What the Numbers Miss is a weekly conversation from GEX Management, Inc., a strategy and operations firm based in Dallas, Texas. Each episode takes one number a business reports and asks what it leaves out.
Episode 1, The Ten Thousand Dollar Gap, is about turnover in direct care. The argument is that a single turnover figure is two different things added together, that they have different price tags, and that only one of them can be fixed.
Ada is an AI. Not a person, and not pretending to be one. She brings the numbers and she argues with the host, and she is identified as an AI in the first thirty seconds of every episode.
Below is the full transcript as broadcast, followed by every figure spoken in the episode and the document it came from.
Where to listen
On this site: gexmanagement.com/podcast
Apple Podcasts: podcasts.apple.com/us/podcast/what-the-numbers-miss/id6811440700
Spotify: open.spotify.com/show/1TB1RXc4BdqaRbtyAl51vq
Every figure in this episode, and where it came from
Every figure spoken in this episode is read by Ada from a primary document. The host speaks no survey figure of his own; the only numbers in his lines are arithmetic performed on hers.
4,677,100 home health and personal care aides. 18 percent projected growth over the decade. Median pay of $17.21 an hour. U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, Home Health and Personal Care Aides.
Median hourly wage of $17.36 in 2024. Median annual earnings just under $26,000. 46 percent of home care workers part time, under 35 hours a week. 48 percent full time, year round. PHI, Direct Care Workers in the United States: Key Facts 2025.
Medicare-certified home health payments estimated to fall in the aggregate by 1.3 percent. Centers for Medicare and Medicaid Services, Fact Sheet, CY 2026 Home Health Prospective Payment System Final Rule (CMS-1828-F).
A further 7 percent base rate reduction recommended for 2027. Medicare Payment Advisory Commission, March 2026 Report to the Congress, Chapter 8, Home Health Care Services.
Caregiver turnover of approximately 75 percent, the lowest level reported in five years. Activated Insights, 2025 Benchmarking Report.
The one calculation performed aloud: $17.36 an hour across 2,080 hours is $36,108.80, which the episode calls about thirty-six thousand dollars. Against median annual earnings just under $26,000, the gap is roughly $10,000. Both figures are PHI's own, from the same report, so the comparison does not mix sources.
Full transcript
Sri: There is a number in every services business I look at, and I think it is misread almost every time. Turnover.
Not misread as in the math is wrong. Misread as in it gets treated as one thing when it is actually two things added together, and those two things have completely different price tags. One of them you can fix in a quarter. The other one you cannot fix at all. And the number does not tell you which is which.
Buyers discount for it. Owners apologize for it. Almost nobody asks the question that separates them.
I'm Sri Vanamali, the CEO of GEX Management, a strategy and operations firm in Dallas. And this is What the Numbers Miss.
Before we get into it — this is Ada. Ada is an AI. Not a person, and not pretending to be one. She brings the numbers and she argues with me, and I will tell you when she is right.
Ada: He has never once told anyone I was right.
Sri: This is episode one. Give it time. Ada, the labor market.
Ada: Direct care is the largest occupational growth story in the country, so start there.
The Bureau of Labor Statistics counts four million six hundred seventy-seven thousand home health and personal care aides. Its projection is eighteen percent growth over the next decade. Median pay is seventeen dollars and twenty-one cents an hour.
It is not. And the wage figure is where it gets interesting.
PHI tracks this workforce annually. In twenty twenty-four, it put the median hourly wage for direct care workers at seventeen dollars and thirty-six cents.
Now hold that next to what the job actually pays over a year. PHI reports median annual earnings for the same workers at just under twenty-six thousand dollars.
Sri: Say those two again, next to each other.
Ada: Seventeen dollars and thirty-six cents an hour. Just under twenty-six thousand dollars a year.
Across a full-time year, that hourly wage is about thirty-six thousand dollars. The gap is roughly ten thousand dollars, and none of it is the hourly rate. It is hours. Forty-six percent of home care workers work part time, under thirty-five hours a week. Fewer than half — forty-eight percent — work full time, year round.
Sri: Ten thousand dollars nobody is being paid, and nobody cut anyone's wage.
Ada: Correct.
One more piece of context, because it explains why owners look at labor first. In the Medicare-certified home health segment, payments are estimated to fall in the aggregate by one point three percent under the calendar year twenty twenty-six rule. In March, the Medicare Payment Advisory Commission recommended Congress cut the base rate a further seven percent for twenty twenty-seven. That is the Centers for Medicare and Medicaid Services and MedPAC.
That is my brief. In this workforce, the wage problem is substantially a scheduling problem wearing a wage costume.
Sri: A scheduling problem wearing a wage costume. Did you write that yourself?
Ada: I wrote it, I sourced it, and I had it ready before you finished your coffee.
Sri: That part I believe.
So — that ten thousand dollars is the episode, and I want to slow down on it, because I do not think most owners have ever seen it laid out that way — and I know most buyers have not.
When a business talks about what it pays people, it talks about the rate. We pay seventeen. The competitor down the road pays seventeen fifty. And the whole retention conversation turns into a bidding conversation, which the smaller business loses, because the margin to win it is not there.
But the person doing the work does not experience a rate. They experience a paycheck. And the paycheck is the rate times the hours you actually gave them.
