Blog/Operations
OperationsSeptember 4, 2026·11 min read

Training SOP: how to document the way you train

Most advice on this describes the culture you should build. This is the procedure itself: the steps, the owner, and the one check that tells you training actually worked.

AY
Anand Yadav · Founder, ReccordSOP
·Last reviewed September 4, 2026

The short answer

A training SOP is the documented procedure for teaching someone a process. It names who trains, what the trainee watches first, what they do supervised, how you check they can run it alone, and what gets recorded when they can. It is the SOP that governs how your other SOPs get into people's heads.

Written once, it makes every future handover the same regardless of who runs it. Without it, training quality tracks whoever happened to be free that week, and nobody can tell you whether the new hire is ready or merely finished.

Two things people mean by this

The phrase splits two ways and the answers are different, so it is worth being clear which one you are after.

  • A procedure for training itself. One document describing how anyone at your company teaches anyone else a task. This is the one almost nobody writes down, and the one this article is about.
  • Using your existing SOPs as training material. Handing a new hire the refunds SOP instead of talking them through refunds. Useful, and it is a delivery method rather than a procedure.

The second only works if the first exists. Give someone a document and no procedure around it and you have not trained them, you have forwarded them a link. The training SOP is what turns a document into a person who can do the job.

The training SOP itself

Here is the procedure, written the way any other SOP should be. Adapt the specifics, keep the shape.

  1. Trigger: a new hire starts, someone changes role, or an existing process gets reassigned. Training is scheduled the same week, not left to the first quiet afternoon.
  2. Name the trainer. One person, not the team. The trainer is whoever currently runs the process most often, and they own the outcome until the trainee passes step 7.
  3. Confirm the SOP is current before you teach from it. Open it and check the last review date. Training someone on a stale document teaches them the wrong process and gives it your authority.
  4. Trainee reads the SOP alone first, then lists what they do not understand. This takes fifteen minutes and it changes the session from a lecture into answers to real questions.
  5. Trainer performs the task once while the trainee watches and asks questions. Record this run if you can, because it becomes the training material for the next person.
  6. Trainee performs the task while the trainer watches and says nothing unless something is about to go wrong. Silence is the point. Prompting produces a trainee who can be prompted.
  7. Trainee performs the task alone on real work. The trainer checks the output afterwards, not the process during. This is the pass or fail step.
  8. Record the outcome: who was trained, on what, by whom, on what date, and against which version of the SOP. That last field is the one everyone omits and the one you will want later.
  9. Book the follow-up. Two weeks out, check one piece of real work. Most training decays at exactly the point everyone stops looking.
The field almost nobody records

Which version of the SOP someone was trained against. Six months later, when a step has changed twice, that field is the difference between knowing who needs retraining and retraining everybody to be safe.

Watch, do, teach

Steps 5 through 7 above are a progression, and the order matters more than the content.

StageWho actsWhat it proves
WatchTrainer runs it, trainee observesThe trainee has seen the real thing, including the exceptions
Do supervisedTrainee runs it, trainee is watchedThey can execute with a safety net
Do aloneTrainee runs it on real workThey can execute without one. This is the only stage that counts
TeachTrainee trains the next personThey understand it well enough to explain the why, not just the clicks

Most teams stop after the supervised run because it feels finished. It is the stage that proves the least. Someone who can do a task while an expert watches has demonstrated that an expert was present.

The fourth stage is optional and it is the best test in the list. A trainee who trains the next person will find the gaps in the SOP within an hour, because explaining a step out loud is where you notice it does not make sense.

Try it on one of your own procedures.

Record a process once, AI writes the structured SOP. 3 free SOPs, no credit card.

Start for free →

How to know training worked

The standard answer is a quiz. Quizzes are popular because they produce a number, and the number is close to meaningless.

A quiz tests whether someone read a document. The failure you actually care about is someone who read it, scored well, and still cannot handle the order that arrives with a missing line item. Reading comprehension and operational competence are different things, and only one of them shows up in your refund rate.

Check the work instead. Take one piece of real output from the trainee's first unsupervised run and inspect it against what the SOP says should have happened. Three things worth looking at:

  • Did they follow the steps in order, or find a shortcut? A shortcut is not automatically wrong. It is worth understanding before it spreads.
  • What did they do at the decision point? Every process has one step where judgement is required. That is where training either took or did not.
  • What did they escalate, and was it the right thing? Escalating too much is a trainee who does not trust the document. Escalating too little is worse.

