The automation is built and the team has taken it over. So did it earn its keep? This lesson goes back to the baseline you wrote in the first lesson and fills in the measurement record: which measure, which value, how it was measured and which decision it was tied to.
This series started by choosing one task. Then you put the process into writing, defined the data, chose the tool, built the flow, connected the systems, settled the exceptions and handed the work to the team. This lesson looks back and takes on what comes after.
The reason for measuring is not to produce a report. Measurement answers two questions: is the thing that was built doing its work, and is the next step growing it or stopping it. Without numbers both turn into argument.
Prerequisite: a baseline written in the first lesson and a flow that has been running for at least a month. Without a baseline this lesson still earns its keep, but the first measurement is not a comparison; it becomes a new baseline.
The easiest things to measure in automation are often the ones that earn the least: how many records went through, how many messages went out, how many hours the system stayed up. These show that the system is running, not that the work got better.
The measures that earn their keep come out of the work itself. Which side of the task you chose in the first lesson was bothering you? The measure sits there.
How long a task takes end to end, how much of it gets corrected by hand, how long the customer waits. These are the measures that feed a decision.
How many records passed, how many alerts came up, how many exceptions are in the log. Needed to explain the measure of the work, not enough on its own.
How the team finds the work, what the customer says. It does not turn into a number but goes onto the form as a line; it carries the reason behind the number.
The form fits on one page and each row holds one measure. The columns match the questions a measurement has to answer. If a column stays empty, that row cannot be compared at the next measurement.
Filling the form in before the measurement differs from filling it in after. Filled in first, the line for how it was measured stays fixed; filled in after, each round gets measured a little differently.
The fifth column is the decision: what this row led to. Left empty, the form stops being a measurement record and becomes a list of numbers.
In the first lesson you drew up the task inventory, chose one task and wrote its baseline: how many times a week the task repeated, how long each run took, in how many of them a correction was needed.
Open that page now. The ground of the measurement is there. Without a baseline the number today is on its own neither good nor bad; it is only a number.
If the baseline was written by estimate, it goes onto the form that way. Comparing an estimate with a measurement is not the same as comparing two measurements; this note keeps a wrong reading away later on.
The four measures below are enough to tell the work in most automations. Measuring all of them at once is not called for; you start from whichever side was bothering you in the first lesson.
Each measure has a counting rule. Measuring without writing the rule leads to measuring something else at the second round.
How long the task takes end to end and how much of that is waiting. Waiting usually moves more than duration does.
How many records were corrected by hand, how many went into the exception log. The two get counted apart: one is a fault, the other a case outside the rule.
How much of the task goes through the flow. In a flow that started through a narrow door this share changes over time and touches the other numbers.
How often to measure depends on how fast the measured thing changes. Daily measurement produces noise in most work: the number moves but the reason is which day of the week it is. Monthly measurement shows a problem late.
The common arrangement is this: weekly for the first month, monthly after that. The reason for measuring often in the first month is not to see improvement but to settle the counting rule itself.
Writing down the days left out matters. Left unwritten, six months on the reason for a jump in the number is not remembered and the jump gets tied to the automation.
A measurement sometimes does not come out in the wanted direction. That does not mean the automation failed; first you look at what the number is telling.
There are three possibilities and each has a different answer. They get ruled out in order.
The counting rule may have changed, a day left out may have been counted, or the source may have changed. This gets ruled out first, otherwise the wrong work gets started.
The flow may be carrying a small part of the task. In that case the total duration hardly moves; the coverage measure shows it.
The bottleneck may sit before or after the automated step rather than in it. The process map gets opened again here.
The most common result coming out of an automation is time. Left unwritten, this time turns invisible: the work gets lighter but nobody can show it, and the next automation calls for persuading.
Writing down where the time went makes what the automation returned concrete and feeds the growing decision.
Writing the newly born work down as well keeps the table honest. The ownerless work problem from the ninth lesson usually shows up here, next to the time gained.
When a measurement comes out well the first thought is usually to set the same thing up everywhere. There are three separate paths and the three are not the same thing; which one was chosen goes onto the form.
