Observation Checklists: A Step-by-Step Guide with Examples and Templates

Train teams faster and more effectively with AI

Start Free

Before somebody on my team does a job on their own for the first time, I stand there and watch them do it once, ticking off each thing I see them do. The form I tick is an observation checklist. The first one I used had no box for "didn't see it".

Some weeks a job does not come up at all, so there was nothing to tick. I'd leave those rows empty, and whoever read the form after me took the gap for a fail. Then I started ticking them instead, to be fair to the person, and my form ended up saying somebody had done a job I never watched them do. It took me far too long to work out that the fix was a column on the form, not more discipline from me.

The three checklists that you can download above all have that column. There are 39 items between them, covering a lesson, a workplace task and a clinical skill, and every item is written as something you can watch happen. After the file I'll go through the rating scale, what to do about the rows you never saw, and how to build your own list if none of the three fit.

What an observation checklist is, and when you would use one

An observation checklist is a list of specific, visible actions that an assessor ticks while watching somebody perform a task, rather than rating them afterwards from memory. Down one side of the form are the things you expect to see the person do. Next to each one you record whether they did it, and how well.

I reach for one when I have to decide whether a person can be left alone with a job. A written test would tell me what they know. A performance review would tell me what I think of them over six months. This tells me one thing: on that day, doing that job, did they do these things. I can answer that from what I wrote down at the time.

You'll see these anywhere somebody gets signed off before working unsupervised. A mentor watching a trainee take a class. A supervisor watching someone drive a forklift before the ticket is handed over. A nurse educator watching a cannula go in. An assessor collecting proof that a student can do what their certificate says.

What makes an item observable

Every line has to name something you can watch happen. Most of the forms people send me fail that in the first three rows.

"Understands the safety procedure" asks me to guess what is going on in someone's head. Put two of us in the same room for twenty minutes and we'll guess differently, and afterwards neither of us can point at anything. "Identified the hazards before starting the task" asks whether a thing happened, and we can both check.

So I test each line the same way before it goes on a form. Could the person standing next to me tick it the other way? If they could, I rewrite it until they can't.

The rating scale

A yes/no box is enough for anything binary, like whether somebody put the PPE on. For work that can be done badly or well, I use four levels. I took the shape from the Millersville lesson observation form, one of the few templates ranking for this search that bothers to define its own scale.

Four levels rather than five, on purpose. With five I had a middle box, and I ticked it every time I couldn't decide. Four makes me pick a side.

Only one line on that table changes anything. A 3 means I'd put the person on the job alone tomorrow. A 2 means I wouldn't, and I owe them a sentence saying why: "Set the guard correctly, after I pointed at it".

The NA column, and why it's excluded from the score

That Millersville form spells NA out as "Not Applicable or Not Able to be Observed". The second half is the part I care about, and I could not find it on any other template ranking for this search. The page sitting at number two links to twenty-two pages of checklists and says nothing at all about how to score them.

Tick NA when the task never came up, or when you were dealing with something else at the moment it did. Leave the rating blank. In the file above, NA rows drop out of the maths, so the percentage counts only what you actually watched.

The filled example works out at 11 items observed, 3 not observed, 9 of those at 3 or above, so 82%. Scored the other way it reads "9 out of 14", which is 64%, and three of those fourteen never happened. I hand over "11 of the 11 I saw, 3 not seen", because I can defend every number in it.

Who observes, and how often

One visit tells you about one session. It does not tell you whether the person can do the job. I've caught myself forgetting that, usually on a week when I wanted the sign-off finished.

The MET project put a number on it. Across 67 teachers with four lessons each, one observation by one observer scored between .27 and .45 for reliability, where .65 is the usual floor for a decision that affects somebody's job.

The same study found the fix, and it is not the one most sign-off policies ask for: "having one video watched by two different observers produces more reliable scores than having one observer watch two videos from the same teacher". A second person beats a second visit. So when I can only spare one extra observation, I send somebody else into the same session rather than going back myself. The full paper is worth reading if you're writing a policy rather than a form.

When two of us disagree by more than one level on an item, I now assume the item is badly worded rather than one of us being wrong. Fixing the wording is assessment validation in miniature, and it is far easier before the form reaches a compliance file.

Fill it in during the observation. I know how obvious that sounds, and I still got it wrong for years. By Friday all I had left was a general impression of the person, so I was filling the form in from the impression rather than the other way round.

Writing your own list

Start from the task, not from a competency framework. Watch someone do the job once with no form in your hand, write down what they did in the order they did it, then turn each step into an item. For a person in their first weeks, the list usually falls out of whatever onboarding plan you already run.

Keep it to one page, which is usually twelve to twenty items. Longer forms get skimmed, and skimming produces ticks instead of observations. Needing forty items means you're watching two jobs at once, so split them into two forms.

Group the items into three or four sections and put a comments box under each. Require a note on every 1 and every 2, as a rule rather than a suggestion. When somebody queries a rating three weeks later, that note is the only thing I've got to show them.

When it doesn't work

When a checklist isn't the right tool

A checklist records that a person did a job while you watched. It will not tell you whether they can still do it in six months, whether they know why the steps go in that order, or whether they'd cope with a situation you didn't see.

Watching them again later answers the first. A written assessment answers the second, which is where formative and summative stop being a glossary entry. Nothing answers the third properly, and I'd rather say so than have a sign-off fall over later.

It is also not a training record. To show an auditor that a named person was assessed on a date by a named assessor and passed, you need that held somewhere with an audit trail, not in a spreadsheet where any row can be edited afterwards. That is the gap compliance training software fills.

Coursebox is a training platform rather than an assessment tool, so it will not replace the form. What it holds is the record: who was assessed, when, by whom, and what they scored, in the same place as the training itself. If all you need is the checklist, download it and go. You don't need us for that.

What to do next

Open the checklist that matches what you're observing and read the items before you use them. You'll want to cut three and add two, because the jobs in your workplace are not the ones in mine.

Then run one observation with it, live, and count how many rows come back NA. More than a third means either the list is too ambitious for a single session, or you watched at the wrong moment. Better to find that out now than on the day the form is deciding somebody's sign-off.