Nobody Knows Their Realization Rate, Including Us

I set out this month to publish the first Realization Rate benchmark for pharma, and I could not build one. What stopped me is more useful than the number would have been, so this is the finding instead of the benchmark.

I set out to publish a benchmark and could not build one

Realization Rate is the measure my practice is built on. It asks what share of the therapeutic and commercial value a company earns at the moment of a clinical decision actually becomes a patient on sustained therapy. The arithmetic is simple. Realized value divided by earned value, with the clinical decision as the denominator, because that decision is what the commercial system worked to produce.

The plan was to assemble that measure from published evidence across several therapeutic categories, publish the arithmetic openly, and let anyone check it. Five stages, in the order a patient moves through them. A prescription written, then filled, then started, then held at ninety days, then held at twelve months. Each stage expressed as a share of the prescriptions written, so the loss accumulates visibly down the ladder.

Four researchers worked four category sets against that structure. They searched the peer reviewed literature, the vendor claims analyses, the payer disclosures and the academic repositories. They came back independently with the same conclusion, which was that the ladder cannot be built from published evidence, and that the reason it cannot be built is not a gap in the search. It is a gap in what the industry measures at all.

The audit covered six categories and one of four rungs is measured

Below each category is a prescription written, which is the denominator and is measured by construction rather than by evidence. Beneath that sit four stages where a figure could exist. Here is what exists.

Of the four stages below the prescription, one is measured. Three are not, in any category, anywhere in the published literature. That pattern appears to hold across every category my researchers examined, and it held for the cross category view as well, where the best available figure is a vendor estimate that roughly three in ten new branded prescriptions go unfilled.

Read the one measured column on its own and it is already a serious finding. Depending on the category, somewhere between ten and sixty-nine percent of the clinical decisions a commercial organization earned never reach the patient's hands at all. The PCSK9 inhibitor figure is the one I keep returning to. Of 45,029 patients newly prescribed one of those drugs, 30.9 percent ever received it.

Claims data records a dispense and has never recorded a dose

The missing stages are probably not missing by accident, and the second one explains most of it. Pharmacy claims record a dispensing event. They cannot record an ingestion event, because nothing in the payment system observes a patient taking a tablet or pressing an autoinjector against her thigh. So the population that collected a prescription and never started it has never been counted at scale in any category.

This matters more than it sounds. Every study published under the heading of primary non-adherence turns out, on reading the methods, to measure prescribed and never dispensed, which is the fill stage again under a different name. I had assumed those studies measured the filled-but-never-started population. They do not, and stacking one figure beneath the other would have double counted the same patients. That mistake is easy to make and I nearly made it.

The ninety day and twelve month stages are missing for a different reason. Figures do exist, in volume, but every one of them is published as a share of the patients who started therapy rather than as a share of the prescriptions written. The denominator changes between stages, and once it changes the stages cannot be chained. Multiplying a persistence rate by a fill rate from a different database and a different cohort produces a number, and the number is an artefact of which sources were chosen rather than a measurement of anything.

A rejected prescription is more often extinguished than redirected to something else

Where the literature is strong is at the fill stage, and what it shows there has changed in the last six years. A study published in JAMA in July of 2026 looked at more than two million first fill attempts for single-source branded drugs across 1.17 million people, spanning commercial, Medicare, Medicaid and marketplace coverage from January 2018 to September 2024.

Rejections rose from 24.3 percent of initial attempts in 2018 to 40.7 percent in 2024, which is an increase of sixty-seven percent in six years. Thirty-two percent of attempts were rejected on formulary exclusion or utilization management grounds. The rate varied twelvefold by class, from 85 percent for incretin weight-loss therapies down to 6.7 percent for oral anticoagulants.

The figure that should stop a commercial leader is the next one. Of the attempts that were rejected, 48.4 percent were followed by no fill of that drug and no fill of any drug in the same therapeutic class within ninety days. The common assumption is that a rejection redirects a patient toward a formulary alternative, and the data suggests that is what happens slightly less than half the time. The rest of the time the clinical decision is simply extinguished. A physician reached a judgment about a patient, the system declined it, and ninety days later that patient is on nothing in that class.

