Kooth is the only publicly traded mental-health business I know of that prints software margins on a clinical product. The London market does not appear to have noticed. FY24 revenue: £93.7 million1. Gross margin: 41%2. The customer is, almost entirely, the state — the NHS in the United Kingdom, a growing list of state Medicaid plans in the United States — and the unit cost curve looks like a SaaS business, not a provider.
The interesting question isn't whether Kooth's revenue is good. (It is.) It's whether the model is durable, replicable, and rerateable. My read across the last four annual reports, two trading updates, and the Pennsylvania contract documents is that two of those three are true, and the third — the rerate — is the trade.
22.01 · OpeningWhat Kooth actually is
Kooth is a digital mental health platform sold to public-sector buyers — NHS commissioners and US state Medicaid plans — and made available, free at the point of use, to a defined population. In the UK, that's typically all 10-to-25-year-olds in a given Integrated Care Board catchment. In the US, it's all Medicaid-eligible 13-to-25-year-olds in the contracted state. The unit they sell isn't a seat. It is a population.
Per-population pricing vs. per-seat pricing
Per-seat economics — the BetterHelp / Cerebral / Talkspace D2C model — bill the customer for each registered user. COGS is linear in users. Per-population economics — Kooth's model — bill the customer once per contracted population (an ICB, a state). COGS is linear in peak concurrent users, which grows far slower than enrolled population. Gross margin compounds with density. This is the whole thesis in one sentence.
That distinction is the entire investment case. A per-seat business has linear cost of revenue; a per-population business has step-function cost of revenue. Kooth's clinical team is sized to peak concurrent users — not to enrolled population — and as the contracted population grows faster than peak concurrency, the gross margin walks up. This is, mechanically, why a 2018-vintage NHS contract still prints a higher margin than a 2024 US state contract: time and density3.
This is not a model that BetterHelp, Talkspace or Cerebral can run on their existing balance sheet. Their unit economics are sized for D2C paid acquisition. Kooth's are sized for a procurement cycle. I wrote about that re-orientation at Talkspace four issues ago — the read across is that the consumer operators cannot get here from there4.
22.02 · Unit mathWhat the numbers actually do
Annual recurring revenue, 2020 → 2024
The mechanical reason this re-rate hasn't happened yet is that the UK side of the business obscures the US side in the consolidated P&L. Strip the UK out and what's left is a SaaS-shaped revenue ramp with a single signed multi-state contract and a credible pipeline of eight more. The market is reading the consolidated number and missing the second derivative.
22.03 · The moatThe procurement moat
State procurement is slow, sequential, and politically expensive. It is also — once won — one of the most durable customer relationships in the economy. The reason Kooth is interesting and not, say, the fifteen US start-ups currently pitching state Medicaid plans is that Kooth already has Pennsylvania. It already has California6. It already has a deployment playbook that has been signed off by the procurement legal teams of two of the largest state governments in the United States. The next twelve states will not require Kooth to invent a category. They will require Kooth to copy-paste an existing contract.