Built on Genna's signal model, with what we can actually build today.
Three campaigns are live. Finance and Quality sent their first email to 37 leads today. That is the pre-release wave, and nothing here gates it.
Genna's nine signals are a tool for what we build from here: the signal columns we stand up in Clay, and how we pick who enters a lemlist campaign next. Sourcing machinery, not a review of what is already running.
What is added below is which of the nine we can actually compute, three places the arithmetic needs a decision, and the angle each cohort carries.
Nate's Finance, Quality and Resurrection campaigns hold 72 leads. Every one was selected on a contract under 4.0 with a gap in one of the six service measures that survived the CMS recalculation. The 13 leads whose plans cleared 4.0 on the July re-issue came out this morning.
Finance and Quality have sent Email 1 to all 37 of their leads, split across both A/B variants, all delivered and none bounced. Resurrection's 35 are loaded and have sent nothing, because its first two steps are manual LinkedIn actions waiting on Nate. Opens read zero on all three, which is the tracking domain still not being configured rather than a result.
Under the scoring model those 72 are the measure-drop and near-the-line cohorts, already assembled. What is still owed on them is everyone signing off on the copy rather than Nate alone, and that sign-off now governs Resurrection and every wave after it.
The model as proposed: strong signals 3 points, moderate 2, supporting 1, plus 2 when two signals land together. Tier 1 at 8 or more, Tier 2 at 5 to 6, Tier 3 at 3 to 4, low priority under 3. ICP fit stays separate from the score.
Nothing catches a 7. The bands are 8 and up for Tier 1, 5 to 6 for Tier 2, 3 to 4 for Tier 3. An account scoring exactly 7 lands in none of them.
That is not a corner case. Scoring the six named combinations against the scale, four of them come to exactly 7:
| Combination | The arithmetic | Lands in |
|---|---|---|
| New exec and a declining rating | 2 moderate, 3 strong, 2 for the pair | 7, no tier |
| New exec and a 3.0 to 3.5 rating | 2 moderate, 3 strong, 2 for the pair | 7, no tier |
| A 3.5 rating and a large Medicare book | 3 strong, 2 moderate, 2 for the pair | 7, no tier |
| A rating decline across several contracts | 3 strong, 2 moderate, 2 for the pair | 7, no tier |
| A rating decline while a peer improved | 3 strong, 3 strong, 2 for the pair | 8, Tier 1 |
| A public statement and a poor rating | 3 strong, 3 strong, 2 for the pair | 8, Tier 1 |
Only the two pairs made of two strong signals clear to 8. Every pair built on a moderate signal, which is four of the six, scores 7 and falls out of the model. Proposed fix, Tier 2 covers 5 to 7. That closes the gap without promoting a two-signal account into the top tier, which should stay for accounts carrying three real signals.
The bonus fires on two different rules. The summary says the bonus is for two strong signals together, but four of the six named combinations pair a moderate with a strong, since a new executive is a moderate. Proposed fix, the bonus fires on any pair from the named combination list.
Five supporting signals make a Tier 2 with no trigger in it. Relevant executive identified, recent hiring, large company, significant Medicare presence and recent restructuring are 1 point each and close to always true for a large payer. Five of them is a 5. Proposed fix, supporting signals cannot lift an account above Tier 3 on their own.
Asking whether any contract qualifies turns the score into a measure of size. This one only appeared once the model was actually run. Scored as written, the seven largest payers all tied at 22, because a parent running 83 contracts always has one that declined, one sitting at 3.5 and one whose peers improved. The fix is to evaluate the per-contract signals on the read contract, the single contract the account would actually be written about, which is the same contract the measure read already names. Enrollment, contract counts and growth stay at parent level, since those genuinely are parent facts. With that change the largest payers fall back into the field and mid-size plans with a real gap rise.
All nine are now built. Four came from files we already held, three from the CMS county enrollment file pulled this afternoon, one is partial for structural reasons, and the executive signal went in this afternoon too.
