Kalderos Found a $1 Billion Leak in America’s Drug-Discount Plumbing

Somewhere between a pharmacy counter and a drugmaker’s finance department, one prescription can cast two financial shadows. A safety-net provider receives the 340B price. A state Medicaid program later asks the manufacturer for a rebate on the same unit. The policy says both concessions should not land on the same drug. The databases do not always get the memo.

That collision is called a duplicate discount. It sounds like an accounting nuisance. Multiply it across manufacturers, pharmacies, hospitals, clinics, insurers, state agencies and millions of claims, and it becomes a large, angry reconciliation problem. Kalderos, founded in Chicago in 2016, made that problem its entire personality.

The company’s pitch was not that America needed another drug-pricing opinion. It needed a shared record. Kalderos ingests fragmented claims files, standardizes them, applies rules and machine-learning models, flags likely noncompliance, and routes the questionable claims to the people who can verify or dispute them. The paying customer is typically a pharmaceutical manufacturer. The workflow also reaches 340B covered entities and state Medicaid teams - counterparties that have different incentives and rarely begin with the same data.

The machine under the medicine cabinet

Founder Jeremy Docken had spent a decade consulting around the pharmaceutical “drug channel.” He knew the oddity of the market: everyone could agree that discounts should reach the right transaction, while each participant kept a different ledger. Docken founded Kalderos with Dave De Vogel and Jim Bonkowski. The first prototype, launched in July 2016, was called Grappa and let covered entities review claims. Later that year, the team began training an algorithm on their responses.

The sequence matters. Kalderos did not begin with a universal health-data platform and hunt for a use case. It began with an expensive exception - claims that appeared to receive overlapping concessions - and built outward. A 2017 pilot produced its first verified duplicate. By November of that year, the company said states had acknowledged and credited its first $1 million of duplicate discounts. By 2018, it had identified more than $100 million in misapplied discounts across programs.

Today the product family reflects those parties. Discount Hub gives covered entities one place to answer good-faith inquiries from multiple manufacturers. MDRP Discount Monitoring helps manufacturers examine Medicaid rebates. Commercial Discount Monitoring tests rebates against contract exclusions. Truzo, the name now at the center of the company’s platform, extends the logic into claims-level gross-to-net intelligence and 340B workflows.

Its first interface failed politely. Its biggest policy bet failed loudly.

Grappa did not survive as a brand. That was useful failure. Covered-entity beta testing showed Kalderos where the review flow was clumsy, so in 2020 the company retired the name and launched Review with a faster, clearer interface. The lesson was mundane and valuable: in enterprise healthcare, the model may find an anomaly, but a tired program manager still has to understand it, gather evidence and answer it.

The noisier setback concerned how the 340B discount itself should be delivered. Kalderos and several manufacturers backed a rebate approach in which a covered entity would initially pay the commercial price, submit claims data through a platform such as Truzo, then receive cash equal to the 340B discount. Supporters argued that claims-level validation would reduce duplicates and diversion. Regulators worried that changing a three-decade-old upfront-discount system would force safety-net providers to float higher acquisition costs and could interrupt access.

HRSA required prior approval and rejected proposed models. Kalderos sued. In 2025, a federal court sided with the government on the core approval question. The fight did not make the data engine worthless; it clarified the boundary. Software can make a disputed market design easier to operate. It cannot grant that design legal permission.

What roughly $50 million bought

Kalderos raised $7 million from Mercato Partners in 2019, when it employed 14 people and worked with seven of the 15 largest drug manufacturers. It raised another $28 million in 2020, led by Bain Capital Ventures with Mercato participating. A roughly $15 million late-venture round followed in 2022, at a reported enterprise value around $140 million. Later private financings brought total reported capital to about $76 million, though their individual terms were not public.

That money bought the unglamorous requirements of becoming financial infrastructure: automated acquisition, a claims data warehouse, security controls, role management, customer-success teams split by stakeholder, and the integrations needed to compare records that were never designed to meet. Kalderos says its warehouse can process more than 100 million claims per hour and that its platform validates against more than 20 third-party sources. Those are not consumer-growth numbers. They are the capacity statistics of a tollbooth.

