The Monday KPI scorecard: 10 numbers, read in pairs, in 15 minutes

An owner's KPI scorecard should fit on one page, take 15 minutes to read, and contain 10 numbers arranged in five pairs, where each number in a pair keeps the other one honest. Booking rate sits next to speed to lead. Close rate sits next to presentation rate. Average ticket sits next to callbacks. Read one without its partner and you'll eventually reward the wrong behavior.

The other half of the method is knowing when a change is real. A tech's close rate falling from 45% to 35% in a week looks like a problem. On 25 proposals it is almost certainly noise, and the math below shows why. Most Monday meetings go wrong in exactly that gap: reacting to random weeks and missing slow trends.

Why a one-page KPI scorecard beats a dashboard

Dashboards are built to show everything. A scorecard is built to force a decision. The difference matters on a Monday at 7 a.m., when you have 15 minutes before the board fills up and a tech calls in with a sick kid.

Robert Kaplan and David Norton made this argument in their 1992 Harvard Business Review article on the balanced scorecard: managers should look at a small set of measures across several views of the business at once, so that improving one area doesn't quietly damage another¹. Andy Grove, who ran Intel, gave the practical version in High Output Management: pair indicators so that both an effect and its counter-effect get measured, so that pushing one number doesn't quietly wreck another².

That pairing rule is the whole design of the scorecard below. It is also the best defense against the problem the anthropologist Marilyn Strathern summarized in 1997, building on economist Charles Goodhart:

Put a single number on a whiteboard and people will move it. Put two numbers that pull against each other on the board and the only way to move both is to do the work better.

The 10 metrics, in five pairs

Here's the page. Every formula is written so anyone in the office would get the same answer from the same records.

Glossary definitions for several of these live at booking rate, close rate, average ticket, callback and speed to lead if you want to align your team on wording.

Pair 1: the front door (booking rate and speed to lead)

Booking rate tells you whether the office converts demand you've already paid for. The hard part is the denominator. Wrong numbers, vendors, job-status calls and "is my tech on the way?" aren't opportunities. If you can't yet tag calls by reason, use call-to-book rate and accept that it reads lower.

Speed to lead is the guard because a CSR team can keep booking rate high on the phones while web forms and texts sit until lunch. The best-known study on this is old but still clear: researchers writing in Harvard Business Review concluded that most companies weren't responding to online leads nearly fast enough, and that the first hour after a lead arrives matters most³. Their data came from business sales leads, not homeowners with a dead furnace, which if anything makes the case stronger for home services. Use the median, not the average, so a single weekend form answered on Monday doesn't swamp the number.

When booking rate slips, the first question: did the call mix change (a new marketing source, a storm, a price increase), or did the conversation change (a new CSR, a script change)? Listen to five unbooked calls before deciding.

When speed to lead slips: look at the hour of day. A slide usually comes from one gap, like lunch, the 4 to 6 p.m. window or Saturday morning, and the fix is a coverage rule, not a pep talk. The speed-to-lead post covers that playbook.

Pair 2: load (capacity booked and overtime share)

Capacity booked looks forward. Measure booked hours in the next 10 working days against the hours your techs are actually scheduled to work, after time off. That number tells you whether to turn marketing up or down this week, which no lagging revenue number can.

Overtime share is the guard because a full board can be built by quietly borrowing hours from the same people every week. Overtime is not bad in itself; in a heat wave it's the job. What you're watching is the trend. Rising overtime share with flat capacity booked means jobs are running longer than you scheduled them, or one tech is absorbing the whole on-call load. Both are early warnings for the turnover problem covered in why technicians quit.

When capacity booked slips: check membership visits due and open estimates first. Those are customers who already know you. Our arrival window capacity math post covers how much buffer to keep.

When overtime share climbs for three straight weeks: stop debating and start hiring, or move the overflow work type (often installs) to a fixed day.

Pair 3: selling (close rate and presentation rate)

Close rate is the sales number every owner already watches. Count a proposal as presented when the customer actually saw the options, and count it once even if it had three options. Good, better, best proposals are designed to raise this number honestly.

Presentation rate is the guard that almost nobody tracks. A tech who quietly stops presenting to "tire kickers" will post a beautiful close rate and a shrinking pipeline. The denominator is visits where a repair or replacement was possible: diagnostic and repair calls, and maintenance visits that turned up a problem. Leave out installs and warranty work.

When close rate slips but presentation rate holds: look at the options, not the tech. If every tech drops at once, the price book or the options moved.

When presentation rate slips: ride along. A falling presentation rate is almost always a confidence problem or a time problem (too many calls on the board), and you can't see either from a report.

Pair 4: quality of revenue (average ticket and callback rate)

Average ticket belongs on the page, but only for service and repair work. One replacement can equal 30 repair tickets, so a blended number tells you more about the install mix than about anyone's selling. Keep installs on their own line or leave them for the monthly review.

Callback rate is the guard. A rising ticket with rising callbacks means parts are being replaced on hope. Define it from the job record (a job booked at the same address within 30 days of a completed one, excluding planned membership visits) rather than from a box the tech ticks, which keeps the number from being negotiated. Note that it lags: this week's callbacks are mostly about work done in the last month. The full method is in our post on first-time fix rate, and the money side is in the real cost of a callback.

When average ticket slips: split it by job type before reacting. Most "drops" are a week with more maintenance visits.

When callback rate climbs: sort the callbacks by reason. Parts not on the truck is a stocking fix; misdiagnosis is a training fix; the same tech three times is a conversation.

Pair 5: cash (revenue invoiced and A/R over 30 days)

Revenue invoiced is the number most owners already know. Count invoices on their invoice date, before tax and after discounts, and don't mix it with "completed revenue" on job date. They're both useful; they just aren't the same number.

