Measuring Monday Automation: Five Numbers Worth Tracking

Most people measure Monday automation by how clever the setup feels. A better question is whether the job is getting done correctly with less of your attention. That takes a few numbers, kept simply.

This post gives you five. You can track them in a notebook or a spreadsheet with one row per week. No benchmarks are quoted, because the right numbers are your own and you will build the baseline yourself.

Lesson 6.6 walks you through it step by step: Hand off the job you do every Monday.

Track outcomes, not activity: did it run, was it right, how much did you fix, how long did you spend, and how fast would you notice a failure.

Activity versus outcome

Activity numbers are easy and misleading. "The automation ran 52 times" tells you nothing about whether the output was usable. Outcome numbers ask whether the job did what you needed.

A useful rule: every metric should lead to a decision. If a number does not change what you do next, drop it.

Metric 1: Run rate

What it is. The share of scheduled weeks where the job actually ran.

How to track it. One tick per week in your sheet. If you have a completion note, a missing note means a missed run.

What it tells you. A drop means something broke, such as a changed file, an expired sign in or a paused schedule. If you use a scheduler that supports pausing and manual triggers, as the Hermes Agent documentation describes for scheduled tasks, check that it was not left paused after a test.

Metric 2: Correct on first look

What it is. Whether the result was right without any editing.

How to track it. Yes or no each week. Be strict. If you fixed anything, it is a no.

What it tells you. This is the most important number. A job that runs every week but needs fixing every week is not saving you much.

Metric 3: Correction count

What it is. How many things you changed in the output.

How to track it. A tally. "Two edits" is enough. Add a few words on what they were.

What it tells you. Patterns. If the same correction appears three weeks in a row, it belongs in the written steps. Fix the note and the count should fall.

Metric 4: Your time on the job

What it is. The minutes you spend on the job after handing it off, including reading the note and fixing issues.

How to track it. A rough number. Compare it with the baseline you timed by hand before you started. Write the baseline down early, because you cannot reconstruct it later.

What it tells you. Whether the effort was worth it. If your time has not dropped after a month, either the job is unstable or the steps need work. It is not unusual for the first weeks to take longer than doing it by hand, because you are checking everything.

Metric 5: Time to notice a failure

What it is. How long a failure would sit unseen.

How to track it. You will not have data until something fails, so run a drill. Pause the job on purpose one week and see how long it takes you to notice.

What it tells you. Whether your completion note works. If you did not notice for two weeks, the note is too easy to ignore. Make it shorter, put it where you look first, or have it say something loud when the job did not run.

The weekly review table

Fill it in on Monday afternoon. It takes about two minutes. Once a month, read down the columns.

How to read the table

• Ran is mostly yes, right first time is mostly no: the steps are weak. Rewrite them.
• Edits are falling: the job is stabilising. Consider spot checks instead of full checks.
• Minutes are not falling: the job may be unstable, or the checks are too heavy for a job this small.
• Everything is yes and low for months: you could widen the job, or move on to a new one.

Do not draw conclusions from one week. Look at four.

Metrics to skip

• Number of automations built. It rewards quantity.
• Hours saved estimated by a tool. You cannot verify it, so use your own timed baseline.
• Anything you will not look at.

A short example, illustrative

Say a freelance designer automates a weekly client status note. After four weeks her sheet shows it ran four times, was right first time twice, with edits of three, one, zero and zero. Her time on it fell from her baseline once she added a line about tone. These numbers are invented to show how the table reads. Yours will differ.

Setting your baseline

Metrics only mean something against a starting point. Before you hand a job over, spend one Monday collecting four facts about it.

First, how long it takes by hand, start to finish. Time it, even roughly. Second, how often you currently make mistakes in it, or how often you have to redo part of it. Even a guess like "about once a month" is better than nothing. Third, how long it sits undone when you are busy or away. Fourth, how much you dislike it, on a scale of one to five, because motivation affects whether you keep the setup going.

Write all four on one line and date it. Four weeks later, record the same four things for the automated version and compare. The comparison is personal. Someone else's job and someone else's time cost tell you very little about yours.

One more tip: record the metrics for the first run even if they look bad. Early weeks nearly always cost extra attention while you check everything. Seeing that on paper stops you from giving up after week one, and it shows you how much the later weeks improve.

Pair the numbers with a routine

Measuring only helps if a review follows. The HAND framework gives you a weekly loop to use these numbers in, and Lesson 6.6 covers the review step: Hand off the job you do every Monday. For a daily version of the same habit, see Lesson 6.1, plan your day in five minutes, which ends with a short look at what moved.

Questions people ask

How long should I track before judging?

Four weeks at minimum. One week is noise.

What if I do not have a baseline?

Time the manual job once, next Monday, before you hand it off. It is the best number you will collect.

Do I need a spreadsheet?

No. A notebook with one line per week works.