A commercial insurance brokerage. Eleven brokers, two claims administrators, 1,900 commercial clients, and a book that had grown steadily for two decades on the strength of personal relationships. Renewal dates lived in Outlook calendars. Client history lived in email. What a client had been advised, and why, lived in whichever broker's memory it had been formed in.
The trigger was not ambition. It was a resignation. A broker with eleven years of relationships left, and it took four months and a lot of awkward phone calls to work out what he had been in the middle of.
The unusual part of this project was not the software. It was working out what a CRM should even mean in a business that does not really have a sales pipeline.
Why a brokerage is not a sales team
Most CRM systems are built around a deal: a prospect appears, moves through stages, and closes or dies. That model describes almost none of how a brokerage earns money.
- The pipeline is a calendar. Revenue is renewals, spread across twelve months and mostly predictable. The question is never "will it close", it is "has anybody started on it yet".
- The relationship is owned by a person, not the firm. That is the commercial reality and pretending otherwise makes brokers hostile to the system on day one.
- Advice has to be evidenced. What was recommended, what the client declined, what was disclosed. A CRM that cannot hold that is just an address book.
- A single client is a bundle of policies with different insurers, dates, and claims histories, and the natural record is the policy, not the deal.
- Claims are relationship events. The client's experience of the brokerage is formed almost entirely during a claim, and the claim is usually handled by somebody other than the broker.
Configure a brokerage CRM around deals and it will be abandoned inside a quarter. Configure it around renewal dates, policies and evidence of advice, and brokers use it because it stops them losing money.
What was actually going wrong
- Renewals were being started late. A little over a third were being worked more than sixty days out. The rest became scrambles, and a scramble is where a client shops around.
- Nothing survived a handover. When somebody was ill, on leave, or gone, the client got a broker who knew nothing about them.
- Cross-sell was invisible. Nobody could produce a list of commercial clients without, say, cyber cover, so nobody ever offered it systematically.
- Records of advice lived in sent items. Retrievable, in theory, by the person who sent them.
- Commission was reconciled by hand, statement by statement, against a spreadsheet that only one person fully understood.
What we configured
A mainstream CRM, not an insurance-specific platform. The specialist products were better out of the box and worse everywhere else, and the gap was closable with configuration. The structure that mattered:
| What a generic CRM assumes | What the brokerage actually needed |
|---|---|
| Deals with stages and a close date | Policies with inception and renewal dates, and a renewal workflow that starts automatically at ninety days |
| One contact per company | A client group with subsidiaries, plus the broker, the insurer, the underwriter and the claims handler on every record |
| Notes as free text | A structured record of advice attached to the policy, with the recommendation and anything the client declined |
| Forecast by weighted pipeline | Forecast by renewal calendar and commission rate, which is far more accurate and needs no guessing |
| Activity metrics per salesperson | Retention and lapse per broker, per class of business, which is the number the business actually lives on |
On top of that: email and calendar sync so correspondence files itself against the client, a document store for schedules and records of advice, a renewal workflow with tasks at ninety, sixty and thirty days, and a monthly import of insurer commission statements matched to policies so the reconciliation stopped being manual.
Give people a reason before you give them a login
The brokers did not adopt the CRM because management asked. They adopted it because the first thing it did was produce their own renewal list, ninety days out, sorted by value, with the ones nobody had touched at the top. Build the thing that makes somebody's own week easier first, and the rest of the data capture follows it in.
Ten weeks, in order
| Weeks | Phase | What actually happened |
|---|---|---|
| 1–2 | Define a client | Harder than it sounds. Groups, subsidiaries, trusts and the same person appearing four times. The rules were agreed before anything was imported. |
| 2–4 | Configure | Policy and claim objects, renewal workflow, records of advice, permissions so brokers see the whole book but only own their own. |
| 3–6 | Data | Policies from the administration system, renewal dates from eleven Outlook calendars, contacts deduplicated from 4,100 rows to 2,600. |
| 6 | Pilot | Two brokers, chosen because one was enthusiastic and one was the loudest sceptic in the firm. The sceptic found four real problems. |
| 7–8 | Train | Two hours each, in pairs, on their own book rather than on demo data. |
| 9 | Live | Outlook renewal reminders switched off deliberately, so there was one diary rather than two. |
| 10 | The meeting moves | The weekly renewal meeting started being run off the system's own report. That, more than training, is what made it permanent. |
The part that was actually difficult
Brokers do not resist CRM because they dislike software. They resist because a documented client book changes the balance of power in a firm where relationships are the asset. Three things defused it.
The first was saying it out loud. The managing director opened the training by acknowledging that the book is built by the brokers, and committing that the system would not be used to reassign clients or to measure activity for its own sake. The second was permissions that matched that promise. The third was making the system genuinely useful to the individual before it was useful to management: the renewal list, the claim status without phoning the administrator, the client history when covering for a colleague on leave.
Two brokers still kept a private spreadsheet for the first six months. That is normal, and it is not worth a fight. Both stopped when the system's version became better than theirs.
What it cost
| Line | Once-off | Notes |
|---|---|---|
| Configuration and workflow build | R208,500 | Policy and claim structures, renewal automation, permissions |
| Data migration and deduplication | R74,800 | Administration system, eleven calendars, two spreadsheets |
| Commission statement import | R53,600 | Four insurer formats, matched to policies |
| Training and pilot support | R37,400 | Paired sessions, then floor support through first renewals |
| Total once-off | R374,300 | Over ten weeks |
| Licences and support | R11,380 / month | Sixteen users at mid-tier, plus support retainer |
For scale: a five-person firm on a standard CRM with light configuration lands nearer R80,000, and a large brokerage with a specialist platform and regulatory reporting lands several times higher.
What changed, measured
- Renewals worked more than sixty days out went from 38% to 90%, which was the whole point.
- Retention improved from 84% to 91% over the following year. On a book this size, that single change paid for the project several times over.
- A broker handover went from four months to a day, tested unhappily but successfully when somebody resigned eight months later.
- Records of advice were attached to essentially all new business, from a starting point nobody could measure.
- Commission reconciliation went from three days a month to four hours, and found R190,400 a year in commission that had simply never been paid.
- Cross-sell became a list rather than an intention. The first campaign, on a single missing cover across the commercial book, produced more new premium in two months than the previous year of ad-hoc effort.
What we would do differently
We would settle the definition of a client before touching data, not during. A week of arguments about groups, subsidiaries and trusts up front would have saved three weeks of rework.
We would also import far less email history. Nine years of correspondence was pulled in because it was technically possible. Two years would have been more than enough, and the rest simply makes the search results worse.
And we would appoint the data owner on day one. The firm eventually gave one administrator explicit authority over record quality, and the difference was immediate. Before that, everybody owned the data, which means nobody did.
If your book lives in people's heads
Start with the renewal calendar. Get every date, every policy and every owner into one place, and run one meeting off it. That alone will change how the next twelve months go, and it will show you what the rest of the system needs to do. The general case for this, if you are earlier in the decision, is in do you need a CRM yet, and the reason so many of these projects stall is covered in adoption is the project.