A/R over 30 days is the guard because cash, not revenue, is what small firms run short of. In the Federal Reserve's latest Small Business Credit Survey, fielded in late 2025, 50% of employer firms reported uneven cash flow as a financial challenge in the prior year and 54% reported trouble paying operating expenses such as payroll and rent; the survey counts collecting on receivables as part of uneven cash flow⁴. For residential service collected on site, this number should be small and boring. If it isn't, a commercial account or a builder is usually the cause. Track residential and commercial separately once you have both.

When A/R over 30 days rises: open the list, not the chart. Five invoices are usually most of it, and the right move is a phone call to each one this week.

When a change is real: the noise problem nobody explains

This is the part that changes how Monday meetings go. Rates built on small weekly samples swing a lot by chance alone.

Quality engineers have dealt with this for a century with the proportions control chart, or p-chart. NIST's engineering statistics handbook gives the limits as the average rate plus or minus three times the square root of p(1 − p) ÷ n, where p is the long-run rate and n is the sample size⁵. Inside those limits, a change is consistent with chance.

Run it for a tech whose long-run close rate is 40%:

Three practical rules fall out of that chart:

• Decide on four-week rolling numbers. A tech who presents 25 times a week has about 100 proposals in four weeks, which narrows the normal swing to about 15 points. Still wide, but you can see a real problem.
• Use the single week to spot a break, not a trend. A week outside the three-sigma limits is worth a question. A week inside them is not, however bad it feels.
• Watch for runs. Even inside the limits, several weeks in a row moving the same way is a signal. Many quality teams treat a long run on one side of the average as worth investigating; pick a rule (we'd use six weeks) and write it on the scorecard.

The same logic applies to booking rate, presentation rate and callback rate. It's why shop-wide numbers are usually fine to read weekly, while per-tech numbers need the rolling window. It also gives you a minimum sample: if a tech had fewer than about 20 proposals in the window, show a dash for their close rate and don't discuss it.

What good looks like (and why we won't give you a number)

Search for HVAC KPI benchmarks and you'll find plenty of confident targets. Almost none of them come from a published method with a known sample. Most are software vendor surveys or consultant rules of thumb. The government statistics that do exist (wages, employment, compensation costs) don't measure booking rate or close rate at all.

So the honest target for each metric is your own baseline:

1. 1

   Pull 13 weeks of history

   One full quarter, so a single holiday week doesn't dominate. If your system can't produce a number for a past week, leave it blank rather than estimating.

2. 2

   Take the median as the center line

   The median ignores the week the server was down or the week of the ice storm.

3. 3

   Compute the normal range

   Use the p-chart formula for rates. For dollar numbers like average ticket, use the range of the middle 80% of your weekly values.

4. 4

   Set a target only for the push metric

   Raise the target a little each quarter, and leave the guard metric as a limit that must not get worse.

5. 5

   Restate when definitions change

   If you change what counts as an opportunity, recompute the history with the new rule before comparing.

Once you have a year of your own data, compare against it, not against a number from a webinar.

What stays off the Monday page

Some important numbers move too slowly to read weekly. Reading them every Monday trains people to ignore them. Put these on a monthly page instead:

• Gross margin by job type, after the books close, from job costing. Weekly gross margin mostly measures when parts invoices arrived.
• Memberships sold, lost and net change. Renewals cluster by month; read them monthly with the renewal rate.
• Days sales outstanding and the full A/R aging.
• Performance pay as a percentage of revenue, once a pay period has been approved.
• Technician turnover and time to productivity for new hires.

Leading indicators on Monday, lagging ones monthly. That split is most of what Kaplan and Norton were arguing for.

The 15-minute Monday routine

The scorecard only works if the reading is the same every week. Here's the order we'd use:

1. 1

   Minutes 0 to 3: cash

   Revenue invoiced last week, A/R over 30 days. Open the list of overdue invoices and pick the five calls to make.

2. 2

   Minutes 3 to 6: load

   Capacity booked for the next 10 working days and last week's overtime share. Decide: marketing up, marketing down, or hold.

3. 3

   Minutes 6 to 10: front door and selling

   Booking rate with speed to lead, close rate with presentation rate. Shop-wide first, then any tech outside their normal range on the four-week view.

4. 4

   Minutes 10 to 13: quality

   Average ticket for service and repair, callback rate. Open the callbacks and read the reasons.

5. 5

   Minutes 13 to 15: three actions

   Write no more than three actions, each with an owner and a date. Last week's three get checked first next Monday.

Three actions is a deliberate limit. An owner who leaves the Monday review with eleven action items will finish none of them.

A worked example scorecard

Here's what one Monday might look like for a hypothetical 8-tech residential shop. Every number is invented to show the reading, not a benchmark.

The instinct would be to talk about the 31% close rate. The scorecard says the close rate is fine and the real story is capacity, speed to lead and callbacks. That's three actions, which is exactly the limit.

Getting the numbers without a spreadsheet afternoon

You can run this scorecard from a spreadsheet; the free scorecard above has the formulas built in. The cost is the Sunday-night export. The bigger risk is definitions drifting between whoever built the spreadsheet and whoever runs it next year.

In TradesWork, most of these metrics come from one metric dictionary of 34 KPIs, each with its definition, formula and date basis shown on hover, and every value opens to the invoices or jobs behind it. The owner morning brief can be set to arrive weekly on Mondays with the headline numbers and the top items that need doing. Honest gaps: there's no scheduled email of an arbitrary custom report yet, and presentation rate as defined here isn't a named metric, so you'd track it from the proposals list or the reports drill-down. A few dictionary metrics are also defined slightly differently from this page (speed to lead is an average there, and A/R over 30 days is a dollar amount rather than a share), so note which version your scorecard uses.

app.tradeswork.app · Reports with metric definitions