently come from three separate applications, moving the boards, dates and discussion into one place is what makes the weekly count survive, and Pinateca is free for up to five people and ten boards while the shape of the sheet is still being worked out.

Q1. How many KPIs should a small team track?

Four to six is the workable range for a team of five to thirty people. Each measure needs a named owner, a written definition and a slot in the review, and beyond six the review stops fitting in the time available. Company level dashboards carry more, but those are read by people who are not also doing the delivery work.

Q2. What is the difference between a KPI and a metric?

A metric is anything countable. A KPI is a metric someone has committed to act on, with a target or a direction and a person attached. The practical test is whether a bad reading forces a specific conversation. If it does not, the number is worth logging but it is not a key indicator.

Q3. Should a KPI template live in a spreadsheet or in a project tool?

A spreadsheet is fine for the template itself, as long as the numbers come from where the work happens rather than from memory. The failure mode is a sheet whose measures each need a different export, because the collection cost is paid weekly. Measures that can be read off a board, such as items past due or items stuck in one column, survive far longer.

Q4. How often should KPIs be reviewed and changed?

Read them weekly, change the list quarterly. Weekly reading is what keeps the numbers honest, because a stale figure becomes obvious when someone has to say it out loud. Changing the list more often than quarterly makes trends unreadable, and changing it less often leaves measures in place after the process they described has gone.

Q5. What should be done when a measure has been red for weeks?

Either act on it or take it off the sheet. A measure that stays red without any decision teaches the team that the review does not drive anything, which is what kills the whole sheet rather than just that one row. If the team genuinely cannot move it, the measure belongs on a quarterly outcome list instead of the weekly one.