Tracking training completion across 50 locations

At fifty sites the spreadsheet breaks

At fifty locations the spreadsheet model breaks. One operations director cannot chase fifty site managers every Monday for completion screenshots, and by the time the numbers are merged they are a week old while the auditor is asking about this quarter.

Fifty sites means somewhere between 750 and 1,500 frontline workers, three to five regional managers each owning ten to fifteen sites, and in any given week starters being inducted, a procedure change rolling out and a recurring safety or hygiene topic coming due for half the network. The volume is not the problem. Each moving part has its own deadline, and the only people who know the real status are site managers who do not have time to fill in your tracker.

The spreadsheet fails in three predictable ways.

• Freshness. It is only as current as the last person who updated it. If managers update on Mondays, the data is a week old by Friday.
• Versions. Somebody copies it locally, edits it and sends it back. Somebody else has a different copy. The single source of truth becomes a folder of near-duplicates.
• The chase. The whole model depends on managers updating it, so head office becomes a status-collection office instead of an operations function.

None of these is solved by a better template. The model itself is the problem: it is a copy of the truth made by hand, and it starts drifting the moment it is saved.

From asking to receiving

The move is from pull, where head office asks and sites answer, to push, where people train and the record exists because they did. Every interaction is already a data point: the module opened, the knowledge check answered, the course finished. Nobody has to report it.

A worker finishes the hygiene refresher on a phone at site 37 on a Tuesday afternoon, and it is on the operations director's view before that manager has clocked out. No screenshot, no chase.

There is a second reason to prefer the automatic record. A manually confirmed completion is a manager's statement that something happened. A system record is a timestamp attached to a person and a course version. When the gap between the two matters, it always matters in the same direction: the manual list is the one that says everyone was present.

What the view is quietly answering

The dashboard is not really for you. It is for the person who will one day ask, out loud, what you can prove.

In the Netherlands, article 8 lid 4 of the Arbowet requires the employer to supervise compliance with the instructions given: a continuing duty, not discharged by a session delivered once. In Belgium, the onthaal adds a signed record per starter. A VCA auditor asks for dated toolbox topics with attendance per site, covered in the toolbox guide. The full picture per country is in workplace instruction law in Belgium and the Netherlands.

So the design brief is not "show completion". It is: current status per person, retrievable by site and by name months after the fact, with the version of the material attached. What an auditor then does with that record is the subject of training records an auditor can check.

Track the version, not just the topic

Most trackers record that someone completed "the cleaning procedure", not which version of it. Procedures change. When the question is whether a worker was instructed on the method in force on the day of the incident, "completed, March" is not an answer. Hold the version in the record: it costs nothing when the course is built from the document, and it is the difference between a training log and evidence.

Five numbers carry the weight

The trap at scale is tracking everything. Five numbers do the work.

Everything else is derivative. If regional managers want more, give them filters rather than more figures.

One question per view, and three framings

Before a view is built, write down the one question it answers and the action it triggers when the answer is bad. If you cannot write the action, do not build the view. Applied honestly, this kills time on platform, average score across the network and login counts, none of which answer a question anyone in operations has asked. Write the question in the title if you can: "Sites below threshold, this week" documents itself in a way "Compliance overview" never will.

A rollup serves at least three people, and they want different things. One screen filtered by permission works structurally and fails behaviourally, because the site manager opening a network-shaped view has to find themselves in it, and most will not bother twice.

Pick the cuts your interventions actually use: by site when the fix is a conversation with one manager, by team when a site looks healthy but one team inside it was never assigned, by starting month when you want to know whether induction works, and by version of the material when an incident asks who knew the current method. A cut that produces a number you would never act on dilutes attention.

Put the time of the last update on every view. Overdue counts want to be live; starter ramp is fine daily; refresher coverage is fine weekly. And refuse the numbers that only look like measurements: averages across sites of very different sizes, percentages without a visible denominator, and comparisons with an outside benchmark, which tell you nothing about your own operation.

The weekly note the operations director reads

If the weekly training report is twelve pages of charts, it is not being read. A useful weekly note is one screen long.

The one-line cause is the diagnostic; without it the table is a list of percentages. Break the format only in three weeks: quarter end, the six weeks before an audit (add refresher coverage) and the week of a network-wide rollout (lead with early uptake).

When a site falls behind

A single round-number threshold either flags everything or nothing. Flag relative to your own network instead, with three separate signals: a site well below the network average, a site that has slid for two weeks in a row even if it is still above the line, and any overdue compliance topic at all, which is regulatory exposure rather than a performance question.

Send each flag to the closest accountable person and their manager, never to a mailing list that nobody reads. Then give the flag an owner over time: in week one the regional manager finds the cause with the site manager (starters not assigned, a new manager, a broken shared device); in week two the operations director joins, because the cause is usually structural; by week four it belongs on the agenda of whoever runs the relationship with that site. If the same three sites are flagged every week, the view is working and the intervention is not.

Replacing the spreadsheet

Replacing the tracker is a setup exercise, not a migration project. Do not map five years of dirty rows. Answer the question that matters now, who is inside their valid window for compliance topics and who needs onboarding, by importing the current roster and assigning the active courses. Keep the old files as a read-only archive in case an audit reaches back. The tabs never come back.

How Aristotl handles it

In Aristotl the record is the training. Courses of 3-5 minutes are done on the worker's own phone, joined through a link or QR code per site or team, and each completion lands per person, per site and per version of the material without anyone reporting it. The platform shows completion per site, region and person in one view, with the sites that are behind standing out and drop-off per module showing where people stop. Checklists and certifications report into the same place, so the Monday review is one screen rather than three sources.

A site manager can be given a view of their own site and their own people, without access to the central courses. That removes the weekly status email in both directions and takes head office out of the middle.

Take the site you are least sure about and one recurring topic, and check how long it takes today to name the people at that site who are overdue.