How a South African brokerage splits its book into renewal segments by risk score, engagement and renewal window, then picks templates and tests timing on WhatsApp.
A renewal campaign sent to the whole book is a campaign sent to nobody in particular. The client who has paid by debit order for nine years without a query gets the same message as the one whose last two collections bounced, and the one who replied "cancel" in March gets it too. South African brokerages that get real retention out of WhatsApp do one thing differently: they segment first, and they decide the template type, the timing and the wording segment by segment. This guide shows how, using the filters a CRM such as ORIS exposes in Customers & Segments and the constraints that POPIA and Meta's template categories impose.
The natural instinct is to segment by product: motor here, household there, funeral in a third list. It is the wrong first cut. What determines the message is when the policy renews and whether the client is at risk of leaving, not whether the asset is a car or a house. A useful first pass for a personal-lines and funeral book looks like this:
| Segment | Filters | Goal of the campaign | Template type |
|---|---|---|---|
| Renewing in 30 to 45 days, low risk, engaged | Renewal window 30-45 days; Risk Score less than 40; Engagement medium or high | Confirm continuity, invite a cover review | Utility (renewal reminder) |
| Renewing in 30 to 45 days, medium or high risk | Renewal window 30-45 days; Risk Score greater than 40 | Open a conversation before the premium increase lands | Utility, followed by a personal message |
| Renewing in under 14 days, no reply to first reminder | Renewal window under 14 days; no inbound message in the last 30 days | Get a yes, a question or a reason | Utility (second reminder with a button) |
| Lapsed in the last 90 days | Policy status lapsed; marketing opt-in recorded | Win back, or learn why they left | Marketing |
| Long-standing, single policy, engaged | Days Since Inception greater than 1 095; one active policy; Engagement high | Cross-sell a second line at renewal | Marketing |
| Critical risk, any window | Risk Score greater than 80 | No campaign: a call from the account handler | None |
The last row is the important one. A segment is as much a list of who not to message as of who to message. The clients the AI has flagged as critical, usually after a reply such as "I'm thinking of moving to a direct insurer" or two failed debit orders, should be removed from every automated campaign and handled by a person. The guide to lapse risk on WhatsApp describes that path.
Two numbers do most of the work. The risk score (0 to 100, shown as Very Low to Critical) is updated from the client's replies, payment signals and flags: a negative sentiment in a reply, a question about cancelling, a coverage gap detected by the AI all move it. The engagement score reflects how the client actually behaves on the channel: whether messages are read, whether they reply, how recently. A high-engagement, low-risk client can be reminded once and left alone. A low-engagement client is not necessarily unhappy, but a template they never open will not retain them, and the data should push the brokerage to pick up the phone.
Three practical rules when building segments on these scores:
Meta's template categories are not a formality. A utility template carries a transactional message about an existing relationship: a renewal date, a premium, a document. A marketing template carries an offer. Meta reviews the category at submission, can recategorise a template that reads like an offer, and treats the two differently in pricing and in how quickly a business's quality rating can fall when recipients block or report messages. POPIA adds the other half of the rule: section 69 on direct marketing by electronic means requires consent from prospects and allows existing clients to be marketed similar products only with an opt-out at every message. The existing guide to POPIA consent on WhatsApp sets out the detail.
Applied to renewals, the decision is usually clear:
In ORIS each campaign carries a compliance tag of service or marketing; a marketing campaign automatically excludes opted-out clients, and a STOP reply records the opt-out on the client record so that the next segment refresh drops them. That is the mechanism; the judgement about which wording is which still belongs to the compliance officer who approves the template before it goes to Meta.
Timing is where most renewal campaigns are either too early or too late. Too early, and the client has not received the insurer's renewal notice and cannot act; too late, and the debit order has already run at the new premium. A working default for South African personal lines is a first utility message when the renewal window opens, a second message with a quick-reply button about ten days before the renewal date for those who have not replied, and a personal message from the handler for anyone who answered with a question. Send windows matter too: a reminder at 07:30 on a weekday is read on the way to work; one at 21:00 on a Sunday is read, resented and blocked.
Rather than argue about wording in a meeting, test it. A meaningful A/B test on a renewal segment changes one variable only:
The campaign view shows delivered, read and replied counts for version A and version B, the classification of the replies and the opportunities created; on a segment of a few hundred clients, one renewal cycle is enough to see a difference of a few points in reply rate, which over a book is a lot of retained premium. The Smart Learning Mode in ORIS can shift traffic towards the better variant during the campaign rather than after it. Whatever the result, write it down: next year's segment starts from this year's winner. More on the sales side of the channel is collected under sales and prospecting.
Usually four to six. Fewer, and the messages become generic; more, and the team cannot follow up on the replies. Start with the renewal window and the risk score, and add engagement and product only where it changes the message.
It can, but it then becomes a marketing message for both Meta and POPIA: a marketing template, clients without a marketing opt-out only, and an opt-out line. Many brokerages keep the reminder clean and send the cross-sell separately.
Check that the number is on WhatsApp and that messages are delivered and read. If they are read but never answered, move those clients to a call list for renewals rather than sending a fourth template; if they are not delivered, the number is wrong and the client record needs fixing.
Segmentation for servicing and retention is processing for the brokerage's legitimate purpose and the client relationship. What POPIA restricts is direct marketing without consent and automated decisions with legal effect on the client; a segment used to decide who gets a reminder or a call is neither, but the purpose should be in the privacy notice.
It depends on when the insurer issues its renewal terms, typically thirty to forty-five days before the renewal date for personal lines. The first message should follow those terms, not precede them, so that the client can act on it.
Shared WhatsApp inbox, client records, follow-ups and opportunities for the whole brokerage. 15-minute demo.
Fire cover is only the statutory floor. What the STSMA and management rule 23 oblige a body corporate to insure, and the checks that decide if a claim pays.
Sections 8 and 9 of the FAIS General Code do not change because the client bought on WhatsApp: the suitability steps, the record, and the copy the client gets.
Who holds the policy, what an unlicensed group may not do, the member data an insurer needs to quote, and how to run contributions and claims on WhatsApp.
