Methodology Companion — a reading guide for the reviewing actuary.
A short companion to the methodology document, written for the actuary preparing to read it ahead of a renewal or strategy meeting. Anticipated reviewer questions, the seven most-frequently-raised technical objections to v1.0 and their responses with explicit section references, a pre-meeting checklist, a post-meeting follow-up template. The companion adds no methodology of its own; it is a reading aid.
Why this exists
The methodology document is, by design, written at the audit bar for a credentialed actuary. The companion exists to make the act of reading that document — and of carrying its implications into a renewal or underwriting meeting — efficient. It does not duplicate the methodology, nor does it offer alternative formulations of any quantity specified in the methodology. It is a reading guide that anticipates the most-frequently-raised reviewer questions, points to where each is answered in the methodology, and provides operational scaffolding (pre-meeting checklist, post-meeting follow-up template) for an actuary who is folding the buyer-side instrument into their own work.
How to use this companion
Read the methodology first, in full, end-to-end. Approximately 30–45 minutes. Then return to this companion. Use the seven anticipated-objection blocks to confirm or sharpen your own initial reactions; use the pre-meeting checklist to organize what you carry into the renewal or strategy conversation; use the post-meeting follow-up template to capture the technical points raised across the table for the v1.1 drafting input.
The seven most-frequently-raised technical objections to v1.0
“A unified Bühlmann constant k = 400 PQ across all four markers is too clean. Real markers have different variance ratios.”
Acknowledged. The variance-component derivation in §4.3 is rigorous for weight. The analogous derivation across the other three markers yields k values in the 350–480 PQ range. v1.0 adopts the unified figure as a conservative anchor with two mitigations: the per-marker bootstrap CI (§5) absorbs additional variance naturally, and the 12-month lag haircut λ = 0.50 (§6) provides further cushion at the long horizon. The unified figure is also explicitly named in Limitations L1. v1.1 will publish per-marker k values derived from production data.
→ Methodology v1.0, §4.3 (derivation), §5 (CI absorption), §11 / L1 (named limitation)
“A zero prior on biometric drift is too optimistic — NHANES shows positive deterioration in working-age adults; you should be crediting the cohort for avoiding it.”
Inverted, intentionally. A zero prior is the conservative choice from the Forecast's perspective: the cohort does not receive credit for avoiding the slow background NHANES deterioration. If the prior were set to NHANES drift, the credibility-weighted estimate would be larger in absolute dollar terms (the cohort would be credited with both its observed improvement and the avoided baseline deterioration). v1.0 declines that credit. Limitations L2 names the choice and notes v1.1 may move to a population-baseline prior with documented effect on the point estimate.
→ Methodology v1.0, §3 (prior choice), §4.1 (formula), §11 / L2 (named limitation)
“You treat each cost coefficient as a point value at the lower bound, but the underlying literature publishes a range. The CI you report doesn't include coefficient uncertainty.”
Correct, and named. v1.0 propagates only sampling uncertainty on the cohort delta through the bootstrap. The reported CI does not include coefficient uncertainty. The choice is conservative for the point estimate (lower-bound anchor on every coefficient — see §9) but means the reported CI is narrower than it would be under full propagation. v1.1 will propagate each cm as a uniform distribution over its published range. The full source-literature range is disclosed in the §9 coefficient table.
→ Methodology v1.0, §5.2 (sources of uncertainty), §9 (coefficient table with ranges), §11 / L3 (named limitation)
“A linear trajectory model for 12-month projection is naïve. Real cohort biometric trajectories saturate, plateau, or regress.”
Acknowledged, and conservative at the long horizon. The linear model generally projects continued improvement rather than asymptote, which means at 12 months it is more likely to over-project the trajectory than under-project it. v1.0 absorbs this through the lag haircut λ = 0.50 and the widening trajectory PI (§6.2). Limitations L4 names the choice and notes v1.1 may move to a piecewise or saturating-curve model. The bootstrap procedure (§5, §6.2) re-fits the trajectory on each resampled cohort, so the prediction interval reflects fit-uncertainty as well as cohort-sampling uncertainty.
→ Methodology v1.0, §6.2 (PI formula), §6.3 (λ haircut), §11 / L4 (named limitation)
“A unified lag haircut λ = 0.50 — by what literature is this defensible per marker?”
Anchored to three sources, with median. §6.3.1 cites the DPP cost-effectiveness modeling (per-mg/dL fasting-glucose realization fraction ≈ 0.45), Million Hearts modeling (per-mmHg systolic-BP realization fraction ≈ 0.50), and SOA general guidance on novel risk-mitigation instruments (30–50% first-year discount). The median across these is ≈ 0.50. Per-marker realization fractions in the source literature span 0.40–0.55. v1.0 adopts the unified median as the v1.0 haircut. Limitations L5 names the choice and notes v1.1 will publish per-marker λm from claim-experience linkage.
→ Methodology v1.0, §6.3.1 (λ defense), §8 (pathway literature), §11 / L5 (named limitation)
“Cross-marker correlation is real. How are you handling it in the total-exposure CI?”
Empirically, via shared bootstrap indices. §5.3 specifies that each bootstrap iteration resamples the same employee indices across all four markers. Whatever cross-marker correlation exists in the underlying cohort (high-improver employees tending to show coherent improvement across markers) is preserved without requiring a correlation matrix to be assumed or estimated separately. The total-exposure CI is read off the bootstrap distribution of Etotal(b) directly. A reviewing actuary who wants to inspect the implied correlation structure can request the per-iteration bootstrap output. Limitations L8 names this choice.
