How 48 hours on the floor reveal how a technical business runs

Since 16 September 2026, ISO 9001 has asked leaders to promote quality culture and ethical behaviour, in clauses 5.1 and 7.3 of the new edition (ISO, 2026). Culture is hard to evidence in a procedure manual. It shows up in how work moves across the floor and in whom people call when a job stalls.

In a technical business of 10 to 50 people specifically, the operator who arrives as a new operations lead or interim manager inherits thin paperwork and busy colleagues. The real picture of the business sits on benches and in a handful of heads, which is where any work on operational excellence in small technical businesses has to begin.

This article argues that two days of structured observation, aimed at five specific points, give a more reliable first read than a week spent with reports. You get a two-day plan and a five-question scorecard for rating what you saw.

A misread business is a planning problem

A misread business is a planning problem: newcomers spend their first days where information is weakest. Three conditions make reports a poor guide here.

• Reports are thin. Use of key performance indicators was the lowest-scoring UK management practice in 2023, at 0.42 on a scale of 0 to 1 (ONS, 2024).
• Know-how stays with a few people. In workshops with 23 practitioners from 14 manufacturers, social factors such as staying indispensable made up 42% of the barriers to capturing know-how (Finkel and Wurster, 2026).
• Formal audits sample briefly. International guidance sets the initial certification audit for a business of 26 to 45 people at 4 auditor-days across two stages (IAF MD 5, 2019).

In summary, the floor holds better evidence than the files, so the first two days should be planned around it.

Why does floor observation matter for a first read?

Floor observation matters because the smaller the business, the less its management data can carry a newcomer. Businesses with 10 to 19 employees scored 0.51 on the ONS management practices index in 2023, against 0.68 for companies with 250 or more (ONS, 2024). The index measures practices such as monitoring and target-setting, the very tools a newcomer would otherwise lean on.

ONS also finds that higher labour productivity goes with higher management scores, with a 1% difference in productivity matched by 0.02 points on the index. That is a correlation and leaves cause open. It still justifies looking at practice directly when the data is thin.

What goes wrong when you read the reports first?

Reading the reports first makes a stretched operation look calm, because reports count the official queue and miss the one beside it. A job system may show 40 orders in progress while the benches hold 20 more tagged “waiting for customer” or “needs a second check”. Call this the second queue: work that has dropped out of the official count and sits on the floor waiting.

Little’s Law explains the damage. The average number of items in a system equals their arrival rate multiplied by the time each spends there (Little, 2011). Leave 20 of 60 jobs out of the count and the lead time the system calculates falls by a third, while customers still wait the full time. Because uncounted jobs leave the system figure, the dashboard can improve while the second queue grows.

A menu and a kitchen pass: two views of the same business

Imagine judging a restaurant by reading its menu. The menu tells you what the kitchen intends to cook, while the pass, where finished plates wait for a waiter, tells you how the kitchen is coping tonight.

A technical business has both. Its procedures and reports are the menu. Its benches and phone lines are the pass, and they change hour by hour as work arrives and stalls.

Three actions for operators to read a business in 48 hours

Plan — splitting two days between flow and exceptions

Plan the 48 hours before you arrive, because unplanned time drifts towards meetings. Spend day one following a single order from enquiry to invoice, and day two on the exceptions, such as rework, late jobs and interrupting calls. Writing the five observation points on one card keeps the second day from becoming a tour.

Count — walking the floor to size the second queue

Count physical work in progress yourself and compare it with the system figure. Divide the floor count by daily output to get the lead time customers actually experience, using the arithmetic of Little’s Law. The gap between the two figures is the size of the second queue, and its location tells you where capacity is leaking.

Trace — mapping the escalation path to the people who hold the know-how

Ask each person whom they call when something breaks, then draw the answers as a map. Names that recur mark where knowledge and decisions concentrate, including customer calls that route straight to the founder. The map shows whom to support in the first month and which know-how to write down first.

How readable is your floor in 48 hours?

Score each question 0, 1 or 2 after your two days. Then read the band underneath.

Question	0	1	2
Does the system’s work-in-progress figure match your physical count within 10%?	No	Partly	Yes, documented
Are the steps you saw done from memory written where a new starter would find them?	No	Partly	Yes, documented
Do routine customer calls reach someone other than the founder?	No	Partly	Yes, documented
Did the work running at mid-morning match the day’s plan?	No	Partly	Yes, documented
Can three or more people resolve a common breakdown?	No	Partly	Yes, documented

0–3: Treat the reports as unreliable and plan your first month around the floor. 4–7: Fix the lowest-scoring point before adding any new measures. 8–10: The reports are a fair guide; test them again at month-end, when pressure peaks.

What to do next

Two days on the floor show an ordinary week. This method stops working for month-end close and seasonal peaks, and people behave differently while a new manager watches. Treat the first read as a hypothesis and test it again at the next peak.

A 48-hour floor read turns two days of observation into a working diagnosis by following the second queue and the escalation path. This week, pick one live order and follow it from enquiry to invoice, noting every place it waits.