How to prove your team actually read the SOP

Read receipts, acknowledgements and what they are genuinely evidence of, which is less than most teams assume.

Record the run, do not write from memory

When you build the training material, do not ask the person who knows the job to write it down. Ask them to do the job while it records.

This is not a preference. Someone who has performed a task four hundred times has stopped experiencing several of its steps as steps. They check a field by reflex. They know which supplier needs a phone call. Asked to write the procedure, they omit exactly those parts, because to them they are not parts, they are just what doing the job feels like.

Those omissions are precisely what a newcomer gets wrong. The unobvious step is unobvious to the expert too, right up until somebody skips it. Recording the real run captures the reflexes along with the clicks.

SOP vs work instruction: what's the difference?

Which document holds the rules and which holds the clicks. Worth settling before you decide what to record for training.

Seasonal hires and VAs

A DTC brand trains under two conditions that break the standard advice.

The first is peak season. You onboard a cohort in October for the eight weeks when mistakes are most expensive and your trainers have least time. The training SOP has to survive being run by someone who is busy, which is an argument for a short documented procedure over a well-intentioned long one. If your training process needs a calm week to work, it does not work.

The second is remote contractors and virtual assistants. You cannot sit behind them for the supervised run, and screen-sharing for an hour is expensive on both sides. The practical substitute: they record their first unsupervised run and send it. You watch it at speed, and you are checking the same three things as above. It is asynchronous, it takes ten minutes, and it produces better evidence than watching live because you can rewatch the decision point.

How to train a Shopify virtual assistant: the 30-day handoff

The longer version for one specific case: which tasks to hand over in what order, and where the handoff usually fails.

What they trained on stops being true

The part of training nobody plans for is that it expires, quietly, without anyone being told.

Someone trained in September on a refunds process learned a screen layout, an approval threshold and a set of steps. By January the tool has shipped a redesign, the threshold has moved, and one step is gone. Nothing announced any of it. They are still doing the job the way they were taught, which is now the wrong way, and they are confident about it because they were trained.

This is why step 8 records the SOP version. When a procedure changes, you can list exactly who was trained against the old one and send them the difference rather than the whole document. A person who trained on version 3 needs to know that steps 4 and 7 moved. Handing them version 6 and hoping they spot it is not retraining, it is filing.

The delta is the training material. The document is only where it lives.

None of this needs a platform. It needs a named trainer, a procedure short enough to run in a busy week, one check on real work, and a record of which version was taught. The teams that get this right are not the ones with the best documentation. They are the ones where somebody owns the answer to a simple question: who was taught what, and is it still true?

Frequently asked questions

What is a training SOP?

A training SOP is a documented procedure for teaching someone a process. It names the trainer, the order of the session, the point at which the trainee works unsupervised, how you verify they are ready, and what gets recorded afterwards. It is distinct from using SOPs as training material: the SOP is the content, the training SOP is the method for getting it into someone's head.

How do you write a training SOP?

Start from the progression rather than the topic. Name the trigger that starts training, name one trainer, have the trainee read the existing SOP alone and list their questions, then run watch, do supervised, do alone. Verify by checking one piece of real work from the unsupervised run. Record who was trained on what, by whom, and against which version of the document. Nine steps is enough.

How do you know if training worked?

Check output, not comprehension. Take one piece of real work from the trainee's first unsupervised run and compare it against the SOP: did they follow the steps in order, what did they do at the decision point, and did they escalate the right things? A quiz tells you they read the document. Neither the customer nor the refund rate cares whether they read it.

Should you use quizzes to test SOP training?

Sparingly, and never as the only check. A quiz measures recall of a document, which correlates loosely with the ability to run the process when something unusual arrives. They are genuinely useful for compliance topics where you need a record that specific information was communicated. For operational competence, a supervised run followed by an inspected piece of real work tells you far more.

How often should people be retrained?

On change rather than on a calendar. When a procedure is revised, look up who was trained against the previous version and send them what changed, not the whole document again. That only works if you recorded the version at training time. Beyond that, a light check on one piece of real work about two weeks after training catches most of the decay, which happens early rather than gradually.

Ready to record your first SOP?

3 free SOPs to start. No credit card required. See if drift detection keeps your docs honest.

Start for free
AY
Anand YadavFounder, ReccordSOP

I built ReccordSOP after watching too many DTC ops teams lose months to undocumented workflows. These SOPs are battle-tested with Shopify operators running $1M to $50M brands.

Last reviewed September 4, 2026

Related reading