Bloomberg ran a piece this week about Claimable. An AI startup that helps patients fight denied health insurance claims. The story is genuinely moving. A kid with a rare neurological condition finally got her treatment. A PGA caddy with Crohn's got his biologic back. Mark Cuban personally intervened to get a four-month-old leukemia patient on a chartered medical flight.
Then I started doing the math.
And that is where many business analyses go sideways. The product is fantastic. The mission is righteous. The founder has real skin in the game. All of that is true. None of it tells you whether the business is durable.
This is the best case study I've seen in a long time on the difference between a great idea and a great business. Investors who lose money confuse them.
What Claimable actually is
The product is simple. You got denied by your insurance company. You pay Claimable $50. They run you through a questionnaire and generate a professional-sounding appeal letter that cites federal law, medical literature, and your own clinical record. They send it to the insurer and, if things escalate, they CC the CEO, the governor, a reporter, and Mark Cuban.
It works. About 3 in 4 appeals get reversed. The industry baseline for internal appeals is 44%. They're beating average by 30 points.
Founded in 2023. Around 20 employees. $10M raised, including from Cuban. About 4,000 total appeals filed as of the Bloomberg article.
Great founder. Great technology. Great narrative.
Now the math.
The market isn't what they say it is
The headline number you see everywhere is "850 million claims denied annually." The Wall Street Journal used it. NBC used it. Bloomberg used it. It sounds enormous.
Track the source. That 850 million number came from Claimable's own calculations, based on KFF and CMS data. Claimable made up the statistic that quantifies Claimable's market. Every journalist downstream just credited "KFF" because it sounded more official.
I don't blame them for doing it. It's a reasonable estimate. But decompose it like any acquisition pro forma.
The Massachusetts Health Policy Commission published clean data for 2024. Of all claims denied in their state, about 16.6% were denied for purely administrative reasons: duplicate filings, coding errors, missing information, wrong member ID. None of those are winnable with a smart appeal letter. You just fix the paperwork and resubmit.
Only about 6% of denials hinged on medical necessity. That's the slice where an evidence-backed letter actually moves the needle.
The rest are things like excluded services, out-of-network care, and benefit design choices. Most of those are legitimate denials. The plan doesn't cover the thing. No AI letter is going to fix a benefit that was never purchased.
So out of 850 million denials, the genuinely addressable slice where a good appeal letter can change the outcome is maybe 6-15%. Call it 50 to 130 million claims per year.
Cut it again. A lot of those aren't worth appealing. A $200 lab bill isn't worth anyone's time. A self-insured employer can usually just override a denial with one phone call. A doctor's office calling once resolves a big chunk.
Realistic appeal-worthy market? I'd put it at 20-40 million denials per year. At $50 each, that's a $1-2 billion ceiling if you capture everything, which nobody ever does.
Realistic market share for a category leader? 1-5%.
Real revenue ceiling for this business model? $10-100 million annually.
A $100 million business is a great outcome for most founders and most employees. But it isn't a venture-scale outcome for investors who priced Claimable expecting something much bigger. And it means the business has to execute nearly flawlessly to hit even the high end.
The TAM story you're reading is the founder's pitch, not reality.
The moat is a wrapper
What does Claimable actually own?
Their claimed edge is a curated knowledge base of insurance law, prior successful appeals, medical studies, and insurer clinical policies. They feed that knowledge into a large language model with a smart prompt and out pops a polished letter.
Audit each layer honestly.
Federal insurance law is public information. Any LLM can cite it. Zero moat.
Medical studies come from PubMed, which is free. RAG systems that pull the right studies for a given denial are a weekend project for a competent developer. Zero moat.
Insurer clinical policies are semi-public. They're annoying to aggregate, but a hospital system or law firm could replicate the database in about a quarter. Small moat, rented not owned.
The one real asset is the outcomes dataset. Which insurer denied which treatment, with what rationale, and what happened after the appeal. That's proprietary and it compounds.
At 4,000 appeals, the dataset is too thin to matter yet. A PBM or a large hospital system could accumulate the same data in six months if they wanted to.
Run the experiment yourself. Take a denial letter. Feed it to Claude or ChatGPT with a decent prompt. Compare the output to what Claimable produces. My guess is Claimable wins by maybe 15-25% right now.
