How a Self-Funded Renewal Projection Is Built: Experience, Pooling, Credibility, Trend, and Fixed Costs

Every fall, self-funded employers receive a renewal projection: from the carrier, from the stop-loss underwriter, sometimes from both at once. The numbers rarely agree. The projection itself is not mysterious. It is a chain of six checkable steps, and anyone responsible for a health plan budget can follow every link. That matters more than usual right now. Surveys published in the summer of 2026 put 2027 medical trend between 9 and 11 percent before plan changes, with PwC at 9.0 percent for the group market and Segal at 9.9 percent for PPO plans, the highest readings in nearly twenty years. This post walks through the standard actuarial build, in order, with a worked example carried from the first line to the last.

The Experience Period vs. the Projection Period

A renewal projection answers one question: what will this group's claims cost during the next plan year? The raw material is the experience period, usually the most recent 12 months of the group's own claims, with a second 12-month period added and weighted less heavily when the data supports it. The projection period is the plan year being priced.

Two adjustments make the experience usable. First, claims are organized by when care was incurred, not when the check was paid, and recent months are always incomplete. That is why the experience window in carrier reports typically ends a few months before the analysis date. Second, raw dollars are divided by enrollment to get a per-employee per-month (PEPM) figure. A group that grew 10 percent will show 10 percent more claims dollars with no change in underlying cost; PEPM strips that out and makes periods comparable.

Complete the Experience

Recent incurred months are partially paid, so complete them before anything else. A plan with a run-in year of paid history can use its own lag pattern; where it cannot, an industry completion curve is the standard fallback. A 12-month period ending three months before the analysis date is typically 97 to 99 percent complete in aggregate but only 60 to 80 percent complete in its final two months. Pooling and trending an uncompleted period understates the base.

Pool the Large Claims First

One transplant or one cell-therapy course can add hundreds of dollars PEPM to a mid-size group's experience. Left in the experience, a single outlier year sets the budget. Removed without replacement, the budget ignores a cost that will eventually arrive. For episodic events like these, the odds of a repeat next year are largely independent of what happened this year. Known ongoing conditions, such as hemophilia or dialysis, are a different story. Stop-loss underwriters price those claimant by claimant, which is where lasers come from.

The exposure is real and growing. Stop-loss carrier Sun Life reports that $1 million-plus stop-loss claims rose 46 percent in frequency from 2022 to 2026. In Segal's SHAPE claims database covering 2019 through 2023, claimants above $100,000 were fewer than 1 percent of claimants and drove roughly 30 percent of total plan spend.

The standard answer is pooling: choose a pooling point, remove each claimant's dollars above it from the experience, and add back the expected cost of that excess layer later as a predictable charge. For a self-funded plan the pooling point is usually set at the specific stop-loss deductible, which means the add-back arrives naturally as the stop-loss premium rather than as a separate pooling charge.

Credibility: How Much Does the Group's Own History Mean?

A 5,000-life group's claims history is a statistically reliable predictor of its own future. A 100-life group's history earns little credibility under standard size-based formulas. Credibility weighting handles the middle ground with the standard blend found in every group insurance textbook:

The manual rate is an independently built estimate of expected claims for a group with this age mix, family size, and plan design (a full underwriting manual adds geography and industry on top), and it must be pooled on the same basis as the experience. Credibility Z is driven by size. Full credibility under limited-fluctuation criteria is usually quoted in the low thousands of life-years for medical, and the exact standard depends on the pooling point: the lower you pool, the less severity variance remains and the fewer life-years full credibility requires. Below the threshold, weight grows roughly with the square root of size. Publish the standard you used, because a credibility weight without its full-credibility basis is not checkable.

Trend: Four Components, Not One Number

The blended figure still describes the past. Trend carries it forward to the projection period. Published 2027 projections cluster between 9 and 11 percent for medical trend, with pharmacy trend running higher. Quote the survey you are using by name, edition and number, and link it, because the published figures differ by several points depending on whose book of business is behind them. Behind any single trend number sit four distinct forces:

Deductible leverage, worked through

Leverage is the component most renewals gloss over, and it is pure arithmetic. Take a plan that pays 100 percent after a $1,000 deductible, and a member with $5,000 of allowed cost. Underlying costs trend 8 percent:

The plan's cost trended 10 percent while the underlying cost trended 8, because the deductible stood still. Run the same member through a richer plan with a $250 deductible and the plan's cost rises only 8.4 percent. The principle: the leaner the plan, the more of its cost sharing is fixed in dollars, and the faster its paid claims trend. Applying one market-average trend rate to a high-deductible plan systematically under-projects it.

Published trend surveys measure plan-paid claims at average plan richness. A plan leaner than average trends above the survey number; a richer plan trends below it.

Mechanically, trend is applied midpoint to midpoint: from the middle of the experience period to the middle of the projection period, compounded, not added. For a January renewal, the span depends on how fresh the experience is: a 12-month period ending the prior September puts 15 months between midpoints, and one ending the prior June puts 18.

Stop-Loss and Admin: The Fixed-Cost Stack

Expected claims are the variable half of the budget. The fixed half is added on top: the specific and aggregate stop-loss premiums, third-party administration, network access, and any consulting or vendor fees. Because the example pooled at the $100,000 specific deductible, the specific premium is exactly where the cost of the removed large-claim layer re-enters the build. This is also where a bad claims year leaves its mark: the claims above the pooling point never touched the expected-claims line, but the stop-loss carrier saw them, and they return as premium pressure or a laser at renewal.

Reading the Final Number

Assembled in order, the whole projection fits on one card:

That is roughly $18,200 per employee per year, in line with where the national average is heading: Aon put total plan cost per employee at $17,562 for 2026 and projects a 9.5 percent rise for 2027, which takes the average above $19,000. If the current budget rate is $1,395 PEPM, the projection lands at an 8.8 percent increase, and every dollar of it traces to a named assumption.

This is the real payoff of building the projection rather than receiving it. When a carrier report arrives at a 12 percent increase and an independent build says 8.8, the disagreement lives on a specific line: a higher trend assumption, a lower credibility weight, a different pooling charge. Those are questions with answers, and asking them line by line is the core of a defensible renewal strategy.