In the organizations I have worked inside, that outcome is recorded as an access problem and routed to a market access team, where it becomes a contracting conversation. It is also an experience failure, a measurement failure and a therapeutic failure at the same time, and no single function is accountable for the compound.

Every persistence figure in pharma is computed on the survivors of access

Here is the consequence that I think the industry may not have reckoned with. If only the fill stage is measured, and every later measure is published as a share of the patients who got through that stage, then the entire apparatus pharma uses to judge patient behaviour is computed on a surviving cohort.

Return to the PCSK9 inhibitor study, because it makes the point in a single set of numbers. Of 45,029 prescriptions, 79.2 percent were rejected on the first day after submission. Forty-seven percent were eventually approved. Of those approved, 34.7 percent were abandoned at the pharmacy counter, leaving 30.9 percent that ever reached a patient. Abandonment among approved prescriptions ran from 7.5 percent where the copay was zero to above seventy-five percent where it exceeded $350.

Any adherence or persistence figure published for that drug class is calculated on the roughly three in ten who survived. That surviving group was filtered twice, once by the payer and once by the price. It is selected for coverage, for income and for the determination to persist through two refusals, which may well be the most motivated cohort in the entire prescribed population. A persistence rate computed on them is partly a measure of their wealth.

The distortion does not even run in a consistent direction, which is what makes it so difficult to work around. One study of 125,474 patients found twelve month discontinuation of GLP-1 therapy at 64.8 percent among patients without type 2 diabetes and 46.5 percent among those with it. Same molecules, same sixty day gap rule, an eighteen point spread. The population facing the harsher access gate also shows the worse persistence there. A measure that flatters performance in one category and penalizes it in another is not measuring commercial performance at all, and it should not be sitting on a commercial scorecard.

The most quoted adherence statistic rests on a 1979 assertion

One more thing surfaced that I had not gone looking for. The figure everyone in this field quotes, that adherence to long-term therapy in chronic disease averages around fifty percent, comes from a World Health Organization report published in 2003. I asked a researcher to find its primary source rather than cite the report, and the chain is not what its fame implies.

The WHO report states the figure as a narrative assertion with no sample size, no confidence interval, no stated measure and no time horizon. It cites a Cochrane review, whose own background section attributes the number to work published in 1979. That Cochrane review was an assessment of interventions across thirty-three trials, it produced no pooled prevalence estimate, and its authors noted the studies were too disparate to combine.

So the most repeated number in adherence is an inherited claim from the 1970s, it describes adherence rather than twelve month persistence, and it is not specific to the United States. I have quoted it myself in the past. I will not again, and saying that plainly seems more useful than quietly dropping it.

A company holds the data for its own brand that the literature lacks

None of this means a manufacturer cannot know its own Realization Rate. The gap I have described is an industry level gap in published evidence, and it is not the same thing as a gap inside a company.

A brand team can see its own rejection rate, because the rejections arrive as claims data. It can see its own abandonment at the counter through its hub and specialty pharmacy partners. It can see refill behaviour through dispensing data it already buys. The one stage nobody can see, the patient who collected a prescription and never started it, can be approached through its patient support programme and its field organization, which are in contact with the people the claims cannot observe. That is not a perfect instrument, though it tends to be a great deal better than nothing, which is the current state.

What a company cannot do is benchmark the result, and I want to be straight about that limitation rather than sell past it. There is no industry median to compare against, because the work to produce one has not been done by anyone, including me. The first useful comparison is a brand against itself over time, and the second is one brand against another inside the same portfolio, where the data definitions are at least under one roof.

This is also why I think the measure matters more than its competitors rather than less. Realization Rate is not a better metric among a field of adequate ones. It is a measure anchored at the clinical decision because every alternative the industry currently runs begins counting after the loss has already happened, on a cohort that the loss itself selected.

The admission is the standard rather than a weakness to manage

I expected to publish a number this month and I am publishing an empty table instead. The table is the more honest artefact, and I suspect it is the more useful one, since an executive can act on a known blind spot in a way she cannot act on a borrowed benchmark.

