← Stage-Gate NPD Suite Product & Regulatory · The Document That Was Wrong

Actuarial Pricing & Assumption Summary

Download Word
Program timeline · status 16 Oct 2026Read the full story →
Harborline
Aug 2025
Cancelled
Gate 0
Feb 2026
Go
Stage 1
Business case
Gate 1
Apr 2026
Recycled
Gate 1
Jun 2026
Go w/ conditions
Stage 2
Development
You are here
Gate 2
Apr 2027
Gate 3
Oct 2027
Gate 4
Feb 2028
Launch
Mar 2028
Gate 5
Sep 2028

Lighthouse Financial Services Company — The pricing basis for Beacon Index Advantage on the corrected capital footing, with every assumption stated alongside how it could be wrong and what would reveal it. Version 2.0, 11 June 2026. Status as at 16 October 2026.

87 bp
Understatement this document carried
14.6% → 13.4%
Return, tabled → corrected
8
Assumptions, each with a failure mode
2
Not testable before the gate that depends on them
Contents
  1. Why This Version Is Written Differently
  2. Pricing Basis
  3. The Capital Charge, Corrected
  4. Assumptions, With Failure Modes
  5. Sensitivity
  6. What Is Still Not Covered
  7. External Review and Control

1. Why This Version Is Written Differently

This is the document that was wrong. The Gate 1 recycle turned on an 87 basis point understatement of the required capital charge, and that misstatement lived here — in the version tabled to the Board on 30 April 2026. It was not a rounding error or a presentational slip. It was a confident, internally consistent, peer-unreviewed number that moved the program's reported return by 120 basis points, and nobody caught it until a gate refused to score the package.

The correction itself was straightforward once found. What was not straightforward was the question the Board asked afterwards: how would we have known? Nothing in the previous version of this document offered an answer, because it stated what the actuarial team believed and stopped there.

So the organizing principle of this version: an assumption summary's job is not to state assumptions — it is to state how each one could be wrong and what would reveal it. Stating an assumption tells a Board what you believe. Stating its failure mode and its detection point tells a Board what to watch, which is the only part of it they can act on.

2. Pricing Basis

ElementValueBasis
Illustrated cap, capped index strategy9.25%Against a peer illustrated 9.00%. Competitive floor 8.50% fixed at Gate 1 — the cap may not be set below it without returning to the Board.
GLWB rider election62%Share of policies electing the optional rider at issue. Drives fee income and guarantee exposure together, in the same direction.
Average case size$118,000A distribution and pricing assumption, not a contractual minimum. Never carried on any register — see §6.
Required capital as a share of account value4.2%The corrected basis. Tabled at 3.33% on 30 Apr 2026; corrected at the second convening. Held against account value throughout the life of the block, not taken once at issue.
Net product margin0.9332%Solved from the approved return rather than asserted. The model refuses to load if it solves outside a defensible 0.6–1.6% band.
Fixed annual run cost1,237,276The only element that does not scale with volume — and therefore the entire reason the downside scenario returns a different number from the base.

Net margin and fixed run cost are solved from the approved return rather than asserted: the model calibrates them so it reproduces 13.4% base and 10.1% downside exactly, and aborts if either lands outside a defensible range. A pricing document whose figures cannot be reproduced from its own outputs is the failure mode §1 describes.

3. The Capital Charge, Corrected

BasisCapital ratioReturnHeadroom vs 11.0%
As tabled, 30 Apr 20263.33%14.6%360 bp
Corrected, 11 Jun 20264.2%13.4%240 bp
The error's mechanism is worth stating precisely, because it is what made it invisible. Required capital is held against account value for the life of the block, not taken once at issue. An error in that ratio therefore compounds across every projection year rather than showing up as a single wrong line in year one — which is exactly why it survived internal review. The number looked reasonable in the year anyone would check.

The corrected ratio ties to the fact base by identity: 4.2% of the $185,000,000 Year 1 premium is $7,770,000, the Year 1 capital strain figure the business case carries. That identity is what anchors the model to the program's locked facts rather than to a separate set of actuarial workings.

