Evidence · Pharmaceutical IP
The annual report credits the molecule and the commercial team. The patent estate that kept the revenue alive after the base patent expired appears on no line of the P&L — which is why the department that built it gets discussed as overhead.
Roche is the clearest counter-example in the industry. We reconstructed how it handled the largest patent expiry in its history, entirely from public records. The finding: the lost franchise was not defended. It was replaced — and the replacement was engineered in the patent file.
Public-source analysis · every figure traceable to a dated primary filing
Written for Heads of IP · Chief IP Counsel · Portfolio & Licensing Leadership
The problem
Most IP functions can describe what they protect. Very few can state what that protection earns, in currency, this year. The gap is not a failure of effort — it is structural. Revenue is reported by product, while exclusivity is held in layers of rights that cut across products, and no accounting convention connects the two.
The consequence is predictable and familiar. Prosecution budgets are argued on cost per filing rather than revenue per year of exclusivity preserved. Lifecycle work is approved late, when the cliff is visible, rather than early, when the options are cheap. And the department carries a reputational burden it has not earned: treated as insurance against a bad outcome, rather than as a producer of a good one.
Roche's record over the last seven years is the most useful available rebuttal, because the counterfactual is unusually clean: a patent cliff of known size, on a known date, against a portfolio whose replacement was visible in the filings years in advance.
The case
By 2018, three Genentech-originated oncology products — Avastin, Herceptin and Rituxan/MabThera — represented roughly half of Roche Pharmaceuticals' revenue, about CHF 21 billion. Their primary US patents expired across 2018 and 2019. Roche's own planning assumed biosimilar erosion of 60–70%.
The erosion arrived, and was steeper than forecast: those three products fell to roughly CHF 3.25 billion by FY2025. Yet across all of Roche's expired-patent products, the combined decline in FY2025 was only CHF 0.7 billion — because patients had already been moved onto successor medicines the company controls. Pharmaceuticals grew to CHF 47.7 billion.
The lost franchise was not defended. It was replaced — and every replacement carries protection filed long after the originals'.
The mechanisms
Each is visible in the public record, and each attaches to a revenue figure. Together they are the most complete worked example we know of an IP function operating as a revenue engine rather than a defensive one.
From a single originating antibody whose composition patent has long expired, Roche built a succession of separately protected products: a distinct molecule deployed in combination, the same antibody re-engineered as an antibody-drug conjugate on in-licensed linker technology, and a fixed-dose subcutaneous co-formulation.
HER2 franchise ~CHF 7.4bn in FY2025. The base molecule's patent contributes nothing to it.
Roche's single largest product is protected less by its original product patent than by a sequence of method-of-use rights running more than a decade beyond it, since extended further by a subcutaneous co-formulation. The most valuable revenue line in the portfolio sits on rights that postdate the molecule.
Ocrevus: CHF 7.0bn in FY2025, with method-of-use protection documented to 2035/36.
Rather than defend an eroding asset, Roche displaced it with a patented successor in the same indication — recapturing biosimilar-exposed revenue under a new exclusivity term. In FY2025 the successor demonstrated a 41% reduction in risk of death measured directly against the product it replaces.
Revenue exposed to biosimilars, returned to protected status.
Subcutaneous and fixed-dose co-formulation now runs across at least five franchises. Each resets an exclusivity clock without a new molecular entity, and each is justified clinically as well as legally: the newest cuts administration from a two-to-four-hour infusion to roughly one minute.
Five franchises converted. Fastest-growing product in the top ten grew 48% in FY2025.
Sources: Roche FY2025 results (29 January 2026); Rapid Alpha Part 1 analysis; public USPTO and EPO records. Product-level figures are listed with their sources in the full report.
The method
Most readers assess a company by looking backwards at reported revenue, which is a lagging indicator. We read the same company forwards, from its patent and pipeline record. We call the result the αSignal.
The method is straightforward to state and laborious to execute. For every molecule entering a pipeline, code three things: who originated it, how the company accessed it, and the stage at which it entered. Then track the patent activity around it. The combination identifies which franchises will carry revenue years before that revenue appears in the accounts.
It works because of timing. Patents and access decisions sit upstream of revenue by years — and Roche's own behaviour makes the point vividly.
| Access mode | NMEs | Share | What it means for the patent estate |
|---|---|---|---|
| Build (internal) | 94 | 58% | Originates and owns foundational rights outright, unencumbered by in-licensing terms |
| Partner | 53 | 33% | Jointly-held filings, staged as options rather than commitments |
| Buy | 15 | 9% | Acquired estates, used selectively |
Roughly 86% of those molecules entered at Phase 1 — the earliest and riskiest point, where most candidates fail. That looks like risk-taking, and it is. Roche manages it through the mode of access rather than the timing: building internally where its own science gives a selection edge, and using partnering and acquisition as staged options.
The evidence that it works is the hit rate. Roche converted new molecular entities at roughly 8%, against about 4% for the industry — twice the success rate on the riskiest shots on goal. Most large pharmaceutical companies do the opposite, de-risking by acquiring validated assets late and paying a premium for science they must then license or share.
For an IP function the implication is direct: entering early and building means the patent position is created at the point of lowest cost and maximum optionality, and held free of the encumbrances that late-stage in-licensing carries.
