Article
Operations KPIs: on-time, cycle time, and rework
An operations board that reports units and hours describes how busy the department was. It says nothing about whether the customer got what was promised on the day it was promised, or what it cost to fix the parts that went wrong. Five rows answer that weekly. Seven more, once the month is closed, say what the answer cost you.
By the Key Performance Indicators editorial teamPublished Updated 10 min read
01
What is an operations team accountable for?
Delivering what was promised, when it was promised, without doing it twice. Those three clauses are the whole department. On-time delivery covers the promise, cycle time covers the speed, and rework covers the cost of getting it wrong the first time. Everything else an operations lead measures is a way of explaining one of those three.
Keeping it to three clauses matters because operations is where a board grows fastest. Every incident produces somebody suggesting a new row, and a year later there are thirty, nobody reads any, and the meeting is a status update with numbers attached.
The three trade against each other, which is the conversation worth having weekly. A team can hit its dates by cutting corners, and rework hours show that inside a month. It can eliminate rework by taking twice as long, and cycle time shows that sooner.
02
Which five operations numbers belong on the weekly board?
On-time delivery, median cycle time, rework hours, capacity booked four weeks out, and jobs already past their promised date. Four of them are readable off the job list on a Friday without anybody building a report, and the fifth is a phone call somebody should be making today rather than a number to admire.
Read on-time delivery and jobs past due together or you will misread both. On-time delivery is a percentage of what finished this week, so a team can hold it high while a handful of impossible jobs sit unfinished outside it. The overdue count is where those jobs live.
Jobs delivered on the promised date
Jobs completed on or before their promised date, divided by all jobs completed that week. Lagging by a few days, weekly, owned by the operations lead. Band is structural: the percentage you are prepared to promise a customer out loud. Gamed by moving the promised date, so lock the first date given.
02Median cycle time
Median working days from order accepted to job delivered. Median rather than mean, since one large job will drag an average anywhere. Lagging, weekly, owned by the operations lead. Band from your own trailing 13 weeks. Gamed by starting the clock at scheduling rather than at order.
Rework hours
Hours spent redoing work already recorded as finished. No formula, just a booking code. Lagging, weekly, owned by the operations lead. Band from your own trailing 13 weeks, driven downward. Gamed by booking the fix as new work, which is why the code has to be easy and never used against an individual.
Capacity booked for the next four weeks
Committed hours in the next four weeks divided by available hours. Leading, weekly, owned by the scheduler. Band from your trailing 13 weeks, with a ceiling as well as a floor: fully booked means every late job pushes the next one. Gamed by counting unconfirmed work as committed.
05Jobs past their promised date
Open jobs whose promised date has gone by, counted rather than valued. Leading, weekly, owned by the project manager. Structural target of zero, which almost nobody hits. Gamed by quietly rescheduling on a Friday afternoon, so read the count of dates changed next to it for the first few months.
03
What belongs in the monthly operations review?
Cost and quality, which both need a closed month behind them. Billable utilization, gross margin on delivered work, cost per unit, first-time-through rate, complaints per hundred jobs, change orders per job and supplier on-time delivery. Read weekly they swing on whichever large job happened to finish, which teaches a team to explain rather than act.
Give the review one hour and go job by job on the worst three rather than department by department on the averages. An average margin tells you the month was acceptable. The three jobs that lost money tell you which quote template, which crew or which customer to do something about.
Billable utilization
Hours booked to customer work divided by hours available. Lagging, monthly, owned by the operations lead. Band from your own trailing 13 weeks, with a ceiling. Gamed by booking rework and internal fixing to the customer code, which is why this row only means anything read beside rework hours.
02Gross margin on delivered work
Revenue on delivered jobs less the direct cost of delivering them, over that revenue. Lagging, monthly, owned by the operations lead with finance agreeing what is direct. Band from your trailing four quarters. Gamed by leaving supervision out of direct cost, which lifts the percentage and changes no quote.
03Cost per unit delivered
Total delivery cost for the month divided by units, jobs or tickets delivered. Lagging, monthly, owned by the operations lead. Band from your own trailing four quarters. Gamed by changing what counts as a unit halfway through a year, so define the unit once and note the date you defined it.