Its customers pay because the downside is measurable. A manufacturer that cannot spot a duplicate may concede revenue twice; a provider responding to scattered inquiries burns staff time; a state operating with incomplete identifiers can request the wrong rebate. Consultants and legacy systems can investigate pieces of that chain. Kalderos’ wager was that a company-owned cloud stack spanning multiple programs could see a claim’s whole discount life cycle - and let the parties act on it in the same place.

One buyer, three users, several suspicious glances

This is a peculiar enterprise sale because the party signing the software contract does not control the whole workflow. Manufacturers have the clearest financial reason to buy: they owe statutory and negotiated concessions and want to know when a claim does not qualify. Covered entities have a different reason to log in: a single queue is less punishing than separate emails, portals and spreadsheets from every manufacturer. State payers need cleaner evidence to correct rebate requests without turning every exception into a months-long investigation.

That arrangement is also Kalderos’ moat and its headache. The more entities respond, the richer its reference data becomes; the richer the data, the more useful its risk scoring becomes. But network effects do not erase politics. Hospitals and clinics may view manufacturer-led scrutiny as friction placed on a safety-net program. Manufacturers may see missing claims detail as an invitation to overpay. States are accountable for rebate accuracy but inherit uneven data from plans and providers. Kalderos has to be useful to all three while collecting most of its commercial value from one.

The alternatives range from 340B administrators such as Verity Solutions and RxStrategies to 340B ESP, legacy revenue tools, consultants and internal teams stitching together warehouse queries. Kalderos differentiated on cross-program claims intelligence and a shared resolution workflow. It was not merely asking, “Is this provider eligible?” It was asking whether this specific transaction had already collected another concession, who could prove it, and what should happen next.

The exit was hiding in gross-to-net

The phrase “drug discount management” made Kalderos sound like a narrow 340B utility. “Gross-to-net intelligence” placed it inside a larger executive concern: the gap between a medicine’s list-price revenue and what the manufacturer keeps after rebates, discounts, chargebacks and fees. The same claims visibility that catches a duplicate can sharpen forecasting, contract enforcement and revenue control.

That framing made Model N the logical buyer. Model N already sells commercialization, revenue optimization and compliance software to life-sciences companies. On June 17, 2026, it acquired Kalderos for an undisclosed price, specifically calling out earlier duplicate-discount detection and reduced revenue leakage. Kalderos brought Truzo, claims-level 340B visibility and relationships with manufacturers and covered entities. Model N brought a larger customer base and a broader revenue platform.

The acquisition also answered a question every vertical SaaS company eventually faces: is the wedge a company, or a feature in a suite? Kalderos got far enough for the answer to be “both.” Its specialist data and workflows were valuable on their own; their distribution and strategic value increased inside a system already used to manage the rest of pharmaceutical revenue.

The part worth copying

Kalderos’ most reusable idea is not “apply AI to healthcare.” That sentence has funded enough pitch decks. The useful pattern is more specific:

1. Choose a reconciliation failure with an owner and a dollar sign. Duplicate discounts created a known payer, a known workflow and a result finance teams could measure.
2. Make the counterparty part of the product. A risk score alone creates another spreadsheet. Review and Discount Hub let the organization on the other side answer.
3. Keep the durable layer beneath changing interfaces. Grappa became Review; manufacturer tools consolidated; Truzo emerged. The normalized claims engine kept gaining context.
4. Separate operational proof from policy permission. A platform can demonstrate that a rebate workflow is technically possible without proving that regulators or providers will accept it.

The conditions matter. This model works when claims carry enough stable identifiers to match, when participants will share data, when flagged dollars exceed integration and review costs, and when the workflow fits the governing rules. It weakens with poor data, low claim volume, hostile counterparties or regulation that blocks the proposed transaction. Machine learning cannot reconcile a field nobody collected. A network product without willing nodes is an expensive empty room.

Kalderos leaves a tidy lesson from a spectacularly untidy market. The company did not eliminate America’s argument over drug discounts. It made the argument legible, claim by claim. In healthcare finance, that is often the product.