Round 1 is fully answered — 24 comments from two reviewers. AJ Robinson's 18 drove the
v3.0 code corrections; Mark Mondello's 6, plus his discrete Excel example, drove v3.1 and v3.2
(exposition only, no figure moved). Mark's spreadsheet reproduces in our code to the cent and
independently reproduced the cross-sell defect, which is the best evidence we have that the
old wording invited it — see the
response memo.
The work is finished; the routing is not. The paper (v3.2, 49 pp), the executive summary
(3 pp, inside the contractual range) and the toolkit (96-file MPL 2.0 release, gate green) are
all assembled and mutually consistent —
289 tracked figures, 0 unmatched against the
drift gate. What is missing is not research. It is
four routing decisions, three of which
only CAS can make, and which have been open long enough that they are now the schedule risk.
★ = researcher's recommendation.
Q1
Submission venue — Variance or E-Forum?CAS
✓ You committed (kickoff Q6) that CAS
provisions ScholarOne access and the researcher submits directly.
✓ The manuscript is ready — but the venue determines the
Pandoc reference template (paper/pandoc/reference.docx, still missing) and the
abstract's target length (currently 281 words).
This is now the largest schedule risk on the project. Every other
paper task is complete. This one answer unblocks rendering the final submission artifact.
Which venue should the paper be submitted to?
A
Variance — the
contract calls for a peer-reviewed paper, and Variance is the peer-reviewed
venue★ rec
B
E-Forum — faster to
publication, but not peer-reviewed; would need the deliverable language revisited
C
Both — E-Forum now for
availability, Variance submission in parallel
D
POG to consult CAS publications staff
and confirm by email within the week
Q2
RPM 2027 — deliverable or individual proposal?CAS
✓ You chose the RPM Seminar as the venue
(kickoff Q7), with the date to be confirmed.
✓ We built the whole submission: a 45-character
abbreviation-free title, a 225-word description, three Bloom's objectives inside the
150-character limit, and a 22-slide deck generated from the paper's own seeded numbers
so it cannot drift. scripts/check_rpm_proposal.py enforces every webform and
accessibility limit — it caught a real violation on first build.
How is the session submitted, and what is the actual deadline?
A
As a project deliverable
under the research agreement — the agreement contemplates an RPM presentation★ rec
B
As an individual speaker
proposal, with the project acknowledged in the abstract
C
Hold until the paper is accepted, then
submit citing the accepted paper
D
POG co-presents — add a POG member as
second speaker on the form
Q3
Toolkit release — provisioning and reviewCAS
✓ You committed (kickoff Q6) to provisioning
the github.com/casact repository.
✓ The release is built and independently gated:
dist_release/, 96 curated files under MPL 2.0, assembled by an explicit
allowlist. 122 tests pass inside the release subset itself — not just in the working
repo — with ruff, mypy and the drift gate clean, plus a curation audit that finds no internal
file and a secret scan that finds nothing.
Overdue since ~Jun 19, and there is no local work-around. The repo
gates the toolkit release; ScholarOne gates the submission.
How should the release proceed?
A
Provision now; POG reviews
dist_release/ in parallel with paper round 2, so review does not gate
publication★ rec
B
POG reviews the release subset
first, publish after sign-off
C
Publish only on paper acceptance, so
code and paper land together
D
Stage privately under
casact first, make public at acceptance
Q4
Variant naming — ratify the amendmentA. ROBINSON
✓ You settled this on Jul 2 as
clv_unconstrained / clv_constrained.
✓ AJ's review overturned it for a good reason: actuaries
read "constrained" as the regulatorily constrained variant — the opposite of the sense
intended. Renamed clv_base / clv_with_growth, with the old names
shipping as deprecated aliases for one release so nothing already running breaks.
Confirm the naming that goes into the published toolkit.
A
Ratify
clv_base / clv_with_growth, aliases for one release★ rec
B
Ratify the names but keep the
aliases permanently
C
Different pair — e.g.
clv_renewal / clv_total
D
Revert to the Jul 2 names; treat the
review comment as a documentation fix
Still open from Aug 6 — restated, not re-asked
Four decisions from the Call #4 poll are unanswered, and four passages in the
paper still read "provisional" as a result. Recommendations are unchanged. If time runs
short, answering these by email is enough — but please answer Aug-6 Q3.
| From Aug 6 | The decision | Recommendation, unchanged |
|---|
| Q1 |
Tenure-varying loss — what becomes the published default basis? |
Keep flat as the headline; publish the tenure gradient as a sensitivity
dimension. Reinforced since Aug 6: external validation came back negative, so the
curve now ships as a mechanism, not a parameter. |
| Q2 |
Where should the tenure curve come from? |
Fit on the user's own book; ours is a documented illustration only.
Reinforced — neither real panel reproduced our 4.22%/yr magnitude. |
| Q3 |
May a tenure relativity enter a filed exhibit? |
Planning use only; publish the filing-basis inversion as a cautionary finding.
Answer this one if you answer nothing else — §7 currently takes a position a
regulator could contest, without POG ratification. |
| Q4 |
Case studies — which grain goes in the paper? |
Lead with the 7-policyholder cohort, then scale to tenure bands and the
full book. |
Reserve question — for Call #7 (Thu Aug 27), flagged now so you can think about it.
The Aug 27 call takes the remaining review-round-1 dispositions: the
default cross-sell
hazard (
Economics.cross_sell_prob — now that cross-sell is correctly a one-time
event rather than an annuity, its default probability actually matters),
how the tenure curve
ships after the negative validation result, and
which segment the retention-adjusted
loss ratio should use as its published default — the multi-product cut is the one that produced
the paper's strongest new result, but geography is the one regulators will ask about.
Materials:
Call #7 brief ·
Call #7 poll.