How to analyse customer complaints and find what to fix

The goal: fewer complaints, not faster handling

Most complaint processes are built for handling: log it, respond, close it. Knowing how to analyze customer complaints adds the step that reduces them: looking across all complaints for the problems that recur, and fixing their causes. A team that handles complaints brilliantly but never analyses them will handle the same complaints forever.

Collect complaints in one place

• Negative and mixed reviews on Google Maps and app stores.
• Complaints to customer care by phone, email and chat.
• Complaints raised in store and recorded by staff.
• Social media posts and messages.

Keep the date, location, channel and full text for each. Negative public reviews are especially useful because they are written in customers' own words and are already public.

Classify to the specific issue

Broad categories make complaints look manageable and leave them unfixable. Classify each complaint down to a specific issue, the level at which someone could do something about it.

• Too broad: Service. Product. Store.
• Better: Staff behaviour. Waiting at billing. Exchange policy.
• Actionable: Exchange refused for resized rings. Second billing counter closed on weekend evenings.

The broad levels are still useful for reporting; the specific level is where action happens.

Find where complaints cluster

• By location: one store, a city, a format, or everywhere?
• By time: since a date, on weekends, in the evening, during a sale?
• By channel: only in app reviews, only in calls?
• By customer journey: before purchase, at the counter, after delivery?

Where a complaint does not occur is as informative as where it does. If exchange complaints come only from Mumbai stores, the cause is probably something only Mumbai stores share.

Measure the right things

• Volume: number of complaints about the issue in the period.
• Prevalence: complaints about the issue as a share of all reviews or contacts, so busier stores are not penalised.
• Trend: change since the previous period.
• Severity: some issues, such as safety or discrimination, matter whatever the volume.

Trace the cause

For each recurring issue, ask why until you reach something the business can change: a roster, a policy, a process, a system, a supplier, a training gap. Test that cause against the pattern you found. Then give it to an owner with a date to look again.

An example

A restaurant chain sees delivery complaints rise across its app reviews. Classified to specific issues, most are about missing items, not late delivery. They cluster in orders with more than five items, at stores using a new packing station, since its rollout in June. The cause is a packing checklist that was not updated for the new station. A revised checklist goes out in August; by October, missing-item complaints at those stores are back to their earlier level, while late-delivery complaints, a separate issue, are unchanged.

Pitfalls

• Counting complaints without a denominator. A store with twice the customers will have more complaints; compare prevalence.
• Analysing only formal complaints. Many customers never complain to you but write a review; include reviews.
• Closing a complaint when it is answered, not when its cause is fixed.
• Treating each complaint as unique. Most are instances of a few recurring issues.
• Ignoring praise. Positive feedback shows what to protect while you fix the rest.

Reporting complaints upward

Leadership reports usually show complaint volume by month. Add three things: the five specific issues behind most negative feedback, how each moved since last month, and which have an owner and an action. That turns a volume chart into a list of decisions.

Your first month

1. Week 1: pull the last three months of negative reviews and complaints into one sheet with date, location, channel and text.
2. Week 2: tag a few hundred by specific issue and count them. Note where they cluster.
3. Week 3: pick the three issues with the most complaints and a clear cluster, and run a short why chain for each.
4. Week 4: agree an owner and an action for each, with a date to check. Then decide whether to automate the classification.

Doing this with Falcon

Falcon does this for your Google Maps and Google Play reviews. Every review is classified into themes, sub-themes and issues from your own taxonomy, with the sentiment of each mention; specific issues are found by clustering what customers wrote, each with a short summary; and every count opens the reviews behind it, with the supporting quote highlighted.