04First-time-through rate
Jobs delivered with no correction, rework or callback, over all jobs delivered. Lagging, monthly, owned by the quality lead. Band from your own trailing 13 weeks. Gamed by a generous definition of correction, so write down what counts before the first month rather than during the first bad one.
05Complaints per hundred jobs
Customer complaints logged, scaled to a hundred delivered jobs so a busy month is comparable to a quiet one. Lagging, monthly, owned by the operations lead. Band from your trailing four quarters. Gamed by resolving a complaint informally and never logging it, which is the usual reason this row looks calm.
Change orders per job
Change orders raised divided by jobs live in the month. Lagging, monthly, owned by the project manager. Band from your own trailing four quarters. Gamed by bundling six changes into one order, so count changes rather than paperwork and read the row against margin on the same jobs.
Supplier on-time delivery
Supplier deliveries arriving on or before the agreed date, over all deliveries received. Lagging, monthly, owned by the buyer. Band is structural, at whatever you agreed in the terms. Gamed by accepting a revised date as the agreed one, which converts a supplier problem into your own scheduling problem.
04
Where do operations targets come from?
Two sources, and mixing them up is where operations targets go wrong. A promise date is a commitment you chose, so on-time delivery has a structural target of the percentage you are willing to sign up to publicly. Everything else, cycle time and rework included, comes off the median of your own last thirteen weeks.
Set the on-time figure deliberately rather than optimistically. Ninety-five percent and ninety-nine percent are different businesses: the second usually means quoting longer lead times, holding more stock or refusing work, and all three of those are decisions for the leadership team rather than for the scheduler. Decide, then write the number on the board.
For the historical rows, take the median of thirteen weeks rather than the average. Operations data is full of one enormous job, and a mean lets that job set the target for a quarter.
05
Which operations numbers should you drop?
Units produced, hours worked, tickets closed and anything counting effort rather than delivery. A busy shop and a productive one look identical in an hours column. Total jobs completed has the same fault: it rises with a good month for sales and falls with a quiet one, and neither movement tells an operations lead to do anything.
Tickets closed deserves a specific warning, because it is the row most likely to make service worse. Closing is the one step a team fully controls, so a target on it produces tickets closed early, reopened by the customer, and closed again. First-time-through rate catches that.
Dropping a row means taking back the weekly slot rather than deleting the data. The system keeps counting units. The leadership team stops spending four minutes a week reading a number nobody was going to act on.
Vanity metric
Units, hours and closures: counts that rise with activity and say nothing about the promise you made.
02Examples by department
Ten operations measurables with owners and starting goals, plus the lists for every other department.
03Dashboard
Why a screen of operations charts is a different object from a weekly list somebody answers for.
06
How is the on-time delivery row written out?
On-time delivery is the row worth writing out, because two people in the same room will count it differently within a month. Five fields settle it: name, owner, definition, band, source. Most of the work is in the definition, and specifically in whose date counts as the promise once a customer has changed the scope twice.
Name: jobs delivered on the promised date. Owner: operations lead. Definition: jobs marked delivered this week whose delivery date is on or before the date confirmed to the customer at order, where a customer-requested scope change resets the date once and only once, with the reset recorded. Band: 93 to 97 percent, a standard this leadership team agreed rather than an industry figure. Source: the job system export, pulled Friday morning by the operations lead.
Teams running EOS® read that row out on the weekly Scorecard in about ten seconds and drop anything under the band into the issues list with a name against it. The framework is not the point. One owner, a definition nobody can reinterpret in the meeting, and a rule for a bad week agreed before the bad week arrives.
Scorecard
The weekly list these rows are written to be read out in, and why two operations rows on it is usually plenty.
02Building a Scorecard
Assembling the first version, how long it takes to settle, and a picker that drafts one by industry.
03Metric owner
Why the scheduler owns capacity and the operations lead owns the date, even though both sit in one department.
Start with a real question
Your operations board counts activity rather than promises kept?
Send us the rows your operations meeting reads each week and the dates you quote customers. We will mark which numbers describe a promise, which only describe how busy the team was, and where the on-time figure is being held up by a definition nobody agreed. One reply from a person, inside a business day.
A person reads every one.