Twelve months from now? GPT-5 and Claude 5 will be better. An open-source RAG wrapper trained on PubMed and public insurer policies will be free. The gap collapses.
This is what I call a wrapper business. The value lives in the model, not in what you built on top of it. When the underlying model improves faster than you can build proprietary value, your margin evaporates.
Every trick has a shelf life
Let's say the letter is great. Better than what anyone else can produce, today. What happens over time?
Three layers of decay are running at once.
The letter itself gets recognized. Insurance company reviewers see enough Claimable appeals that they start pattern-matching the citations, the federal law callouts, the formatting. Once recognized, it gets routed to the harder review queue with a different playbook.
The distribution hack decays faster. Claimable's secret weapon isn't the letter. It's the CC list: governor, senator, insurer CEO, Mark Cuban, a journalist. That works because it's rare. When a thousand patients all CC the same five people, the governor's office sets up an auto-filter, Cuban quietly blocks the inbox, and the CEO creates a dedicated "appeals" folder that nobody reads. Social pressure is a depleting resource, and Mark Cuban is not infinite.
Then there's the counter-AI problem. Insurers already use algorithms to deny claims. They have infinite incentive to train classifiers that recognize Claimable-generated appeals and route them into a harder review path. It's the obvious move. Any insurance company data science team that isn't already working on it should be fired.
There's a counter-argument worth naming. Claimable's outcomes dataset gets better with scale. More appeals means more learning about which arguments work with which insurers for which conditions. That's a real compounding force running against the decay.
The question is which side runs faster. I think decay wins. UnitedHealth has more data scientists than Claimable has total employees, and the insurer-side incentive to build counter-classifiers is huge. Claimable is trying to build proprietary data faster than the insurance industry can build proprietary counter-AI. That race favors the side with more money.
So the 75% win rate is probably Claimable's peak, not their floor. Every percentage point lost makes the $50 price harder to justify.
Then there's the math that actually kills companies
The unit economics on the consumer side are the reason the founders are already pivoting. You can see the problem when you lay them out.
Claimable's D2C unit economics, as best I can estimate from public data:
Revenue per customer: $50. One-time.
Repeat rate: close to zero. A patient with Crohn's might appeal twice in a decade. A family dealing with a rare pediatric condition might appeal three or four times over years. Most people appeal once in their life and never come back.
Gross margin on the letter itself: high, probably 80%+. Gross margin on the business overall: much lower, because the CEO is personally handling half the cases. That doesn't scale.
Now CAC.
Right now CAC looks fantastic. Bloomberg, NBC, WSJ, CNBC, and TIME have all done free feature pieces. Mark Cuban is personally promoting cases on LinkedIn. That's worth millions in earned media.
None of that is repeatable at scale.
Once the press moves on and Cuban gets a new cause, Claimable has to acquire customers the normal way: SEO, paid search, referral, content marketing. A sick patient who just got denied, who knows to Google for appeal help, who finds Claimable instead of Counterforce Health (free, nonprofit) or Fight Health Insurance (free for patients): the cost to find that person and convert them is tight at best.
Consumer e-commerce CAC benchmarks run anywhere from $10 on the low end (Stripe pegs consumer e-commerce SaaS at an average of $64) to several hundred for anything with a considered purchase. Claimable's in a tough spot: the product is considered, the category has two free competitors, and paid search terms around "insurance denial help" are crowded and expensive. Best case, CAC is $80-150. Worst case, higher.
Compare that to an LTV of $50. The math doesn't work at any point on that range. You can't run a subscription business without a subscription. You can't run a transaction business where CAC exceeds unit revenue.
The SaaS benchmark for a healthy business is a 3:1 LTV to CAC ratio. Claimable's D2C motion runs 0.5:1 or worse as soon as the earned media runs out.
Which is why the founders already know this and are already pivoting. Walkabout's general partner said "direct-to-consumer just doesn't work" on the record, in Bloomberg. That's the investor language for "we're moving the business."
Compare to TurboTax, which Bokhari actually names as his model. TurboTax charges $80-250, has 80% annual retention because everyone files taxes every single year, and upsells into audit defense, tax prep, bookkeeping, and financial products. Every one of those things is missing from the D2C version of Claimable. The difference between today's Claimable and today's TurboTax is the difference between a transaction and a subscription.
