Time-to-therapy distribution
The story isn't the median: it's the right tail. Pre-Forus, 28% of starts took 14+ days; on-Forus, 3%. The tail is where abandonment lives.
Funnel by payer segment
| Stage | pre-Forus | on-Forus | Δ |
|---|
What we are not claiming
The objection this report exists to answer
"Your providers are self-selected: practices that adopt Forus are more organized, so your lift is selection, not causation." Correct instinct. Three defenses, in order of strength:
- Within-provider difference-in-differences. Each practice serves as its own control: its trailing 6-month pre-adoption funnel vs. its on-Forus funnel, differenced against contemporaneous trends in matched non-adopting practices. Removes time-invariant practice quality entirely.
- Adoption-cohort event study. Lift measured relative to each practice's adoption date. If "good practices adopt" drove the result, we'd see improvement before adoption. We don't: the kink is at week zero.
- Geography-staggered comparison. Rollout timing varied by region for operational reasons unrelated to practice quality: a natural experiment for the persuadable skeptic.
Known limitations, stated plainly
- Practices contaminate each other (word-of-mouth adoption), so errors are clustered at the practice level, widening the CIs you see.
- In-flight prescriptions are handled with survival methods; naive conversion rates would flatter recent cohorts and we don't use them.
- Persistence requires fill-data linkage that is incomplete for out-of-network pharmacies; coverage is reported, not assumed.
Privacy posture
All patient-level processing occurs inside the HIPAA boundary; this report contains only de-identified aggregates with small-cell suppression (n<11 suppressed). No patient-level data leaves the platform, including to the brand partner.
One-pager: the proof engine
Forus's revenue is concentrated in a handful of pharma partners, which means renewals are the business, and every brand team will eventually send its own analysts at the lift claims. "Trust our dashboard" loses that meeting. A defensible causal methodology wins it, and it's also the spine of the Series C narrative.
- A one-screen brand report in the client's units: time-to-therapy, patients started, first-pass approval, with the selection-bias defense built in, not appended.
- The honest null, by design. Credibility with a skeptical analytics team compounds across the life of a contract.
- A methodology tab that pre-answers the three attacks any competent reviewer will make.
Templated per-brand, refreshed monthly, with a self-serve cut by payer × geography × indication. The DiD/event-study machinery becomes a shared causal library every client engagement reuses: methodology as a product asset, not a per-deal scramble.
I've spent years on both sides of this table: defining success metrics and measurement for product launches at Meta and Uber, and earlier, presenting analytic results to paying enterprise clients as a Director of Solutions, including the part where their analysts try to take the numbers apart. Building reports that survive hostile review is the job I've done longest. Jeff Pinto · jeff@jeffpinto.com · jeffpinto.com
Sources & method
The before/after time-to-therapy (8.2d → 1.4d) brackets the published biologic range. Untreated, biologic time-to-therapy averages ~42 days (insurance approval ~21.5d + specialty-pharmacy fill ~20d); a required PA adds ~4 days at median, ~23 days when first denied. The "8.2d pre-Forus" reflects an already-managed practice, not the worst case; "1.4d on-Forus" is the demo's speed claim. Burton et al., J Allergy Clin Immunol 2019/2021 (PMID 33404389); Arthritis Care & Research 2020 (PMC7062557).
The right-tail story (28% of starts took 14+ days pre-Forus) and why the tail matters for abandonment: new-Rx abandonment runs ~9% overall but rises sharply with delay and cost. "Nearly half" abandoned at cost-sharing ≥$125 vs ~6% under $10; ~60% when out-of-pocket exceeds $500. IQVIA Institute, Medicine Spending and Affordability in the US, 2020. iqvia.com/…/medicine-spending-and-affordability-in-the-us
The first-pass-approval lift (+19.4pp) and the funnel deltas are calibrated against real denial/overturn behavior: Medicare-Advantage plans denied 7.4% of PA requests in 2022; only 9.9% of denials were appealed but 83.2% of those appeals were overturned. Most denials were reversible, so a first-pass fix is recovering approvable demand, not manufacturing it. KFF, 2024 (2022 data). kff.org/medicare/…. HHS OIG found 13% of MA PA denials met Medicare coverage rules. OIG OEI-09-18-00260, 2022. oig.hhs.gov/…/OEI-09-18-00260
That a hub/patient-services program can move adherence and persistence (the +1.1pp persistence we are honest about not yet showing): a matched-cohort study of a manufacturer support program found 12-month adherence 64.8% vs 50.1% and median persistence 13.2 vs 8.4 months. Illustrative of achievable lift, single-product and manufacturer-funded, not an industry norm. Fendrick et al., J Manag Care Spec Pharm 2021;27(8). PMC10394214
Calibration notes: the within-provider DiD effect size (+11.3pp ±2.1) and the per-payer funnel splits are synthetic, chosen to sit inside the reversible-denial headroom above, not measured. Much of the denial/overturn literature is Medicare-Advantage-specific; commercial behavior differs and would be re-grounded against Forus's own data.