Before you argue about whether it works, check whether the number is possible.
Most vendor savings claims can be tested in about ninety seconds, with four numbers and one division, before anyone opens a clinical study. It settles more meetings than the study does.
Four numbers. Two of them you already have.
Covered lives and annual plan spend are on your renewal packet. The claimed saving is on the vendor’s slide. Engagement is the one they are least specific about, and the one that moves the answer most.
Employees plus dependents on the plan.
What the plan pays out in a year, before this programme.
The headline figure, per member per year. A PMPM number × 12.
Engagement. Vendors publish this reluctantly; some disclose it in SEC filings.
Their figure multiplied by every covered life — which is how a buyer reads it, and how it lands in a board paper.
The same claim on the denominator it was measured against. Often a real number, and worth buying at the right price.
No single programme removes a quarter of everything a plan pays in a year. This is almost never a false claim — it is a different denominator. Ask what the per-member figure is per member of.
Every figure above comes from what you typed. This page holds no benchmark data and sends nothing anywhere — it is four numbers and a division.
The same claim, read two ways.
A large digital physical therapy company published a saving of $2,941 per member per year — a 3.14× return. We have not named it, because the point is the shape of the arithmetic rather than the company. You will recognise the shape.
Of everything the plan pays in a year, from back pain alone. All musculoskeletal care is roughly 13% of plan spend, so the claim as a buyer reads it would erase more than twice the whole category.
The company’s own filing with securities regulators discloses that 3.4% of covered people were active users. On that denominator the same claim is a real number, and a fine thing to buy at the right price.
Same claim, same maths, same company. “Per member” means per member who used it. It is not a lie. It is a different denominator, and it is the single most common way a savings figure misleads without anyone intending it to.
This test takes ninety seconds and it settles a lot of meetings before anyone opens a clinical study.
A test worth running is a test with edges.
A claim can fail this test and the programme can still be worth buying. What fails is the headline, and what that changes is the price you should pay rather than whether to buy at all.
Selection bias, population fit, engagement realism and whether there was a comparison group at all are separate questions. This test just decides whether they are worth asking.
Two of the four inputs are on your renewal packet and one is on the vendor's own slide. That is the point — you can run it in the meeting.
This is one programme. Most employers own six.
The same test across a whole portfolio is the free report — every programme on one scale, the overlapping claims separated, and every assumption written down where you can argue with it. Reviewed by a person, in your inbox within 24 hours, with no call attached.