That's the old business. Now look at the new one.
The money lives at a different address
Where are they pivoting? Drugmaker deals. Hospital system deals.
That's the right move. If you ask Charlie Munger's favorite question, "show me the incentive and I'll show you the outcome," you find the real economic customer is not the patient.
Who makes money when a denial gets reversed?
The patient, who gets their drug
The drugmaker, who keeps a $40,000-a-year prescription
The hospital, who gets paid for a liver transplant that runs into seven figures
Nobody at the insurance company
Claimable charges $50 to the only party in that stack who can't actually afford it. The drugmaker on a biologic prescription would gladly pay several hundred to a couple thousand per saved prescription. Pharma already spends $4-6 billion a year on specialty hub services doing prior authorizations and appeals for exactly this kind of work, per Health Affairs research. Claimable would be one more vendor in a budget line item that already exists.
So the pivot is correct. That doesn't mean it works.
Pharma sales cycles run 12 to 18 months from first conversation to signed contract. Claimable has about 20 employees and $10 million raised. Four signed deals that the company won't name sounds more like pilots than revenue. Pilots do not pay real money and they often die before they scale.
Selling into pharma is a different business than building a consumer app. You need a sales team with pharma access relationships, legal experienced with HIPAA and commercial contracts, a compliance function, and a services layer that can deliver what pharma actually wants (data reporting, outcomes tracking, integration with existing hub providers). None of that is in the Bloomberg story because the Bloomberg story is about the consumer product. Whether the company has any of it internally is an open question.
The companies that already do this at scale (AssistRx, Eversana, Covance, dozens of others) have spent 20 years building those relationships. Claimable has to either take share from them or convince pharma to carve out a new spend category alongside them. Neither is fast.
The optimistic case: Claimable closes $5-10 million in annual pharma contracts within 18 months, proves out a drugmaker-funded appeals motion, raises a Series B at a reasonable valuation, and becomes a real business. The specialty hub market is big enough that a startup with good technology and PR momentum can plausibly win a seat.
The pessimistic case: the runway burns while pilots stay pilots, the D2C motion can't justify its CAC once the press moves on, and the company ends up either acquired at a modest price by Cuban or one of the specialty hub incumbents, or quietly folded. The $10 million raised does not buy 18 months of enterprise pharma sales cycles plus a consumer business that loses money on every customer.
Which case plays out depends on things I can't see from the outside: how many of the four deals are real, how much runway is left, and whether the team can hire enterprise pharma sales leadership fast enough.
Five questions that kill most deals
I want to be careful here. I think Claimable is a great idea. I think Bokhari is doing God's work. If I had a denied claim, I'd use the service.
None of that means it's a great business in its current form.
A business is not a cause. A business is a machine that takes capital in one end and returns more capital out the other, durably, over years. Every part of that machine has to work: the market has to be big enough, the customer has to be willing and able to pay more than it costs you to serve them, the moat has to hold up when the technology underneath you changes, and the customer has to come back or you have to go find the next one cheaper than the last one cost you.
I'm not blaming the founders. This is a reminder for the rest of us.
The same pattern shows up everywhere. I see it in proptech all the time. A fantastic product that solves a real pain, priced to the wrong buyer, with a wrapper moat that will get eaten by the next model upgrade, and no repeat-purchase mechanism. Everyone in the room loves the demo. The economics never work.
Before I write a check or commit my time, I run five questions against the business:
Is the market big enough, decomposed honestly, not in the pitch deck?
Am I charging the person who feels the pain or the person who can pay?
Do I have a moat that isn't just the current model generation?
Does the customer come back, or do I have to find a new one every time?
What does CAC look like when the free press dries up?
Claimable, as currently structured, struggles on every single one of those. The pivot to pharma could fix questions two, four, and five if it closes in time. It doesn't fix three, and three is the one that gets harder every month the underlying models improve.
I hope they make it. Anybody who saves a kid with PANS from a wrongful denial has a seat at my table.
But I don't write checks based on how much I like a founder.
The business behind the headline is much harder than the headline suggests. It usually is.
https://apple.news/ALoUg7CdbThqM29XGtXzzlw