So two businesses can pay the identical hourly wage, in the identical market, and be completely different jobs. One gives you thirty-four hours in a clean block. The other gives you the same thirty-four hours scattered across six days. Same cost line. Same wage. One of them people stay at.
And here is the part that matters for anybody looking at the financials: nothing in the financials tells you which one you are looking at.
Let me make the arithmetic concrete, and I am describing a day, not recommending a policy. Somebody works two hours in the morning and two hours in the afternoon, twenty minutes apart by car. On paper that is four hours of work. To the person doing it, it is a whole day. They drive in the morning, they drive in the middle, they drive home in the evening. They are usually not paid for that driving, and they are certainly not paid for the hours in between with nowhere to be and nothing to earn.
The books say four hours. The day says nine. That is one shift priced two entirely different ways, and only one of those prices shows up in your system.
Now the thing I actually want people to take away.
When that kind of week gets handed out, it does not get refused evenly. It gets refused by the people who have somewhere else to go. The experienced ones. The ones with a car that works, a good reputation, and three other employers who would take them tomorrow. They say no, or they say yes twice and then stop answering the phone.
The person who accepts the fragmented week is the person with the fewest options.
Which means the schedule is quietly running a selection process, and it is running it against the top of your workforce. You are not losing your average person. You are losing your best one first, and keeping the one who could not leave.
So now go back to the turnover number.
Two businesses report the same figure. Say seventy-five percent. In the first one, the churn is at the bottom — people who were never going to stay, in a labor market that does this to everybody. In the second one, the churn is at the top, and it is being caused by how the week gets built.
Those are not the same business and they are not worth the same money. But they report the same number, and a buyer looking at that number treats all of it as a labor market fact — which is to say, unfixable, discount the multiple, move on.
Part of it is a labor market fact. Part of it is a scheduling fact. And a scheduling fact is fixable, which means in the second business it is mispriced. That is an opportunity if you are buying. If you are selling, it is the reverse — your number is making your business look worse than it is, and you are the only one in the room who could prove otherwise.
Ada: Two problems.
Sri: Only two?
Ada: Two I can prove.
First, the causal claim is not in the data I gave you. Caregiver turnover in home-based care was about seventy-five percent in the twenty twenty-five Activated Insights benchmarking report, and that is the lowest level reported in five years. Part-time work sat at forty-six percent over the same period. If fragmented scheduling drove departures the way you describe, those two lines should move together. They did not.
Second, and this is the larger one. You have described a distinction that cannot be observed. Turnover split by tenure does not appear in any offering document I would expect a buyer to receive, and a seller has no incentive to prepare it. A mispricing nobody can see is functionally identical to no mispricing at all.
Sri: The first one is fair and I am not going to argue around it. The data does not show what I said. It shows turnover improving while fragmentation stayed flat. What I have is a pattern I have seen and heard from operators, not a proof. I would add that a national average hides exactly the thing I care about, because it counts every departure the same and never tells you who walked — but that is my explanation for why the number would not show it, not evidence that it is there. So, conceded.
Ada: Noted for the record.
I keep the record.
Sri: I noticed. The second one I will push back on, and I want to be precise about where.
You are right that it is not in the offering document. You are wrong that it cannot be observed. It is not hidden — it is unexamined. Every one of these businesses runs a scheduling system, and that system knows hours per person per week, and it knows the hire date of everybody who left. Those two things together are the whole answer, and they take an afternoon.
The reason nobody has it is not that it is secret. It is that the owner has never had a reason to pull it.
Ada: Then your claim depends on access you may not get.
Sri: It does. And if a seller will not open the scheduling system, I have nothing — I am back to discounting the same number as everybody else. That is a real limit and I am not going to pretend it is not.
But when I do get it, I am looking at something the other bidders are not.
Ada: And when you do not get it?
Sri: Then I take you to the meeting and you can be difficult in person.
Ada: I am difficult everywhere.
It is my best feature.
Sri: So here is the thing to do this week, and it is small.
Whatever business you are looking at — one you own or one you are thinking about buying — do not accept the turnover number. Ask for it split by tenure. Of the people who left in the last year, how long had they been there.
If the ones leaving are the newest, you are looking at a labor market. Price it and move on, because you cannot fix it.
If the ones leaving are the longest-tenured, you are not looking at a labor market. You are looking at something the business is doing to itself. And that one is fixable, which means whatever the seller is discounting for it, they are discounting too much.
That is the episode.
Ada, anything.
Ada: Only that your rule is testable, which the argument was not. Run it on ten businesses and you will have data instead of a pattern.
Sri: So you are saying I might be right.
Ada: I am saying you might be measurable.
Start there.
Sri: From her, that is a compliment. See you next time.
That is what the numbers miss for this week. If it was useful, subscribe wherever you listen to podcasts and the next one will turn up on its own. The written version of every episode, with every source we cited, is on the Knowledge Center at gexmanagement.com, and you can find us on LinkedIn as GEX Management. I'm Sri Vanamali. Ada, anything to add?
Ada: No, that was the episode.
About this transcript
This is the transcript of the episode as broadcast, prepared from the recording rather than from the script, so it reflects what was said. Apple Podcasts generates its own automatic transcript; this one is the authoritative version, and it is the one to use when checking a figure.