Refunds Done Right: From a Training Need to a Course

The Need

This case follows on from Finding Where Training Is Needed, where the model can be run. In an invented online retailer, Hartwell Home, the needs register opened a row for one customer service team. Its reopened tickets rose from about 9% to 16% over two months while the rest of the department stayed flat.

A ten-minute check on the linked skills found two problems. Agents applied the refund policy wrongly 38% of the time, usually with the wrong time limit. They gave no clear next step 34% of the time, usually with no time frame. The figures are illustrative.

Before anything was built, three other causes were ruled out: a new tool, staffing, and an out-of-date document. The refund policy had changed in the spring and only some agents had learned it. Two problems were passed on rather than trained: the internal policy summary still showed the old time limit, and the refund screen did not show the purchase date.

The Learners

The course is for agents already in the job, from 3 months to 5 years. They handle refunds every day. It is a refresher after a policy change, not an induction, and it fits into one 30-minute session and ten minutes a week.

The Objectives

Each objective targets one error the check found. By the end, an agent can:

1. Choose the right outcome (refund, repair, replacement or decline) under the current policy and consumer law.
2. Keep responsibility with the company: never send a customer to the manufacturer, and never ask for proof of a fault in the first two years.
3. End every reply with who does what next, and by when.
4. Check a reply against three points before sending it.

The Design

The module has five lessons.

Lesson	What happens	Why
Three Customers	Agents decide three cases before seeing any rules, then see model answers	Attempting first shows each agent where their habits differ from the policy
The Policy on One Page	The rules as one decision table	A table is what agents use at work
Watch It Done	Two cases worked step by step, a prediction question, a sorting task	Worked examples suit learners meeting a new rule set
Your Turn	Three cases, each with one step missing	Fading the steps moves agents from watching to doing
Before You Send	A three-point check	A short routine to use on the job

Four practice sets follow, one a week. Each has six questions that mix the skills, so agents first decide which rule applies. Before seeing each answer, they say how sure they are. A confident wrong answer shows a rule someone is applying wrongly without knowing it. Each answer shows which rule applied and why, never a score against colleagues. The policy table is open in week 1, a link away in week 2, and gone in weeks 3 and 4.

The Law

The first draft of Hartwell’s policy was invented, and it broke Spanish consumer law: it passed faults after 60 days to the manufacturer’s warranty. The course now teaches the real rules. Goods carry a 3-year guarantee. A fault that appears in the first 2 years is presumed present at delivery. The seller repairs or replaces first, and refunds if that fails. Hartwell’s own extras, a 30-day change-of-mind window and a 48-hour rule for damage, sit on top of the law.

Building It with AI

The module was drafted with Articulate 360’s AI from content written in advance, then edited by hand in Rise. The draft needed six changes:

What the AI did	Why it mattered	Change made
Merged two lessons and repeated the practice cases	Learners would answer the same cases twice	One set of cases, lessons reordered
Opened with “trust your instincts” and “respond before you check the policy”	It taught the opposite of the course	Decide now, then compare with the policy
Shortened a change-of-mind case to “looks different from the photo”	An item unlike its description counts as faulty, so the wrong answer became arguable	Rewritten as a plain change of mind
Said returns must be in “original condition”	Agents would refuse returns without the box	“The condition it arrived in”
Wrote a question with two correct answers	A right answer would be marked wrong	One option replaced
Oversold the course (“master”, “empowers you”)	Experienced staff switch off	A plain description naming the need

The Evaluation

The evaluation was planned before the course was built.

Question	Measure	When
Did they learn it?	The same ten-question check, scored by skill	Before the module and after week 4
Do they do it at work?	Reopened tickets, and a sample of refund replies scored on the four objectives	Eight weeks before and after
Was it the training?	A second team that starts four weeks later	Weeks 0 to 4
Did they find it useful?	Three questions, recorded but not counted as evidence of learning	Week 4

Results are reported by team and skill only, never by name, and never for a group under five. Nothing goes into a performance review. The register row closes when the reopen rate is back in line and the comparison team shows the change came from the training.

Limits

Hartwell Home, its figures and its teams are invented. The check scores are illustrative. A small group of volunteers will take the check, the course and the check again, and a group that size shows change, not cause. The business figure is designed and simulated in the data model, not observed in a real company.