The decision depends on which side of the measurement came out well. Opening a new task while coverage is low leaves a working flow half done.
Growing the coverage: raising the share of the task that goes through the narrow door. If the coverage measure is low the next step is usually this.
A second branch, a second team, a second product line. The process map gets copied but the data and permission lines get written again.
Moving to the second task waiting in the inventory. The scoring table from the first lesson gets opened a second time here.
Growing is not repeating the setup. Which part of the setup carries over as it is and which gets written again is separated at the start. Otherwise the second setup takes half as long as the first and produces twice the faults.
The order is this: first what gets copied, then what gets adapted, and last what gets merged.
The postponed line should not be left empty. A matter left unsolved in the first setup does not solve itself in the second; it doubles.
Not every automation calls for growing. Some do their work once set up but what they return does not meet the maintenance load. Writing this down opens the way to closing an automation as a measurement result rather than as a failure.
The stopping decision goes onto the form as well: which measure, which value, on which date and with which decision it closed.
A stopped flow does not get deleted; it gets closed and its record stays. When the same task comes back a year later the setup sheet and the process map are still sitting there.
In the first lesson the office chose following up the messages coming from portal listings. The baseline was written like this: eighty messages a week, each answered within about two hours on average, and a share of the messages getting lost in the late afternoon.
At the end of the sixth month the office fills in the measurement form for the second time. The comparison gets made against the row from the first lesson.
The waiting row: at the start the first reply averaged two hours, today it sits around forty minutes. How it was measured is written down: the difference between the arrival time of the portal record and the time of the first reply, weekdays, holidays left out.
The coverage row: not all of the messages go through the flow. Requests arriving by telephone still get entered by hand, and this does not pull the total duration measure down. Coverage first goes into the decision column; no new task gets opened.
The exception row: the log holds nineteen records over six months, eleven of them coming from the same three cases. All three have gone into the flow as rules and the remaining eight get settled by hand. This row shows the maintenance load.
The time gained row: six hours a week of the office manager has come free; a part of that goes to going over the pending list and the rest to appointment conversations. The newly born work is written on the same row.
The decision of the office turns out to be widening rather than growing: taking the requests arriving by telephone into the same flow. The second task gets noted on the form as the row to be chosen from the inventory once coverage has risen.
This week open the measurement form and fill in the first row for the flow you have running. Waiting six months is not called for; even if the first measurement is not a comparison, it becomes the ground of the second.
Download the measurement record (xlsx) · a one page form with columns for which measure, the baseline, the value today, how it was measured and which decision; the measurement calendar, the time gained and the growing decision are in it as well.
The ones below come out of how the measurement was set up rather than out of the measurement. What they share is this: none of them shows up as a fault in the system.
The number today gets written but there is no ground to compare it against. The argument runs over feelings rather than over the number.
At the second measurement the counting goes a little differently. The two numbers look comparable but do not tell the same thing.
A thousand records processed gets reported. How long the work takes and how much of it got corrected stay unmeasured.
The flow gets counted a failure because the duration measure does not move. Yet a small part of the task goes through the flow and the problem sits there.
The measurement gets done, the number gets written, no decision gets tied to it. A few months on, the form stops being filled in.
A flow whose return does not meet the maintenance load stays on the maintenance list for years because nobody talks about closing it.
If you can answer every item on this list at the end of a measurement round, the measurement counts as settled.
The words that come up in a measurement conversation. Calling the measures by the same names keeps the form readable in the sixth month as well.
In this lesson you closed the series: you chose the measure, went back to the baseline, wrote down the reason behind the number and separated the growing and stopping decisions. The curriculum comes to its end here.
The next round starts with the second task. For that the inventory and the scoring table from the first lesson get opened again.
The task inventory, scoring with three criteria and the baseline. The ground the measurement compares against was written there.
The handover checklist and the single door. The work newly born next to the time gained gets owned there.
The whole training section and the modules added since.
If you would like to fill in the measurement form together, write to us and we will draw up your baseline.
Get a Quote on WhatsApp