Skip to content
arthova
InsightsTemplate

Measure first: What your automation should actually change

A measurement log for handling time, waiting time and exceptions, with a worked example and a CSV template for your baseline.

Put this to work

Separate active work, waiting and rework before projecting a benefit.

CSV download

Three kinds of time, three different decisions

An enquiry can wait for two hours even though someone spends only eight minutes working on it. Counting the entire elapsed time as labour overstates the capacity that could be released. Looking only at active minutes can hide a long wait for the customer.

Table – scroll sideways if needed

MeasureHow to record itWhat it helps you decide
Active timeActual work by everyone handling the case, including reworkCapacity and handling effort
Waiting timeTime with no active work between stepsHandoffs and response speed
ReworkIdentify corrections separately; they are already included in active timeFind quality problems without double counting

Start with the CSV measurement log. Use pseudonymous case IDs instead of customer names. An empty time field means “not measured yet”; enter 0 only for an observed zero minutes.

Why one average does not tell the whole story

This is a constructed example: five cases take 6, 8, 9, 12 and 45 active minutes, totalling 80 minutes. Investigate the long case as an exception. Do not quietly remove it from the calculation.

One dataset – three different statements

Illustrative example

16 min

Arithmetic mean

80 active minutes divided by 5 cases. The long case affects the mean.

Illustrative example

9 min

Median

The middle value of the sorted observations: 6, 8, 9, 12, 45.

Illustrative example

45 min

Longest observed case

A signal to investigate an exception, not a prediction for all future cases.

Mean and median follow the NIST definitions. Five constructed cases are neither a representative sample nor Arthova project data. Your own planning needs a selection covering normal cases, follow-up questions and relevant exceptions.

A manageable start for your measurement log

From the first observation to a usable baseline

  1. Set the boundary

    Define where the workflow starts and ends. Agree which people and corrections count towards active time.

    Your deliverableOne measurement rule shared by everyone involved.

  2. Collect observations

    Record cases during a normal working period. Five working days can be a starting point; rare or seasonal workflows need a different window. Record missing measurements too.

    Your deliverableA log showing case count, observation period and gaps.

  3. Explain exceptions

    Review long cases, manual handoffs and incorrect results. Record the reason instead of deleting inconvenient observations.

    Your deliverableSeparate views of normal cases and exceptions.

  4. Agree the comparison

    Use the same process boundary and comparable types of cases in a later before-and-after comparison. Check quality and rework as well.

    Your deliverableA baseline that remains comparable after implementation.

Model the benefit afterwards

Put measured active time and realistic volume into the capacity calculator. Waiting time does not become saved labour. Do not count rework twice, and label inputs that remain estimates.

Leopardo’s approximately six hours per week are a modelled capacity estimate based on the owner’s assessment. They do not replace your own baseline. Whether released capacity becomes lower cost, more output or quicker responses must be established in operation.

Your working document

Download working document

Direct download. No sign-up.

CSV

Sources & context

The working template is Arthova’s editorial interpretation. External guidance and our project evidence are linked separately.

  1. NIST: Measures of Location
See the documented project
From reading to doing

Bring one specific workflow.

We will discuss what can usefully be automated, what information is missing and how to check the outcome.

Discuss a workflow

A useful next step

Acceptance before production: A checklist for AI-supported workflowsCapacity before ROI: An honest process model