The unbound quote exists in the rater and nowhere else

Two systems, and only one of them is reported from

An independent agency runs a comparative rater and an agency management system. The rater takes applicant data once and returns carrier options. The agency management system holds the account, the policies, the activity history and every report the principal actually reads. The division of labor is deliberate. Its consequence is not: the earliest and least durable state in the whole revenue path is created inside the system nobody reports from. A quote is state. It records that a named applicant, at a named moment, supplied enough information to be rated, and that a set of carrier options came back at a set of prices. If nothing carries that state across, the applicant exists in the rater and nowhere else — no account to attach a task to, no activity to age, no status field an automation could watch, and no row for a report to count.

A missing object is not a slow follow-up

These two failures look alike from outside the business and behave nothing alike inside it. A slow follow-up is a task that exists and is worked late: it sits in a queue, it can be reassigned, its age can be measured, and a cadence tool can act on it. A missing object cannot be late, because nothing is waiting on anything. Every intervention aimed at speed — a sequence, a reminder, a dialler, an agent — takes records as its input. Where the record was never created, adding cadence tooling changes nothing, and the spend on it produces no observable effect for reasons that have nothing to do with the tool. The reporting consequence follows directly. A quote-to-bind ratio needs both terms measured over the same population. The bound policies sit in the agency management system. The quotes sit in the rater. Computing the ratio from the agency management system alone quietly substitutes a different denominator — the quotes that happened to get logged by hand — so the resulting figure describes logging behaviour rather than sales behaviour. It is the same shape of error as any denominator drawn from a system that only ever saw the successes, and it moves in the wrong direction: the less that gets logged, the healthier the ratio looks.

Why the ad platform's conversion count is not a substitute

A tempting shortcut is to treat the advertising platform's conversion count on the quote-request form as the quote population. It is a different object. Google documents that when users deny consent for storage, consent-aware tags send measurements without cookies, and that Google products use those pings to model your metrics ^[SRC-18] — reported conversions are partly modelled, by design and by documentation. Separately, Google has stated that Chrome will maintain its current approach to offering users third-party cookie choice, and has listed the Attribution Reporting API among the technologies being retired ^[SRC-16]. Neither of those is a defect. Both mean the browser and ad-platform layer is not where a per-applicant operational record can be reconstructed. A quote is an operational object and belongs in an operational system, with a primary key, a status and an owner.

What a write-back would actually have to do

Closing this is an integration problem before it is an automation problem, and send the quote to the CRM is not a specification. The following are separate obligations. A build that satisfies some of them and not the rest produces a record that still cannot be worked:

1. Identity first. Create or match an account at the moment of rating, on a deterministic key both systems hold, rather than reconciling on name and date afterwards. Every later join inherits whatever this step decides, including the commission reconciliation years later.
2. The quote as a durable object. Write the quote with a status, an effective or expiry date, the carriers rated, the lines quoted and the producer responsible — not as a free-text note stapled to a contact.
3. An owner and an ageing threshold. Generate a task or opportunity assigned to a named person, with a threshold that makes an untouched quote surface as an exception rather than merely become old.
4. An explicit terminal state. Record why the quote stopped moving — bound, declined, rated for comparison, ineligible, no response — so the population can be closed rather than left permanently open and permanently ambiguous.
5. Source preservation. Carry the acquisition source onto the account so the policy that eventually binds can be joined back to the inquiry that produced it, rather than to whichever dropdown value the last person edited.

Sequencing matters more than it appears to. Where a vendor gates production API access behind a developer agreement and an approval step, the critical path for this work runs partly through someone else's calendar, and a plan that treats it as pure engineering effort will estimate the wrong dependency. That makes it orchestration work rather than scripting — see Integration & Orchestration Systems for where the capability sits, and the insurance agency revenue architecture page for the surrounding leak surface, including the download suspense queue and commission reconciliation, which depend on the same identity decision made here.

Following up is a permission question before it is a cadence question

Once unbound quotes exist as records with owners, the question stops being whether to follow up and becomes how contact is permitted to be made. The FCC delivery restrictions at 47 CFR 64.1200 require prior express written consent for autodialed or prerecorded telemarketing calls and text messages, treat revocation as valid when made by any reasonable method to clearly express a desire not to receive further calls or text messages, require revocation to be honored within a reasonable time not to exceed ten business days, and require national do-not-call scrubs against registry data no more than 31 days old ^[SRC-14]. Whether a particular quote-request form captured consent of that kind is a question about that form and that record. The structural point is narrower and worth stating on its own: consent captured at the rater and consent visible to the follow-up system are different things unless something carries it across. If the write-back moves the applicant but not the consent state, the follow-up system holds a record whose permission status is unknown to it. That is a field-level decision made at integration time, and it is cheaper to make then than to reconstruct later. Interpretation of the rules themselves belongs with qualified counsel.

The reconciliation that settles it

Export every quote the rater produced across a trailing period with applicant name, date, lines and carriers rated. Export accounts, activities and opportunities created in the agency management system across the same window. Attempt a record-level match. Three counts fall out of it: quotes with a matching account, quotes with an account but no activity and no owner, and quotes with nothing at all. The third group is the population under discussion, and its size is a fact about record creation — exactly the kind of fact an export can establish without anyone having to estimate anything.

• An unmatched quote is not a lost sale. The same applicant may have been rated more than once, quoted for comparison against a policy they intended to keep, found ineligible under carrier appetite, or bound under a name the match cannot see.
• A completed write-back does not create demand. It makes an existing population addressable and countable; whether any of it converts is a separate question no record can answer in advance.
• Nothing here establishes the incidence of this configuration or what it is worth. No verified benchmark for quote volume, bind rate or unbound-quote value was available, and inventing one would be worse than leaving the question open.

Before buying follow-up tooling, run the export and the match. If the records exist and are simply worked late, that is a cadence problem and cadence tooling is the right purchase. If they do not exist, no amount of cadence will find them. Request a Revenue Scan to see what the public side of a quote-request path shows, or read Revenue OS for how a confirmed finding maps to the system that would own the fix.