How the numbers are made

Every figure PlanVantage prints comes from compiled code and versioned tables, built and signed off by an FSA. This page shows the method behind each one, what it is calibrated to, and how it compares with published sources. The formulas are the standard ones; the work is in the tables they run on.

01

Actuarial value and the continuance tables

An actuarial value is the share of allowed cost a plan design pays. The engine reads it off a continuance table: the distribution of members by annual allowed cost, with each band’s average spend by service. For every band it applies the design’s deductible, coinsurance, copays and out-of-pocket maximum, sums what the plan pays, and divides by what was allowed. Each band carries medical and prescription spend by service, so a separate Rx deductible or out-of-pocket maximum is applied to the prescription dollars in every band rather than approximated.

The continuance distribution, with a sample design marked

The continuance table, figures illustrative

The first rows of the continuance table: for each band of annual allowed cost, the share of members in it, their allowed cost, and the average spend by service. The table continues beyond what is shown.
Band, up to	Members	Allowed	Inpatient	Rx generic
No claims	15.50%	$0	$0.00	$0.00
$100	2.98%	$50	$0.00	$19.65
$200	3.27%	$159	$0.00	$28.01
$300	3.61%	$260	$0.01	$35.60
$400	3.31%	$363	$0.01	$49.94
$500	3.12%	$466	$0.00	$62.92
$600	2.92%	$571	$0.02	$78.42
$700	2.69%	$680	$0.02	$92.53

Actuarial value on a continuance table

AV m (a) = \frac{\sum _{i} p _{i} ( a _{i} - m ( a _{i} ) )}{\sum _{i} p _{i} a _{i}} = min (min (a, D) + c max (0, a - D), M)

: the share of allowed cost the plan design pays
: share of members in band i
: allowed cost in band i
: what the member pays at that allowed cost
: deductible
: member coinsurance
: out-of-pocket maximum

Two checks against published figures. The allowed cost of care per person the engine produces at default assumptions, a single adult with no rebates taken out, reads 9% below the Milliman Medical Index on the same rebate basis, whose average person is one of a family of four. And across 1,416 designs scored both ways, the engine agrees with the current CMS AV Calculator within three points 78% of the time, and where the two differ it reads lower on average. The calculator values one fixed standard population with its own rounding rules; the engine values the same design on its own basis. A PlanVantage value tends to be the conservative one. The leanest plan the ACA allows for 2026, deductible and out-of-pocket maximum both $10,600, reads 60.4% on the engine, above the 60% minimum-value line.

PlanVantage beside the published figure for the same measure

PlanVantage results beside the published figure for the same measure: allowed cost per person against the Milliman Medical Index, and actuarial value against the current CMS AV Calculator across 1,416 designs.
Measure	PlanVantage	Published	Read
Allowed cost per person per year−9.2%	$8,496	$9,357, Milliman Medical Index 2026	−9.2%
Actuarial value, 1,416 designs scored both ways78% within 3 points	r² 0.95	the current CMS AV Calculator	78% within 3 points
Mean difference from the calculatorlower on average	−1.4 pt	Platinum +0.1, Gold −1.2, Silver −4.2	lower on average

The leanest plan the ACA allows, 2026, on the engine, employee only

The leanest plan design the ACA allows for plan year 2026, run employee only.
Plan year	2026
Deductible	$10,600
Out-of-pocket maximum	$10,600
Member pays below the maximum	100%
Enrollment	employee only

Actuarial value on the engine

60.4%

50%60% minimum value70%

02

Creditable coverage

A group prescription plan is creditable when it pays at least as much of expected drug cost as the Part D defined standard benefit, measured as an Rx actuarial value against a threshold CMS sets each year: 72% for 2026, 73% for 2027. The engine values the plan and the standard benefit on the same Medicare-aged distribution and reports creditable or not creditable, and flags a result within 2 points of the line for review. Checked against 28 published determinations clear of that band, it matched 23; the 5 it missed it read as not creditable, the conservative reading.

Part D standard benefit and creditable threshold, 2026 and 2027

The Part D defined standard benefit and the creditable-coverage threshold, one column per plan year.
Plan year	2026	2027
Deductible	$615	$700
Member coinsurance	25%	25%
Out-of-pocket threshold	$2,100	$2,400
Member share above it	0%	0%
Creditable threshold	72%	73%

Six prescription design archetypes on the engine, 2026 parameters

Six prescription designs by Rx actuarial value on a scale from 40 to 100 percent, against the 72% threshold for 2026, with the published determination where one exists.
Design	40% to 100%	Rx AV
Rich copay Rx, four tiers		92.2%
High-option group Rxcreditable, published		88.1%
HSA-qualified HDHP, shared deductiblecreditable, published		79.0%
Flat 28% coinsurance		72.0%
Lean copay designnot creditable, published		66.8%
HDHP, family deductible above the Part D cap		46.1%
72% threshold, 2026 below it

