The two-day training is a symptom

A business buys a new system. The vendor offers a two-day training session, everyone nods, because that sounds responsible. Nobody asks the obvious question: why does it take two days to learn how to record a sale?

Our view is blunt. When staff need days of training to do the most common task in a system, the design is the problem, not the staff. People who run a till or a back office are not slow. They are doing a job under time pressure, often with a customer waiting, and the system is supposed to fit that job.

Where the two days come from

In most cases we see, the screens were built in the order the database is organised, not the order the work happens. A developer creates a table for customers, a table for products, a table for transactions. Then one screen per table: add, view, edit, delete. Engineers call this CRUD, and as scaffolding it is perfectly reasonable. It is the fastest way to get a system running.

The trouble starts when that scaffolding is what the cashier sees. Recording one sale now means opening the customer menu, then the product menu, then the transaction menu, and filling in fields that exist because the database needs them, not because the cashier does. A branch code the system already knows from the login. A product code, when the cashier recognises the item by sight. A discount ID that is mandatory and that nobody can explain.

Comparison of two ways to build screens: one screen per database table versus one screen per job, with the result that the first needs two days of training and the second one shift

Each of those fields is a small question the cashier has to answer: do I need this one, and what goes in here? Two days of training is mostly the time it takes to memorise the answers.

Measure the steps, not the features

We judge a system by one number: how many steps its most frequent job takes. Recording a single transaction at the till, for example. Then we cut steps until the job can be taught in one shift. In practice that means a few habits:

  • Remove what the system can fill in by itself, such as branch, date, and operator.
  • Order the screens the way the work happens at the counter, not the way the tables are stored.
  • Move rare cases, such as refunds and special pricing, out of the main path so they do not clutter the daily one.
Three ways to cut steps in the most frequent job: fill in automatically, order screens like the work, move rare cases aside, with step count as the measure

None of this is new. Nielsen Norman Group's well-known usability heuristics say a design should follow real-world conventions, making information appear in a natural and logical order, and should use words familiar to the user rather than internal jargon. It keeps being ignored because following the database is cheaper for the builder, and the cost lands on someone else: the people who use the system every day.

What it costs the business

Training is not a one-off expense. Staff turn over, shifts change, a cashier covers for someone on leave, a new branch opens. Each of those brings another person who has to learn the system. If learning takes two days, the business pays for it again every time. If it takes one shift, most of that cost disappears.

There is a quieter cost too: workarounds. When a field is confusing, people type whatever lets them move on. The reports you rely on later are only as good as those guesses.

A test you can run before you sign

If you are choosing a vendor, you do not need a research programme. Ask for one simple thing: find someone who has never seen the system, ask them to record one transaction without help, and watch.

Four steps of an unguided test: find a fresh user, give one task, stay quiet and watch, note every stop

Say nothing and do not point. Note where they stop, hesitate, or ask a question. Each of those moments is a design problem, not a personal failing of the tester.

You also do not need many people. In a widely cited article, Jakob Nielsen reported that a first study with five participants had found about 85% of the usability problems, while a single user found about 31% on average. Treat those figures as a rule of thumb for one design, not a guarantee. The direction is still clear: a handful of fresh users tells you more than a demo run by the vendor's own people.

Share of usability problems found: 31 percent with one tester and 85 percent after the first study with five participants, per Nielsen Norman Group

What we would do

If a system you already use takes days to learn, you can still act. Pick the most frequent job, count its steps, and ask your vendor or developers which ones can be removed or filled in automatically. Then test again with someone new. When the answer is "the staff just need more training", ask for the step count first.

At XETUP this is the standard we hold our own builds to: a job that cannot be taught in one shift sends us back to the screen design, not to the training calendar. If you want a second opinion on a system you use or are considering, write to bisnis@xetup.id or WhatsApp +62 822 6515 7644.