→ Methodology v1.0, §5.3 (procedure step 1), §11 / L8 (named limitation)
“This is not a substitute for carrier-experience credibility on incurred claims.”
Not the claim made. §1 explicitly states the Forecast is a complementary, buyer-side instrument intended to be carried into the renewal conversation alongside — never instead of — the carrier-side experience-credibility instruments. Limitations L7 names the absence of carrier-claim integration as a deliberate v1.0 scope choice. The cleanest framing for the broker or actuary to carry into the meeting: this is the buyer's forward biometric-signal forecast, to be presented next to (not as a replacement for) the carrier's retrospective claim experience.
→ Methodology v1.0, §1 (scope and intent), §11 / L7 (named limitation), §12 (attestation protocol)
Pre-meeting checklist — for the actuary or broker carrying a Forecast into a renewal conversation
Before the meeting
- Read the methodology v1.0 document in full. 30–45 minutes.
- Read the §11 Limitations list (nine items). Confirm comfort with each.
- Identify the cohort size N and credibility weight Z on the Forecast cover. Confirm Z is the figure you expect from n / (n + 400) with n in person-quarters.
- Identify which tier the Forecast carries on its signature block (Tier 1: employer-retained signer; Tier 2: broker-firm signer; Tier 3: methodology-version attestation only).
- Confirm the methodology version pinned in the Forecast footer matches the version of this document you have read.
- Have the §9 coefficient table available. The reviewing actuary across the table may want to substitute midpoint coefficients to test sensitivity; the methodology supports this and the substitution will yield a higher figure than the published Forecast.
- Have the Actuarial Brief multi-marker appendix at hand (/actuarial-brief.pdf), which carries the same coefficient table in doctrine register.
- If the cohort is below the carrier-side credibility threshold (typically <200–500 employees depending on carrier), be prepared to position the Forecast as a forward biometric signal complementing — not replacing — the carrier's retrospective claim credibility.
In the meeting
- Lead with the §1 framing: this is a complementary, buyer-side instrument. It is not a substitute for the carrier's experience-credibility math.
- Walk through the cover page first — point estimate, 95% PI, cohort size, credibility weight, methodology version, signature block.
- If asked about k, point to §4.3 derivation. If asked about λ, point to §6.3.1 anchors. If asked about CI procedure, point to §5.3.
- If pushed on any limitation, name it openly — every one of the nine §11 items is published in the methodology. Naming a limitation that the methodology already names openly disarms the objection.
- If asked “what is the carrier supposed to do with this?”, the answer is: fold it into your credibility math as a complementary forward signal alongside your existing claim-experience credibility, in whatever weight your professional judgment supports.
- If the reviewing actuary expresses interest in attesting to the methodology, the protocol is in §12: care@usemetra.com, subject “Methodology Attestation — v1.0”.
- If the reviewing actuary raises a substantive critique, log it for the post-meeting follow-up. The v1.1 drafting input is built from these critiques.
Post-meeting follow-up template
Within 48 hours of the meeting, send a short follow-up to the reviewing actuary. The template below is the recommended structure. A second copy goes to care@usemetra.com with subject line “Methodology Review — v1.0 — [meeting context]” so the critique enters the v1.1 input log.
Subject: Methodology v1.0 — follow-up from [date] meeting
[Reviewing actuary name],
Thank you for the time on [date]. A short follow-up so the technical
points you raised are captured cleanly:
1. [The first specific technical point the reviewer raised]
— Reference: Methodology v1.0, §[section]
— Status: [acknowledged and named in L#] / [open for v1.1] / [resolved
in conversation]
2. [Second technical point]
— Reference: …
— Status: …
3. [Third technical point]
— Reference: …
— Status: …
Methodology v1.0 is at usemetra.com/methodology/v1.0/. The attestation
roster is at usemetra.com/methodology/attestations/. If the points above
resolve cleanly on a second read, an attestation submission would be
warmly welcomed at care@usemetra.com.
If any point would benefit from further discussion, I'm available.
[Sender name]
[Firm]
What this companion is not
This companion contains no formula, no coefficient, no constant that is not already specified in the methodology document. Every quantity referenced here is sourced from the methodology with an explicit section reference. The companion is a navigation aid, not a supplement. If a reviewer asks for the methodology “in plain English”, the companion is what to send; if a reviewer asks for the methodology, the methodology v1.0 document is what to send.
Sell-side equity-research register, applied to actuarial review. The companion is written in the register of a sell-side equity research note prepared for a buyer-side analyst about to read a primary document. It assumes the reader is professionally credentialed, technically literate, and time-constrained. It does not summarize what the reader is about to read; it identifies the places where the primary document's argument is most likely to be tested and points the reader at the exact location where each test is answered.
Companion to v1.0
Methodology v1.0 is the primary document. This is the reading aid.
The full mathematical methodology is at usemetra.com/methodology/v1.0/. The public attestation roster is at usemetra.com/methodology/attestations/. The Actuarial Brief with multi-marker coefficient appendix is at /actuarial-brief.pdf. All three documents are part of the public corpus a reviewing actuary works through.