Incorrect commitments in customer service only surface when the customer writes back
Service replies are checked against refund, warranty and delivery rules before sending. Incorrect commitments fell from around 8 to under 2 percent.
- Retail
- Customer service
Key metric
Replies with an incorrect commitment
8%<2%
Share of reviewed replies containing a commitment outside the rules
At a glance
Starting point: around 300 closed service cases with the reply sent and a note on any later follow-ups.
The problem
The retailer's customer service handles around 15,000 written cases a week via email, contact form and messenger, mostly about returns, warranty claims and delivery and refund status. Around 120 employees at several locations answer these requests using text templates and a knowledge base that holds deadlines, goodwill limits and responsibilities. Before sending, team leads spot-check roughly five out of every hundred replies, mainly from new employees. All other replies go out to customers unchecked. Escalations and the monthly review regularly show that individual replies contain commitments that deviate from the applicable rules.
Because almost all replies go out without a second pair of eyes, quality depends on the individual's experience and how their day is going. Incorrect commitments rarely stem from carelessness but from special cases: different warranty periods for certain product groups, temporary goodwill rules, and different deadlines for marketplace and direct orders. Such errors are not noticed when the reply is sent, but only when the customer invokes the written commitment and writes a second time. The case then has to be walked back, escalated internally and decided again, which costs trust and takes more handling time than a correct reply the first time. The team leads' sample is too small to spot patterns early, and too late, because it only takes effect after sending.
- 15,000
- service cases per week
- 5 / 100
- replies checked in advance
Why it was left unsolved
The ticketing system can enforce mandatory fields, but it cannot judge whether a freely worded reply contains a commitment that should not be made. Rule-based checks fall short because the decisive statements are written in everyday language and the exceptions change several times a year. The topic was too small for a dedicated IT project, yet the team felt it every week.
The solution
Every finished reply is read immediately before sending and checked against the company's stored rule set: statutory and commercial warranty periods per product group, current goodwill limits, deadlines for returns and refunds, and responsibilities per ordering channel. If a commitment deviates, a mandatory detail is missing or the tone is inappropriate, the employee gets a short note before sending, showing the affected passage and the relevant rule. Unremarkable replies go out without any extra step. Decisions and wording stay with the person, but every reply is checked instead of only every twentieth.
- Every outgoing reply is checked against deadlines, warranty periods and goodwill limits from the knowledge base
- Note before sending with the passage highlighted and a reference to the relevant rule
- Silent approval of unremarkable replies, so the standard case needs no extra step
- Weekly analysis of the most frequent issues as a basis for training and text templates
Impact and effort
Period: first 3 months after launch.
| Metric | Value | Basis |
|---|---|---|
| Replies with an incorrect commitmentKey metric | around 8% → under 2% | Samplemeasured |
| Replies checked in advance | around 5% sample → all replies | System measurementmeasured |
| Confirmed issuesQuality metric | around 80% of notes confirmed by team leads | Samplemeasured |
| Repeat contacts due to incorrect information | noticeably reduced | User surveycustomer-reported |
Cost-effectiveness
The check only detects deviations from stored rules. It cannot assess new goodwill decisions or statements with no basis in the rule set, and the rule set has to be kept up to date.
Also works for
A similar case in your organization? Let's talk.
Briefly describe your process. We'll get back to you with an initial assessment of whether the pattern can be transferred and what it would realistically take.