How counting rework can turn a weekly report into an early warning

Most weekly management reports in a technical business are built from what the job and finance systems already produce. In 2023, UK businesses scored 0.42 out of 1 on their use of key performance indicators, the weakest of the management practices the Office for National Statistics reports (ONS, 2024).

In technical service businesses specifically, this lands on the technical leader: the operations head or technical founder reading a weekly report of revenue, hours booked and jobs closed. Those numbers arrive on time and look precise. They describe last month well and say little about the weeks ahead.

This article argues that the measure which predicts trouble costs effort to collect, so it is the one left off the report. ISO published ISO 9001:2026 on 16 September 2026, giving every certified business a transition to plan and a reason to revisit what it measures (ISO, 2026). You get a way to find your missing measure, three operational excellence actions and a five-question check you can score today.

Tracking the easy number is a structural issue

Reports fill up with the numbers that systems produce for free, so the gap is built into how the report is made. Rework rarely has its own line. Three drivers keep it that way:

• Systems record what gets billed. Job and finance software captures hours and invoices by default, while a rework loop usually appears only as extra hours on the same job.
• KPI practice is the weakest management habit. UK businesses scored 0.42 on KPI use in 2023, against 0.80 for continuous improvement (ONS, 2024).
• Rework cost is spread across functions. Quality costs fall into four classic categories: prevention, appraisal, internal failure and external failure (ASQ). A 2026 survey of 175 organisations found the cost of poor quality is often measured only in selected functions, and many struggled to express it as a share of sales (Antony et al., 2026). In ASQ’s 2025 cost-of-quality survey, 31% of respondents said they fully understood how quality costs affect financial performance (ASQ, 2025).

In summary, the predictive measure is missing because it lives between systems and has to be counted on purpose.

Why does a predictive measure matter for delivery and cost?

A predictive measure matters because it moves the decision forward by weeks, while the cost of the problem is still small. Businesses with below-median management scores were four times more likely to use little or no analysis to support important decisions (ONS, 2024). A rising redo count in week two is a conversation about staffing or a supplier, while the same problem found in month-end delivery figures is a conversation with an unhappy customer.

The cost side is large and mostly unseen. Antony and colleagues found that internal and external failure costs often sit in double-digit ranges while prevention spending stays low, and that cost-of-quality information is used most at operational level and least in strategic decisions (Antony et al., 2026). Internal failure, such as a repeated test or a second visit, is the share a weekly count can catch before it becomes external failure: a return, a complaint or warranty work. A count of work done right first time brings that cost onto the page where decisions are made.

What goes wrong when the report measures what the system counts?

The report stays green while rework grows. What looks strong is the activity data: hours booked can even rise, because a second visit or a repeated test is billable time. What is weak is the warning. It arrives late, as a complaint or a missed date, after the cost has already been paid.

Researchers call the underlying mechanism surrogation: managers start treating a measure as if it were the goal it was meant to represent (Choi, Hecht and Tayler, 2012). In a small business the effect runs quieter, because the person who built the report is usually the one reading it. Work on Goodhart’s Law adds that an imperfect proxy, pushed hard enough, stops being a useful guide (Manheim and Garrabrant, 2019). We call the version found in small technical businesses the free number: the figure that earns its place on the report because the system produces it at no cost, and then slowly becomes the definition of a good week.

Three actions for technical leaders to put the predictive measure on the report

Trace — following one late job back to its first deviation

Start with your own work. Take the last job that ran late or came back, map its steps as they happened and mark the first point where it left the plan, such as a missing input or a return visit. Repeated across a few jobs, that first deviation shows which measure would have warned you, usually a count of something done twice.

Count — tallying first-time-right by hand before automating it

Count it by hand for four weeks before building it into any system. First-pass yield is the simplest form: the share of work that passes a process without rework or correction (ASQ Phoenix Section, 2024). Where a job passes through several steps, multiply the step yields to get rolled throughput yield, because four steps at 95% each finish right first time only 81% of the time. A tally sheet at each step costs a few minutes a day and tests whether the definition holds on the floor.

Measure	How to calculate it	Use it when
First-pass yield	Jobs right first time ÷ jobs completed	The job has one main step, such as a test or a visit
Rolled throughput yield	Yield of step 1 × step 2 × each further step	Jobs pass through three or more steps
Defects per unit	Defects found ÷ units checked	One job can fail in several ways
Internal failure cost	Hours and materials spent redoing work, in pounds	The board needs the problem in money

Choose — selecting the measure with the people who run the work

Choose the measure with the supervisors and engineers judged by it, and review it each week beside the outcome it predicts. Research on surrogation found that managers who took part in choosing the strategy were less likely to treat its measures as the goal, and that the act of choosing mattered more than discussion (Choi, Hecht and Tayler, 2013). Pairing the predictive measure with its outcome keeps both honest, because a gain in one should show up in the other within weeks.

How ready is your business to spot trouble early?

Score each question 0, 1 or 2, add the five scores, then read the band underneath.

Question	0	1	2
Does your weekly report show one measure of work done right first time?	No	Partly	Yes, documented
Can you trace last month’s most delayed job to its first deviation?	No	Partly	Yes, documented
Is time spent on rework recorded anywhere?	No	Partly	Yes, documented
Did the people who run the work help choose the measures?	No	Partly	Yes, documented
Is each predictive measure reviewed beside the outcome it predicts?	No	Partly	Yes, documented

0–3: Your report describes the past. Trace three late jobs this month before changing any measure. 4–7: The ingredients exist. Pick one predictive measure, tally it by hand for four weeks and place it beside its outcome. 8–10: Your report gives early warning. Check each quarter that the predictive measure still moves before the outcome does.

What to do next

Below roughly 20 jobs a month, one redo swings a percentage by five points, so a monthly count of redos beats a ratio. Small volumes are the weak spot. Delays that start outside the business, such as supplier lead times, belong on a separate watch list.

The measure that predicts trouble is usually the one that takes effort to collect, and the search starts with your late jobs. This week, take the last three jobs that ran late or came back, trace each one to its first deviation and note what you would have counted to see it coming.