The validation
In our 2008–2018 analysis we identified Genentech as originator or co-originator of 32% of all molecules entering Roche's pipeline, and 49% of those in oncology. We also flagged something less remarked at the time: Genentech was building Phase 1 positions in neurology and ophthalmology, areas where it had not historically led.
Several of the assets those entry tables identified were, at the time, unapproved — recorded as having no first approval. By FY2025 they are among Roche's largest products. Seven of the ten best-selling originate with Genentech, and the two highest-growth franchises sit precisely in the neurology and ophthalmology positions the data flagged.
We state the limits plainly, because a reader who checks our work will find them anyway. This is a directional, single-case validation — one company, one seven-year window, predictions that matched. It is not yet a backtested signal with a measured hit rate across a peer set. Establishing that footing is the next phase of the work.
The offer
The Roche analysis is a demonstration, not a product. The work we do inside an IP function begins deliberately small, because the output should be checkable before anything larger is commissioned.
One product family. Every right sustaining it, revenue attributed to the post-expiry layers rather than the base patent, and a dated cliff calendar.
Built fromIP Plan · IP Rights Portfolio Development
The same method across the estate, refreshed annually. Where protection is thinner than assumed, and what the prosecution budget buys in revenue-years.
Built fromIP Plan · Freedom to Operate (FTO) Support · IP Capture
A named competitor read forward from filings and pipeline entries — the αSignal applied continuously rather than once.
Built fromCompetitive Landscape Report · IP Research · Rapid Assessment
Every engagement starts with a single franchise and a scope fixed before any work begins. What it costs depends on the size of the estate and the condition of the records, and we price it in advance, in writing. The entry piece is deliberately small enough to be judged on its output.
Check the work
The mechanisms in the Roche case are specific to regulated life sciences. The method behind them is not. The common thread is the argument this analysis makes about Roche: an IP position built deliberately against the commercial narrative, and made legible to the people who fund it.
10 daysvs. 6 months
Technology and IP diligence that used to take half a year, delivered in two working weeks.
Source: Rapid Alpha client benchmark, 2025
Developer of translational research tools, listed on Nasdaq at the time of our work. We conducted in-depth analysis of competitor patent activity in support of NanoString's litigation position — a position it went on to win on appeal in Europe, reversing an earlier decision — and delivered a competitive landscape report identifying acquisitions that could grow market share.
Built the first end-to-end platform for prescription digital therapeutics. We guided prosecution of the patent portfolio behind that position — protection that supported successive capital raises and a first-of-its-kind FDA authorisation for a prescription digital experience addressing traumatic brain injury.
Develops RNA therapeutics matched to the underlying genetic drivers of disease. We provided IP planning and claiming strategy across the platform, and supported the company through its capital raise.
Builds the industry's leading DLE systems and sorbents, recovering lithium from oil and gas wastewater at commercial scale. We work with Element3 so the breakthroughs driving that commercial success are both protected and valued by investors.
Builds drone systems and the operating software behind them. We developed the IP strategy and architected the portfolio around the company's commercial narrative — AI-assisted operation and sensor fusion, so that anyone can fly, in any environment, and do more once there. The work identified over 47 protectable inventions and supported a USD 20M Series A. Xtend has since gone public.
The deliverable carries the credibility, not the logo. Every conclusion in our work traces to a primary source with a date — which is why the Roche analysis uses no confidential information of any kind.
The report
Accessing Innovation — Roche & Genentech, Part 2: 2019 to Present. This page carries the spine of the argument. The report carries the evidence: the replacement map scored drug by drug, the full transaction set, the patent mechanics behind each franchise, and the source behind every number.
Public-source throughout. Every figure traces to a dated primary filing — Roche results, USPTO and EPO records, published trial data. No confidential information of any kind.
That email address didn't look right — please check it and try again.
No. Every figure is drawn from public sources: Roche's annual and full-year results, public USPTO and EPO patent records, and published clinical trial data. That is deliberate — the method has to work from outside a company for it to be useful against competitors.
The finding that IP-linked pipeline data is a leading indicator of future revenue. Track who originated each molecule, how the company accessed it, and the patent activity around it, and you can identify which franchises will carry revenue years before it appears in the accounts. The name follows the investing sense of alpha: return above the benchmark, here meaning the informational edge from reading a company forwards rather than backwards.
The mechanisms described here — layered rights, method-of-use protection, formulation-based term extension — are specific to regulated life sciences, because that is where exclusivity is granted and contested in those particular forms. The underlying method is not. Coding origination, access mode, entry stage and patent activity works anywhere a portfolio is built ahead of revenue. We have run it across consumer electronics, networking and security, automotive, aerospace, mining, agriculture, autonomous systems, digital health, and direct lithium extraction — where the category itself is new enough that the patent estate is still being set.
The usual starting point is a bounded mapping exercise on a single franchise: attribute its current revenue to the specific rights sustaining it, and identify where exclusivity is thinner than assumed. It is deliberately small, because the output should be checkable before anything larger is commissioned.
Prepared from public sources. Roche, Genentech and product names referenced are trademarks of their respective owners; nothing here implies any relationship with or endorsement by them.
