The Levers Move Less Than the Weather. That Is the Whole Finding.
Ask a managed-care leadership team what they can do to the margin and you will get a list of programs. Ask the published evidence how much each program moves, hold every lever to its band, and draw the annual rate swing beside it — and something uncomfortable appears. Six things from building a P&L simulator that refuses to flatter anyone.
1. Every lever gets a band, never a point. Reference pricing cut outpatient facility prices 25% for a state purchaser; a tiered network cut spending about 5% for another; pharmacy prior authorization moves targeted classes. Real effects, independently measured — and every one arrives as a range, applied to the slice of spend it actually touches, scaled by the reach you can honestly claim. Output is a range because the inputs are ranges. A simulator that returns one number has already lied once.
2. The most confidently sold programs are locked at zero. Wellness: two large randomized trials, two nulls on spending. Telephonic disease management: measured at national scale and null — Medicare Health Support, and thirteen of fifteen coordinated-care demonstrations. Retention outreach and CAHPS uplift: never independently measured at all. The simulator shows these levers and locks them, with the reason printed. A tool that lets you type an effect for an unmeasured program is a pitch deck with sliders.
3. The weather is drawn beside the work. The rate announcement moved 3.1 points in one cycle and 5.2 in another, direction unknowable in advance. Morbidity trend misses run one direction. Risk-model residue is still settling. None of these are manageable, all of them move the margin — so they enter as dials, not levers, drawn on the same chart.
4. The thin span is the finding. Stack every evidence-backed lever at honest reach and compare the total to one year of weather: the span management actually controls is thinner than the span that controls management. That is not fatalism — it is prioritization. A thin span means every Tier-D program funded as a lever eats span the real levers needed.
5. Earnings calls are mostly weather, worn as strategy. Read a season of managed-care earnings calls against this taxonomy and most announced "actions" classify as portfolio moves and weather positioning — exiting counties, repricing bids — not levers. By our count, at most two of ten recurring strategies are things the evidence lets you call a lever. That is not a criticism of the companies; it is a decoder ring for the transcript.
6. Additivity is a declared simplification. Stack six programs and the simulator says out loud that adding their bands assumes they do not fight over the same dollars — which they do. The honest flag is printed, not buried. Every simplification in the tool announces itself, because a model whose assumptions are silent is a model whose errors are yours.
We built this as an instrument: levers with cited bands, unmeasured programs locked, weather dials, the thin span drawn where a board can see it. Browser-only; nothing entered leaves the machine.
A P&L simulator that accepts any coefficient you type is not a model — it is a mirror.
If you stacked only the levers the independent evidence supports, at the reach you could defend to your own actuaries — how wide is your span, and what is currently funded as if it were inside it?