Client satisfaction surveys on WhatsApp: one question, one button, one action

Most brokerages do not measure satisfaction; they infer it from the cancellations that arrive at renewal, which is the most expensive way to find out. Email surveys get a handful of replies, and ten-question web forms are abandoned halfway.

On WhatsApp, a single question with three buttons gets answers within minutes, because it reaches the client where they already read their messages and takes one tap. The point is not the score but what it triggers: an unhappy client identified the same day can still be rescued, and an enthusiastic one is a potential referrer or online review.

This guide is written for brokerages in South Africa, Kenya or Nigeria working to Treating Customers Fairly outcomes, and for UK and Irish firms monitoring customer outcomes under Consumer Duty. It covers when to ask, how to word a question that does not lead the answer, which Meta template category applies, and above all how to organise the handling of replies so that the survey becomes a retention routine rather than a dashboard nobody opens.

The journey, step by step

1. Pick the moments that matter, not an annual survey. Ask after a specific interaction while the memory is fresh: a claim settled, a new policy incepted, a renewal processed, an advice meeting, a complaint resolved. A general survey sent to the whole book once a year gets few replies and does not tell you what worked. The brokerage defines three or four triggers and the delay after each (for instance two to five days after a claim payment, the day after a meeting) and keeps to them.
2. Write one question with three buttons. The question is one sentence about the experience just had, not about the brokerage in general: "How did the settlement of your claim go?" with buttons "Very well", "Average", "Badly". A 1-to-10 scale works too, but every digit then comes back as free text to classify. Three buttons is the maximum a WhatsApp template allows, and it is enough to sort replies. Sign the message with the consultant's first name: a question asked by a person gets more replies than one asked by a brand.
3. Submit the template in the right Meta category. A satisfaction question tied to a specific transaction (claim settled, policy incepted) is generally accepted as a utility template; a brand-perception survey with no recent transaction behind it is likely to be classified as marketing, which requires a marketing opt-in and excludes anyone who has objected under POPIA section 69 or PECR. Submit the template with the explicit context ("following the settlement of claim {contrat}") and wait for approval. A rejected template is reworded, not sent as free text to clients whose window is closed.
4. Deal with every "Badly" the same day. The client's reply opens the 24-hour window: that is when a person, not an auto-reply, writes in free text. The consultant asks what was missing, acknowledges the problem without blaming the insurer, and proposes a concrete action (a call, a check on the file, a gesture if the brokerage has the discretion). A negative reply left unanswered for more than a day is worse than no survey at all: the client has voiced a complaint and watched it go nowhere, which is exactly the kind of outcome TCF and Consumer Duty reviews look for.
5. Use positive replies without pushing. To clients who answer "Very well", the brokerage can suggest, in the same thread, leaving a review or recommending a friend. One message, no chaser, and never a reward conditional on a positive review, which breaches the terms of Google and Trustpilot and consumer-protection rules on misleading practices. These clients also feed an "advocates" segment, the natural starting list for a referral programme.
6. Measure by consultant and by reason, then fix the process. Each month, review the reply rate, the share of "Badly" by interaction type and by consultant, and the verbatims. A high share of "Average" after claims usually points to missing progress updates; after inception, to policy benefits not being explained. Findings become process changes, not individual blame. Keep the results: in South Africa they evidence the TCF outcomes the FSCA expects intermediaries to demonstrate; in the UK they feed the outcomes monitoring the FCA expects under Consumer Duty.

Sample messages

Three days after a claim payment, utility template with buttons

Reply to a client who answered "Badly", inside the window

Reply to a satisfied client, inside the window

Pitfalls to avoid

• Asking five questions: on WhatsApp every extra question halves the number of complete answers; one question, one tap.
• Blasting the whole book at once: with no recent interaction behind it, the message reads as marketing and Meta may classify it that way.
• Letting a bot answer the unhappy clients: someone who has just tapped "Badly" expects a person, not an automated acknowledgement.
• Offering a gift for an online review: against review-platform terms, a misleading practice under consumer law, and damaging if it becomes known.
• Surveying clients who have opted out of all messages: even a utility template must not go to someone who replied STOP.
• Collecting scores without reading them by consultant and by interaction type: the dashboard stays decorative and nothing changes.

How ORIS organises this use case

In ORIS the survey is a campaign with the "Satisfaction / retention" goal, built in five steps or as a Quick Campaign: a client list (for example the month's closed claims, picked by manual selection or imported as CSV), an approved template with up to three quick-reply buttons, a send window and a daily limit. The campaign detail shows reply rate, the split of answers and the sentiment detected.

Every reply lands in the shared inbox. A negative answer is classified by the AI as churn risk: the client gets an attrition_risk flag, their risk score rises and they appear in Opportunities & Risks sorted by "Highest Risk" so that a consultant calls them. Positive answers lift the engagement score and make it easy to create a "satisfied clients" segment in Customers & Segments, reusable for a referral campaign. The CSV export of campaign performance feeds the monthly review by consultant.

Frequently asked questions

Is a satisfaction survey a utility or a marketing template under Meta's rules?

When the question relates to a specific, recent transaction (a settled claim, a new policy), it is generally accepted as utility. A general survey with no reference transaction is likely to be reclassified as marketing, with the opt-in constraints that follow. When in doubt, write the template with the explicit context.

How many buttons can a WhatsApp template carry?

A maximum of three quick-reply buttons per template. That is why the "Very well / Average / Badly" scale works well; a 1-to-10 scale forces the client to type a number, which lowers the reply rate.

How often can I survey the same client?

No more than once per significant interaction, and never two surveys in the same month. A client asked too often stops replying or, worse, opts out and you lose the channel for service messages too.

Should I keep the answers?

Yes, attached to the client record with the date and the interaction concerned. They evidence the monitoring of service quality, useful during an FSCA or FCA review or a complaint, and they let you compare one quarter with the next.