Premium leaves an MGA without a moment you could point at. Nothing gets declined. No broker writes to say they've placed it elsewhere. A bind instruction waits two days, the broker confirms with another market, and the thread stops. The CRM still shows an active relationship. The quarter still shows growth.
What follows is how that works mechanically, and what the one piece of evidence we have does and doesn't tell you about it.
Why lost premium doesn't look like loss
Distribution reporting is built around things that happened. Submissions received, quotes issued, binds, renewals retained. There's no field for business that would have bound if a thread had been answered on Tuesday rather than Friday.
A declined risk is a decision. Somebody weighed it, said no, and left a record. A signal nobody got to produces nothing at all: no exception report, no flag, no line in the monthly review. In the books it's indistinguishable from business that was never on offer.
So the loss shows up somewhere else, later, as a competitor's growth, and by then it's attributed to rate or appetite or a relationship that "went cold", none of which is checkable.
Four ways it goes
The unanswered bind instruction. The highest-intent message a broker can send arrives looking like everything else and competes with everything that has landed since. Every hour it waits the odds shift, and when the broker confirms elsewhere they generally don't mention it. The thread just ends. A broker sent you something and expected an answer, and nobody replied. Before that's a data problem it's a rude one.
The slow quote turnaround. Behind the quote request is the broker's own client, waiting. A three-day turnaround doesn't only risk that risk. It teaches the broker something about your capacity, and they don't complain, they recalibrate. The next one goes first to the market that came back in a day.
The renewal reopen nobody picked up. A broker reopens a renewal thread with a mid-term change, a pricing check or an appetite question, and gets nothing. Renewals are where incumbency does the selling for you, so an unanswered reopen is the quickest way to spend that advantage. Worth saying that renewal season is also when half the desk is on holiday, which is a real constraint and not one anybody has solved.
Inbox drift. A producer's traffic shifts over a few months from submissions and bind instructions to admin queries, chasers and circulars. The relationship looks alive because mail still arrives. The commercial substance left in the spring. Nobody decided anything and the mix just moved.
Why nobody notices
Four reasons, none of them about effort.
- Signals are spread across several underwriters' mailboxes, so each inbox holds a slice of a producer and nobody holds the total. That mechanism on its own is worth an article.
- Email surfaces the most recent message rather than the most valuable one. At volume, recency wins by default, because recency is the only thing an inbox knows how to sort on.
- The CRM records outcomes. There's no way to log an email nobody acted on, so the gap is invisible in the one system built to hold the relationship.
- Watching the space between a signal and a response isn't in anybody's job description, and it isn't obvious whose it should be.
An underwriter carrying a full technical workload can't also run a watching brief across three colleagues' inboxes. Nobody can hold sixty producer relationships and their normal rhythms in their head, and it isn't a reasonable thing to ask. The bordereau arriving late gets chased because somebody owns chasing it. The thread that went quiet is nobody's chase.
What we actually know about how big this is
Less than we'd like, and it's worth being straight about the size of it.
We ran one controlled four-week test against a real MGA's broker traffic. Nine inboxes, 47 brokers active in the window, and 27 threads carrying submission or bind intent with no clear follow-up.
That's one test, on one book, over four weeks. It wasn't a pilot and it wasn't a client engagement. We don't know what those 27 threads were worth in premium, and we don't know how many of them would have gone anywhere if someone had picked them up, because we didn't follow them. Twenty-seven high-intent threads with no follow-up in a month isn't a rounding error, but anyone quoting it as a conversion rate is quoting something we don't have. Our first proper pilots run in August and September, and we'll publish what they show, including if it's less than this.
The absence is the hard part. A firm with real history and a quiet six weeks generates no email to flag it, so the digest carries it as a coverage gap. What it cannot tell you is what that firm was worth.
Three numbers that make it visible
You can't manage the counterfactual. You can manage the window in which it gets decided, and that takes three things, weekly.
- Signals received, by producer. How many commercial signals arrived this week and from whom, separated from the admin traffic. Most operations have never seen this number.
- Engagement trend against each producer's own baseline. Which relationships are running above their own normal and which have dropped below it. Against their own history, not against each other, because a firm that sends you two threads a month is not underperforming a firm that sends twelve.
- How long the broker waited. On the high-intent threads specifically. Aggregate response time is a comforting number and a fairly useless one, because it's dominated by the easy mail.
The 27 threads in that test weren't found by interviewing anyone or auditing the CRM. They were found by reading what the inboxes already held, the metadata of what arrived and what went back, assembled into one view. That's the product we build at BindSignal, for transparency. The three numbers stand on their own.
Something to check this week
Pick a fortnight from last quarter, far enough back that the outcomes are known. Find every bind instruction and every quote request that came in during those two weeks, across whichever inboxes you can get at. For each one, find the first substantive reply and note the gap in days.
Then look at the ones with no reply at all, and check what happened to that account.
It'll take an afternoon and it's genuinely tedious. What it will tell you is whether the gap in your operation is hours or days. What it won't tell you is what the slow ones cost, and no exercise will, which is the whole difficulty with this category of problem.