Which channels produced bound premium?

Comparison sites and aggregators self-report performance, and those figures are rarely reconciled against the policies actually bound in the carrier's own system.

Adding the aggregator lead export to the lead file and matching it against bound premium gives an independent count. It may well be the first neutral measurement of that spend the carrier has had.

Two channels binding premium at the same acquisition cost are not equivalent if one produces policyholders who renew for eight years and the other produces annual shoppers.

Because the match is on the policyholder, every renewal credits the original channel — and once retention is in the ranking, the cheapest acquisition channel is frequently the worst one.

A carrier writing through appointed agents while also advertising direct has two acquisition motions pointed at overlapping markets. A policyholder who saw a television campaign and then bought through a local agent has been produced by both of them, and the internal argument about who gets the credit is generally settled by whoever controls the reporting rather than by anything resembling evidence from the carrier's own records. Both motions are expensive, both are defended internally, and neither has ever been measured against the other on the same basis.

Combining both sides into one lead file — the agency book with contact details and the direct lead exports, each row tagged with its source — puts them into one ranking on bound premium. Where a policyholder appears on both sides, the policy is credited automatically to whichever record has the earlier date; it is not split, and it is not held for a person. Sizing the overlap itself is a separate import: agent-written policies as the sales file against the direct inquiries alone as the lead file, so the matched share on that import's own results is the part of the agency book that also inquired direct; delete that import afterwards so the workspace's reports do not count those policies twice. That is a smaller answer than a model would offer and a considerably more useful one, because it counts the overlap instead of assuming it.

Sizing that overlap tends to change how brand spend gets discussed at the carrier. Finding that a substantial share of agent-written policyholders also inquired direct is evidence that the advertising supports the agency channel rather than competing with it. Finding almost no overlap means the two are genuinely separate markets, which argues for two separate budgets and two separate sets of expectations rather than one contested number. Either finding is more useful than the annual disagreement about whether brand advertising helps the agents or competes with them.

A rate filing takes effect and written premium per policy rises across the entire book. Every channel's revenue figure rises with it, and a report ranking on premium alone will read a pricing decision as a marketing improvement. In a segment where rate changes are frequent and occasionally large, this is the single most likely way for an attribution report to mislead the people who commissioned it and to do so confidently. The rate went up, the revenue went up, and nothing about the advertising changed in the slightest.

Policy count is the control. The report shows both figures, so a channel binding the same number of policies at a higher average premium is visibly a rate effect, while a channel whose policy count grew has done something genuinely different. Comparing a period against the same period a year earlier rather than the preceding quarter removes most of the seasonal noise that would otherwise sit on top of the rate movement. Policy count is the figure that does not move when the actuaries do, which is exactly what makes it useful here.

Mix does the same thing more quietly and is easier to miss. A shift toward a higher-premium line, or toward a state where rates are structurally higher, lifts a channel's average without any change in how it performs. Mapping line of business as Product / service and state as Location shows how premium and its attributed share move across lines and states, which makes a mix shift visible though it does not rank channels within each, and both cost nothing to add because the policy record already carries them. Mix effects are the quietest way for a channel to look better than it is, and among the easiest to check.

A policy is not a fixed object. A vehicle is added, a young driver is rated onto the household, a property is refinanced and the coverage changes, a second policy is written for the same address. Each of those raises the premium written against a policyholder who was acquired once, through one channel, in a campaign period that closed long ago and that nobody is still reporting on. Each of them is new premium written against an existing relationship rather than new acquisition, which is precisely why they belong to whichever channel produced the household.

Because the match is on the policyholder rather than on the policy, those later transactions credit the acquiring channel when they are exported as revenue rows against the same contact. The ranking moves from a measure of new policies written into a measure of premium written against the people each channel brought in, which is a quantity much closer to what the carrier's growth plan is actually built on. For a carrier whose growth comes largely from the book it already holds, that distinction is the whole point.

There is a practical consequence for the export. Pull premium transactions by policyholder and period rather than a single figure taken at bind, and the report handles endorsements, additional policies and renewals without any of them needing to be labeled as such. Where a transaction belongs to a policyholder who never appeared in a lead file, it stays unattributed, which keeps the attributed share honest rather than complete. It is a smaller number than a projection would give, and it is the one that can be checked against the ledger.