Take one broker firm and follow it through a month. This is a worked example rather than a client, but the parts are ordinary enough that most of it will be familiar.
The firm has four partners and a dozen account executives, and it sends you property, casualty and the occasional one-off. You write three of their lines. Kate handles the property, Dan the casualty, and Priya picked up a builder's risk for them last autumn and has been their default for anything odd ever since.
March, one producer, three inboxes
Week one. A property submission lands with Kate. Two messages, a schedule attached, answered the same afternoon. Nothing about it is interesting.
Week two. A casualty enquiry to Dan, which turns into a quote request on the Thursday and a quote out on the Friday. Meanwhile Kate gets a follow-on question about the week one schedule and a separate submission on a small retail risk.
Week three. Priya gets the one-off, contractors all-risks on a conversion, sent to her because she quoted something like it in October. Kate gets a chase on the retail risk. Dan gets a chase on the casualty quote, then a second one two days later from a different account executive at the same firm, who doesn't know the first one was sent.
Week four. The bind instruction on the casualty arrives with Dan on the Tuesday. Kate gets a renewal thread reopened on an account that went quiet in January. Priya gets two more messages on the conversion, one of which is really a coverage question and should probably have gone to Dan.
Fourteen threads across the month. Kate saw five of them, Dan six, Priya three.
Three honest reports, one wrong answer
At the month-end review, all three are asked how that firm is doing.
Kate says steady. Five threads, nothing unusual, they've been about that busy since the spring. Dan says busier than normal but he's had the bind so it's fine. Priya says she's had a couple of bits and doesn't really cover them.
Every one of those answers is accurate. The firm was the most active producer you had that month and nobody in the room knows it, because the number that would tell them is fourteen and the largest number anyone can see is six.
The brokers are routing correctly
Fragmentation happens because the routing is working. Property goes to whoever writes property. The odd risk goes to whoever handled the last odd risk. The account executive emails the person who answered fast in October and has no idea, or reason to care, how you've divided the desk.
Which means the distortion is worst exactly where it costs most. Your single-line producers, one contact and one inbox, are perfectly legible. The multi-line relationships that carry the book are the ones nobody can see whole.
It also gets worse the better your distribution gets, which is an uncomfortable thing for us to have to say out loud, given what we sell. A firm sending you three lines is a firm you've won properly. The reward for winning it is that it becomes harder to read.
What fragmentation does to prioritisation
Your busiest producers get middling attention everywhere. Attention gets allocated inbox by inbox, so a firm is ranked by the slice each person can see. Five threads gets five threads' worth of urgency, in three places at once.
A decline shows up as normal variation. Fourteen threads a month becomes six and each inbox sees its usual trickle get a bit thinner, comfortably inside a normal month. The drop is only visible in the total. This is the mechanism behind how premium leaves without anyone noticing.
Ownership dissolves at the seams. The coverage question that landed with Priya in week four is nobody's, technically. So is a bind instruction copied to two underwriters who each assume the other has it. We don't have a clean answer to that second one, for what it's worth. Seeing the thread whole tells you it's open. It doesn't tell you which of the two of them was going to pick it up.
None of this is fixable by the people inside it. An underwriter can be perfectly responsive to everything in front of them and still be wrong about a producer's priority, because the information needed to be right was never in their inbox. Asking them to cross-check with two colleagues before deciding what matters is asking for a data-integration job on top of a full technical workload.
Why the usual fixes don't hold
The shared mailbox. Route everything through submissions@ and fragmentation goes away, in theory. In practice brokers don't cooperate, because the personal relationship is the entire point. The shared mailbox collects circulars and the occasional first-time submission while the real signals keep going to the person the broker actually knows. You can't process-manage your way out of something your brokers are actively choosing.
CRM logging. If everyone logged every touch the CRM would hold the total. That fails for reasons worth a separate article. And even at implausibly good compliance, the CRM records what got handled. A thread nobody acted on generates no entry, so the signals you most need are the ones least likely to be there.
The monthly review. Which is where we came in. Going round the table collects impressions, and fragmentation is the exact condition under which three accurate impressions add up to a wrong answer.
Seeing the producer whole
What would have to be true to fix it is that somebody, or something, groups the traffic by the broker firm rather than by the inbox it landed in. Every one of those fourteen threads carries the firm's domain on it. Roll them up by that domain across every mailbox you monitor and the total exists, without anyone changing how they work and without asking brokers to route differently.
Three questions become answerable at that point. Which producers sent the most high-intent signals this week across all inboxes combined. Which look quiet in every individual inbox and substantial in the total. And which high-intent threads have had no reply from anyone at all, which is a different question from whether each person answered their own mail.
For transparency, assembling that from Microsoft 365 email metadata, no message content and no body text, and sending it as a weekly email is the product we build at BindSignal. The diagnosis holds without it.
Kate’s five, Dan’s six and Priya’s three, on one line. The total that no individual inbox can show, assembled by grouping on the broker’s own email domain.
Something to check this week
Pick your three largest producing firms by premium. For each one, search your own mail on the firm's email domain rather than on a person's name, and count the threads from the last month. Then ask the other underwriters who deal with that firm to do the same, and add the three numbers up.
Twenty minutes, most of it waiting for colleagues. If your total is close to what you'd have guessed, fragmentation isn't costing you much and you can stop reading us. If it's roughly double, you've been ranking your best relationship on a third of the evidence.
A month is a short window to call anything, mind. Do it for two.