The question I would put to a commercial leadership team is not what their Realization Rate is, since almost nobody can answer that today. It is narrower and rather harder to deflect. For the brand that matters most to the business this year, how many of the prescriptions written for it last quarter reached a patient, and who in the organization is accountable for the gap between those two numbers. If the answer involves three functions and no name, the measurement problem described above is already operating inside the company, and it is being read as an access problem.

I will keep working on the benchmark. If a manufacturer wants to establish the first defensible Realization Rate for a therapeutic category using its own data, under its own definitions, that is work I would take on, and the methodology below is where it would start.

Methodology, sources and the challenges I would make to this myself

Every figure in the table is a published share of prescriptions written or the arithmetic complement of one, with the conversion shown. No figure was interpolated, estimated or carried across populations. Where a stage had only a conditional figure, published as a share of patients who initiated therapy rather than of prescriptions written, the cell was left empty rather than converted.

The GLP-1 figures come from a JAMA Health Forum research letter examining 9,848 GLP-1 orders in Colorado between January 2018 and September 2022. It used UCHealth electronic health records linked to the Colorado All Payer Claims Database, covering Medicare and commercially insured patients. The overall fill rate was 60.1 percent, splitting to 37.2 percent for the obesity indication and 47.5 percent for the diabetes indication. Average out-of-pocket cost on a filled thirty day supply was $71.90, rising to $134.04 where obesity was the sole indication.

The oral oncology figures come from a 2026 Journal of Clinical Oncology study of more than 12,000 Medicare and commercially insured patients newly prescribed specialty oral anticancer drugs for blood cancers in 2022. Initial rejection at first pharmacy submission was 64.9 percent under Medicare and 84.0 percent under commercial coverage. The share both approved and filled within ninety days was 54.5 percent and 45.5 percent respectively.

The biologics figure is derived from a 2018 study in Arthritis Research and Therapy covering 434 patients with rheumatoid arthritis newly prescribed biologics or tofacitinib. It drew on Optum claims linked to electronic health records, and found that 40.6 percent failed to initiate within three months. The PCSK9 inhibitor figures come from a 2017 JAMA Cardiology study of 45,029 patients. The statin range comes from a 2024 systematic review in BMC Primary Care covering eighteen United States studies, which publishes a range rather than a point estimate.

The rejection figures come from Levy, Alexander, Vabson and Ippolito, published in JAMA on 9 July 2026, covering more than two million first fill attempts across 1.17 million individuals. The GLP-1 discontinuation comparison comes from a 2025 JAMA Network Open study using Truveta records from over 800 hospitals.

Four challenges a reader should make to this, which I would make myself. Each one narrows what the table can be used to claim, and none of them is answered by better searching.

First, the fill figures come from different years, payer mixes and geographies, so the table may be read as comparing categories rather than measuring a single market. The biologics figure in particular describes prescribing from 2007 to 2015, before the agents that now dominate that category. Second, a prescription that was written and never filled is not necessarily a commercial loss. Screening requirements, a pregnancy, a patient who improved on prior therapy and a prescriber writing speculatively to open a benefits investigation can all appear in the data as the same event. The fill figures are therefore an upper bound on recoverable loss rather than an estimate of it.

Third, prescriptions written is itself not quite the right denominator, since every claims source counts prescriptions submitted to a pharmacy, so scripts never transmitted from the office are invisible. Fourth, in oncology, death and disease progression compete with abandonment as explanations for a prescription that was never filled, and the published studies handle that inconsistently.

Those four challenges are why I am publishing a table with empty cells rather than a benchmark with a decimal point. A number this fragile would have been taken apart by the first competent reviewer, and correctly.

About the Author

Wayne Simmons is the founder of The Customer Excellence AGENCY and the author of The Customer Excellence Enterprise (Wiley, 2024). He is founding faculty of the MS in Customer Experience Management at Michigan State University's Broad College of Business. He led global customer excellence in Pfizer's first Chief Marketing Organization and in Bayer's Customer Powerhouse. Related reading: What is Realization Rate, and how do you calculate it?, What is value leakage in pharma?, How to measure customer experience in pharma and The commercial system is a link in the therapeutic value chain