| Signal | Build it | Where it comes from |
|---|---|---|
| Deterioration in a specific measure | Yes | 12,240 contract by measure rows. 460 of 800 contracts have a service measure down, and the parent-grain read is send-ready |
| Rating declined year over year | Yes | 120 of 800 contracts. Correct against the Jul 22 re-issue first, it moved 34 |
| Rating is 3.0 to 3.5 | Yes | 264 of 800 contracts |
| Multiple contracts below 4 stars | Yes | 26 parents have two or more. Centene 42, Elevance 27, UnitedHealth 24, Humana 24 |
| Large Medicare Advantage enrollment | Yes, new today | Coverage went from 297 of 800 contracts to 709. The 91 without one are not in the CMS file at all |
| Enrollment growth | Yes, new today | July 2025 against July 2026. 88 parents have a read solid enough to say out loud |
| A peer improved while they declined | Yes, new today | 26 contracts fell while the plans competing for the same members in the same counties rose |
| Public statement on Stars or the bonus | Partly | Roughly 8 of our 93 parents are publicly traded. The rest are Blues, mutuals, county authorities and nonprofit systems |
| New Medicare or quality executive | Built today, in Clay | A published Clay workflow: find the leaders at a domain by title, then keep only those whose current role started inside the window. Ran across all 72 accounts. 9 carry a recent hire, 13 hires in total |
Two things in that enrollment data would have produced bad emails. Contract-level growth is contaminated by members migrating between contracts inside a parent, which is why one Humana contract reads as 13,000 percent growth, so growth is parent grain with a floor. And a peer comparison that only asks whether more plans rose than fell fires on a 68 to 56 split, so a peer now has to share the counties holding 80 percent of a contract's members, with a real margin.
One caveat sits under all of it. All 19 measures we track are marked unknown for the October release, because the Clover ruling and the CMS recalculation pulled a set of operational measures out of the 2027 bonus and CMS took no position on what October uses.
Scored across 182 parent organisations with the four fixes applied:
| Tier | Accounts |
|---|---|
| Tier 1 | 49 |
| Tier 2 | 32 |
| Tier 3 | 22 |
| Low priority | 70 |
| Excluded as current customers | 9 |
Two things worth looking at. The customer gate caught nine, including Humana, which scored 22 before the gate ran and would have been the single highest-scoring account in a cold campaign. That is the case for keeping exclusion outside the score rather than inside it.
And Tier 1 holds 49 of 182, a quarter of the universe. Three signals gets an account to 8, and a plan under 4 stars with a measure gap trips three almost by definition. That is the model behaving as specified, not a defect, but a top tier holding a quarter of the field is not a priority list. Raising the Tier 1 floor is Genna's call, not something to change quietly.
Six accounts score exactly 7, so under the bands as proposed they would have had no tier at all.
The executive signal is a published Clay workflow and it has run across all 72 accounts. Nine carry a leader who started inside the last year, 13 hires in total, out of roughly 300 leaders looked at. The strongest is Blue Shield of California, 4 recent hires among 9 leaders including a Chief Operating Officer who started in August. Nine of seventy-two is the point: a flag that fires everywhere is not a signal. Domains for all 72 were resolved in Clay first, which is what had been blocking this.
This is what the signals become once they are columns in Clay: six ways an account can qualify, each its own campaign. Each carries one idea, and persona changes the framing inside a cohort, never the idea. The trigger decides why we write today; the measure-level read is the payload in all six, because the overall rating is public and the measure read is work.
A service measure dropped. A durable service measure moved on their largest contract, whether or not the overall rating did. Quality and Stars seats get the measure math, Finance gets the same read against the bonus line.
Close to 4 stars with a big Medicare book. Rating at 3.0 to 3.5 with large Medicare Advantage enrollment. The half-star is worth more at their enrollment than at almost anyone else's, so the same fix pays differently here. Finance leads this one.
A new Medicare or quality exec. Hired in the last 6 to 12 months. This is the one cohort here that is not really about Stars. A new operations, service or back-office leader is a why-now for every motion we run, so the signal should be built once and pointed at all of them, with Stars taking its slice. The listening tour ends and the agenda gets written, and the service measures are the part of that agenda that needs no restructure. The email cannot open on the rating. They inherited it, they already know, and opening there makes us the fifth vendor that week to tell them their score. Congratulate, name the moment, offer the first-year win, and let the measure read sit second.
They fell while their market rose. Their rating fell while the plans competing for the same members rose. What changed is not the market. Quality seats. Frame it as the field moving differently for different plans, never as falling behind.
They've talked about Stars publicly. A public executive statement names Stars, the bonus or Medicare Advantage economics, so the email answers a question they already put on the record. Not an earnings-call cohort: for the Blues, mutuals and county plans the surface is a press release, a rate filing or a state insurance filing.
Growing faster than service can keep up. Growth becomes contact volume, and the member-experience scores are set in the weeks when volume peaks. Operations and service seats, not quality.
Customer exclusion sits outside the score. A current customer can score 11 and must never enter a cold campaign, it routes to the account manager. This matters more with scoring than without it, because scoring surfaces the largest and most exposed plans first, and those are the likeliest to already be customers.
Every signal here is a public fact with a date. Each ships with its primary source and an as-of date, re-checked at send time because CMS re-issues. Nothing in this model is an Intradiem claim.
Every Stars number says they know their own numbers better than we do. The number is proof we did the reading, never the argument.