The 12 Signals That Predict Medicare Plan Change — and the 4 That Agents Waste Time On

The T65 signal is real — but overstated

Turning 65 is the most widely used lead signal in Medicare distribution. It's a genuine trigger: this person is newly eligible and needs to make a decision. But it's not as predictive as agencies treat it.

T65 volume in most markets is large enough that agents who work only T65 leads are competing against 6–12 other agencies with the same list. The T65 signal tells you *when* someone becomes eligible — it doesn't tell you *how likely* they are to choose your product or respond to outreach.

T65 is necessary but not sufficient. It needs to combine with at least one enrichment signal to produce a high-confidence lead.

The 4 signals that don't move

These appear in every lead scoring deck. They have almost no predictive value in practice:

1. Age alone (beyond T65). Being 70 vs. 74 doesn't meaningfully predict plan change likelihood. 2. County population density. High-density counties have more leads, not better leads. Density affects agent territory sizing, not lead quality. 3. Time of year outside AEP/SEP windows. Leads acquired outside enrollment windows are real people, but outreach timing is constrained — you're scoring for later, not for now. 4. Generic "engagement" based on email opens. Open rates from cold Medicare lists are near random and have no verified correlation with conversion.

The signals that do

These 12 signals, in rough order of predictive strength:

1. T65 proximity (high if within 90 days, very high if within 45) 2. LIS/Extra Help eligibility — dual-eligible beneficiaries have dramatically higher plan change rates 3. Prior plan change history — a beneficiary who changed plans in the last 3 years is 2.8× more likely to change again 4. Medicaid crossover signal — dual eligible status changes frequently during the year, not just during AEP 5. County-level carrier mix shift — when the dominant carrier in a county loses market share YoY, their existing members become mobile 6. CMS enrollment data pattern — gaps in continuous enrollment signal manual re-enrollment events 7. Income bracket (verified, not inferred) — LIS threshold proximity is more predictive than generic income range 8. Outreach recency and channel — phone engagement in the last 30 days, not just opens 9. ZIP code SNP concentration — high Dual SNP density areas produce different lead behavior than MAPD-dominant areas 10. Agent contact history — first contact vs. follow-up contact have very different conversion profiles 11. Plan benefit change notification — beneficiaries who received an Annual Notice of Change with material benefit reductions in their plan 12. Prescription drug tier sensitivity — beneficiaries with high-tier medication needs are very sensitive to PDP and MAPD formulary changes

The 12-signal model in beneficiaryCONNECT weights these by local market — the signal mix in Dallas County differs from rural West Texas.

What this means for agent time allocation

The average Medicare agency has agents spending 40–50% of their day on lead qualification calls. Most of that time is spent on leads that will never convert — not because the agent failed, but because the lead was never scored correctly.

A 12-signal scoring model, applied before the first outreach call, changes the economics. Agents work the high-confidence leads first. BRIDGETTE AI handles initial engagement on medium-confidence leads before a human call. Low-confidence leads stay in a warming queue.

The result: fewer calls per conversion, not fewer calls total. Agent time concentrates where it converts.