The creditable coverage test

AV_{Rx} creditable = \frac{E [ plan-paid drug cost ]}{E [ allowed drug cost ]} ⟺ AV_{Rx} \geq T_{year}

: the plan’s Rx actuarial value on the Medicare-aged distribution
: the expected value over the Medicare-aged distribution
: the threshold CMS sets for the year

03

Projection

A projection is the standard build-up, and every term is a row on the screen with its source: the engine’s figure, the carrier’s, or yours. Fully insured, the blended claims and the pooling charge are grossed up for retention. Self-funded and level-funded, the same claims leg sits beside the stop-loss premiums, fees and adjustments. Trend is the plan-paid trend the design’s actuarial value implies, so a leaner plan trends harder; at a market-average design the default reproduces the 2026 surveys and sits at the low end of the 2027 surveys published so far.

The fully insured renewal build-up

premium_{PEPM} blend = \frac{blend + pooling charge}{1 - \sum _{i} retention _{i}} = Z \cdot experience + (1 - Z) \cdot manual

: the fully insured premium per employee per month
: the credibility-weighted claims per employee per month
: the group’s own claims, trended and pooled
: the manual rate, net of the pooled layer
: credibility, the weight on experience
: the expected cost of claims above the pooling point
: administration, taxes, commissions and margin, each a share of premium

The default trend beside the 2027 surveys published so far

The default trend beside the 2027 survey figures, medical and prescription.
2027 trend	Medical	Rx
PlanVantage default	9.0%	11.0%
Segal	9.9%	11.5%
PwC	9.0%	–
IFEBP	10.0%	–

Plan-paid trend at each actuarial value breakpoint

Plan-paid trend by actuarial value, medical and prescription, in two column groups: a leaner plan trends harder.
AV	Medical	Rx	AV	Medical	Rx
60%	10.4%	12.8%	85%	9.0%	11.0%
65%	10.1%	12.4%	90%	8.7%	10.7%
70%	9.8%	12.0%	95%	8.4%	10.4%
75%	9.5%	11.7%	100%	8.2%	10.1%
80%	9.2%	11.3%

04

Manual rate

The manual rate starts from a national allowed cost per member, trends it to the plan-year midpoint, and adjusts for the design’s richness, the group’s age and its members per employee. That is the gross rate; the net rate removes the layer the pooling point takes out, so it sits beside experience that was pooled the same way. Credibility weights the group’s own experience against the manual by covered members and months of data.

The manual rate, gross and net

manual_{gross} manual_{net} = allowed_{PMPM} \times trend \times richness \times age \times members per employee = manual_{gross} - pooled layer

: the rate per employee per month, gross and net
: national allowed cost per member per month, from a published index
: the average of medical and Rx trend, compounded from the index year’s midpoint to the plan year’s
: the design’s paid share relative to a full-value plan, with the network discount
: the group’s age factor relative to age 40
: covered members per subscriber, from the group’s own enrollment
: the expected claims above the pooling point, removed for a net rate

One-year group credibility (Fuhrer)

Z = 1 - \frac{0.75}{1 + 0.01 ( n - 1 )}

: covered members; fewer than twelve months of data scales Z down
: the weight on the group’s own experience, 25% at n = 1, rising toward 100%

Credibility by group size, twelve months of experience

Credibility by covered members, twelve months of experience, computed from the rule above.
Covered members	Weight on own experience
50	50%
100	62%
250	79%
500	87%
1,000	93%

05

Large claims

One severity curve answers three questions: how many members will exceed a pooling point, what the layer above it costs, and where the pooling point should sit for a group of a given size. It is a lognormal body spliced to a generalized Pareto tail at $100,000, trended to the plan year before use, and calibrated to published large-claim continuance data. On plans run in PlanVantage in 2026, claimants above the pooling point ran 1.08 times expected and dollars 0.99 times.

Members per thousand above each threshold, published beside the curve

Members per thousand with annual claims above each threshold, in the curve's reference dollars: the published reference beside the curve.
Claims above	Published, per 1,000	Curve, per 1,000
$50K	22.0	20.8
$100K	9.0	9.0
$250K	2.0	2.0
$500K	0.40	0.48
$1M	0.10	0.10

The severity curve, with its published calibration anchors

Expected claimants above a pooling point

λ = n \cdot S (\frac{P}{t}) \cdot u

: the expected number of claimants above the pooling point in the plan year
: covered members
: the pooling point, in plan-year dollars
: the share of members whose annual claims exceed x, read off the severity curve
: the dollar trend from the curve’s reference year to the plan year
: the utilization factor, the growth in claim counts over the same years