Where your sold business comes from

You asked a simple question: which marketing produces sold business? Here is the honest answer, measured directly from your HubSpot. Nothing in this report changes or risks any of your records.

By channel (the website the lead came through):

The report makes the gap visible, but it cannot recover it. The lever is fixing how the source is recorded going forward.

The marketing source for 65% of your sold business was never captured at the time, so there is no honest way to reconstruct it after the fact. We will not guess, because a guessed source is worse than an honest unknown: it would send you spending decisions built on invented data.

The painful part is cbcins. Every cbcins sale is source-blind, and cbcins is your best-converting channel: it turns quotes into sales at 11.6% versus 1.8% on Costcoquote, roughly six to one. You are most effective on the channel you can measure the least. The measurement audit indicates paid social (largely Meta) drives much of cbcins traffic, so you are likely spending there and cannot yet tell which of that spend produces sold business.

So the real lever is not normalizing the past, it is fixing capture going forward, so that next year's sales, especially cbcins, carry their source. The report is the scoreboard. The capture fix is the game.

Within the traceable 35%, two Costco assets do 84% of the work.

Your Costco Business Health Insurance page and a Costco email campaign are 84% of all traceable sold premium. That is evidence you can lean on. Google and the AI assistants are real but tiny so far (6 sales). Read the credit as last-touch, not total: under this model Costco gets about 98% of the traceable credit; a first-touch view would likely shift some toward whatever drives initial awareness.

Everything above is measured, not modeled. Here is the basis.

• Total sold (all-time): 1,386 won deals / $876,600 premium floor, measured live on 2026-06-22. Reconciles to your deal-side numbers ledger to the dollar, plus the few deals that have closed since the prior pull.
• The traceable set is small and clean. Among won deals only 14 distinct recorded source values appear, and every one maps cleanly to a category. No spam or junk value reached a sale.
• Channel split (ties exactly): Costcoquote 832 / $590,463 (618 with no source), Cbcins 510 / $184,000 (all 510 blind), Other 44 / $102,136 (all blind). 832 + 510 + 44 = 1,386; the dollars sum to $876,600.
• The conversion gap (the strategic point): quote-to-sold is 11.6% on cbcins (510 of 4,393 quoted) versus 1.8% on Costcoquote (832 of 46,202 quoted). This is the single biggest reason to fix cbcins capture first.
• Why history is blind (plain terms): your sales records are created by your own system on the server after the visitor has left the page. The marketing source lives in the visitor's browser session, and the system does not carry it through to the record. HubSpot's built-in source field reads "Offline" on 99.96% of your contacts. This is fixable, but only going forward.
• A caution on the dollars: premium is recorded on only about 38% of won deals, so every dollar here is a floor and undercounts. The counts are reliable.

Three pieces. None of them changes or risks a single one of your existing records.

1. The attribution report (visibility). A normalized "marketing source" field now sits in your HubSpot and sorts every sale into clean source buckets. It is read-only, writes nothing, and is removable anytime. This scoreboard is built on it.
2. The going-forward capture fix (the game). A change to how your quote engine records a lead, so the marketing source is carried onto the record at submission, including for cbcins (which has no tracking installed today). Prepared from a read-only copy of your code; your team deploys it.
3. The tagging standard and tool (stop new gaps). A fixed vocabulary for tagging marketing links, plus a link-builder tool, so the team stops creating new untracked links. The tool is built and hosted.

You cannot optimize what you cannot measure. The fix puts you in position to move budget toward what actually sells.

• Your best-converting channel is your blindest one. Making cbcins measurable is the highest-value single improvement, because it lets you put money behind a channel that already sells six times better.
• The two Costco assets that work are now provable. 84% of traceable sold premium runs through the Costco Business Health Insurance page and a Costco email campaign.
• We are not putting a dollar figure on the upside yet, because an honest one depends on data we are still building. We will add it here, sourced, once going-forward capture has run long enough to measure.

What we need from CBC (minimal)

• For the report: nothing. We build it; your team reviews it.
• For the capture fix: your developers deploy the prepared change through your normal process, and confirm where cbcins form submissions currently land.
• For the tagging standard: name one person to own the approved tag list going forward.

• Live and read-only. All numbers come from live read-only queries against your HubSpot (portal 20781567), run 2026-06-22. No data was changed. The normalized source field that powers this scoreboard is live in your portal, so a fresh pull updates this picture on demand.
• Scope: your HubSpot view. Every figure is measured from HubSpot. If your team also records final sold outcomes in another system (for example Monday), those are not reflected here, so this is your HubSpot sold-by-source picture, not necessarily your entire sold book. Reconciling the two is a separate step we would flag, never silently fold in.
• Money is premium, a floor, not revenue. The deal Amount is policy premium volume, recorded on about 38% of won deals. Counts are reliable; dollar totals undercount. Agency revenue is a commission on premium at a rate not reflected here.
• Channel means the website the lead came through (Costcoquote versus cbcins), read from the quote link on the deal, not the internal pipeline name.
• The attribution model, named: single-source, last-non-direct, deal-level. HubSpot's native multi-touch revenue attribution (Marketing Hub Enterprise) is not in place, so this answers "the last marketing touch on record," not "every touch that contributed."
• Unknown means unknown. Where a sale's source was never recorded, we report it as unknown rather than guess.

Totals reconcile to your deal-side numbers ledger to the dollar; the small day-to-day difference from a prior pull is normal closings, explained in our working notes.