4. Assumptions, With Failure Modes

IDAssumptionHow it could be wrongWhat would reveal itReadable
A-01Licensed and in good standing across all 46 target states through launchA state license lapses or a market-conduct action suspends new business in a target state.Quarterly license attestation from Compliance.Continuously — already readable
A-02The Compact accepts the product under existing uniform standards; no new standard neededThe Compact requires a new uniform standard for the rider feature rather than accepting it under existing ones.The first substantive Compact objection letter.Stage 3 — not before the gate that depends on it
A-03New York is out of launch scopeDistribution pressure to add New York arrives after advisor training is built.A channel escalation, most likely from a national partner.Stage 4 or post-launch
A-04The existing admin platform is extended, not replacedConfiguration reveals the platform cannot carry a launch-scope feature without custom build.Cordelane configuration testing under WP-1.Stage 2 — readable now
A-05Distribution runs through existing IMO/broker-dealer relationships; no new channelExisting partners decline to add the product, requiring a new channel the case did not fund.Written commitments against the Year 1 plan — already partially failed.Already reading against it
A-06Hedging is executed by the in-house Investments deskThe in-house desk cannot achieve readiness, forcing an external hedging provider and a cost the pricing does not carry.ISDA execution progress under GC-05 (DEP-06).Stage 2–3
A-07Rates stay inside the Gate 1 pricing corridorRates move outside the Gate 1 corridor, compressing the option budget and the illustrated cap.The monthly option-budget monitor against the corridor.Continuously
A-08No competing internal launch draws the same actuarial and IT people in 2027A competing internal launch draws the same actuarial and IT people in 2027.The 2027 portfolio planning round.Stage 3 — after the gate that commits to it
Two assumptions cannot be tested before the gate that depends on them — A-02 and A-08. A-02 is binary and, if it fails, invalidates the filing route rather than delaying it. A-08 is tested by the 2027 portfolio planning round, which happens after Gate 2 has committed the program to a resourcing profile the program has no authority to hold. Both are recorded as untested rather than assumed sound, and the honest position is that the program will commit money to both before it can check either.

A-05 deserves separate mention: it is the only assumption already reading against itself. Written channel commitments stand at $168,000,000 against a $185,000,000 Year 1 target — a 9% shortfall disclosed at Gate 1 and carried rather than closed.

5. Sensitivity

Volume is the single variable that moves the return, because fixed run cost is the only element that does not scale with it. Margin and required capital both scale, so a proportional shortfall would leave the return untouched — it is operating leverage alone that converts lower volume into a lower return.

Volume vs planProjected IRRNPV at hurdleVerdict
+0%13.40%$11,009,118clears
-5%12.94%$8,618,608clears
-10%12.45%$6,228,098clears
-15%11.92%$3,837,588clears
-20%11.36%$1,447,077clears
-25%10.76%$-943,433FAILS
-30%10.10%$-3,333,943FAILS

The case clears down to a 23% shortfall. The Benefits Realization Plan converts this table into pre-committed monitoring thresholds, and the Cost-Benefit Analysis carries the full scenario record.

6. What Is Still Not Covered

IDGap
AG-01Average case size is not on any register. It is a pricing assumption of $118,000 with no risk, assumption or condition attached to it. Nothing in the program's governance would surface a movement in it before launch, and it decomposes volume alongside policy count — two different problems with two different remedies.
AG-02Rider election and persistency are modeled independently. Higher election with worse persistency can look neutral in aggregate while changing the shape of the guarantee exposure materially. The annual experience study handles the interaction; nothing before launch does.
AG-03The model is calibrated, not validated. It reproduces the program's approved returns by construction — that proves internal consistency, not that the underlying pricing is right. GC-03's external review addresses the pricing basis; no review addresses the model.
AG-04No assumption carries an owner distinct from the team that set it. Actuarial sets the pricing assumptions and Actuarial monitors them. That is normal in a carrier this size and it is precisely the arrangement that let an 87 bp error reach a gate.

7. External Review and Control

ControlStatus
GC-03 — independent external actuarial review of the rider pricing basis, funded by a $180,000 contingency drawOn track — Ardmore engaged; due 26 Feb 2027
Gate 2 must-meet: pricing signed and supported by the external reviewOn track
Model reproduces all three locked returns to within half a basis point, or refuses to loadEnforced in code
Independent review of the model itselfNone — see AG-03
GC-03 exists because of the error described in §1. The Board's response to a peer-unreviewed number reaching a gate was to require that the most consequential pricing element be reviewed by somebody outside the carrier, and to pay for it from contingency rather than absorb it into the actuarial workplan — which makes the control visible in the budget rather than invisible in a team's capacity.

Prepared by S. Ravichandran, Lead Pricing Actuary, under N. Adeyemi, Chief Actuary. Tabled through the gate process by C. Tyrrell, NPD Program Manager. Related: Cost-Benefit Analysis · Gate 1 Recycle Memorandum · Gate 1 Recycle — Executive Session · Gate Conditions Register (GC-03) · RAIDD Log.