Nov 01 2026
/
Analog Selection in Pharma Forecasting: Getting It Right and Getting It Wrong
Every pharma forecast eventually leans on a comparison. Somewhere in the model there is a line that says, in effect, “our product will behave roughly like that one.” That product is the analog, and the choice of it can move the final number as much as any other assumption in the model.
Yet in our experience, analog selection is often the least documented part of a forecast. The epidemiology gets pages of sources. The pricing gets a scenario table. The analog gets a sentence on a slide: “Based on launch curves of comparable products.” Comparable in what way? Nobody asks until the sales come in at half of what was predicted.
Why We Need Analogs at All
A new drug has no sales history. Before launch, the only way to estimate uptake, peak share, or time to peak is to look at what happened to products that faced a similar situation. Analogs give a forecast something to stand on that isn’t pure opinion.
They are useful in four places in particular:
- Uptake curve: how fast a product climbs and when it flattens
- Peak market share: where the product settles against its competitors
- Persistence and compliance: how long patients stay on therapy and how consistently they take it
- Time to reimbursement: how long it takes for access to open up
A good analog set informs all of these. A poor one quietly distorts all of them.
Know more about our Market Forecasting Consulting Service
The Usual Mistake: Matching on What is Easy to Match
The most common way to pick an analog is to match on a therapeutic area, which is quick: oncology drug, oncology analog; rare disease drug, rare disease analog.
The trouble is that the therapeutic area says very little about how a launch will go. Two oncology drugs can have almost nothing in common commercially. One may be a fifth-line option in a small, heavily specialized population. The other may be a first-line therapy prescribed in community practices across the country. Putting them in the same set because both are “oncology” is a bit like comparing two restaurants because both serve food.
What actually drives uptake is a mix of factors that are less tidy to classify:
- Unmet need and standard of care. Is the product entering an empty space, or replacing something physicians already trust?
- Clinical profile versus the incumbent. A modest efficacy gain with a worse tolerability story behaves very differently from a clear step change
- Route and burden of administration. An infusion needing a specialist center grows differently from a daily oral tablet
- Population size and prescriber concentration: Reaching 300 physicians is a different task from reaching 30,000
- Payer environment at launch: Prior authorization, step edits, and reimbursement timelines can slow even a very good drug
- Competitive context: Launching alone, or alongside two other products, changes the shape of the curve
Picking analogs by indication alone captures almost none of these.
Start by Defining What You Are Trying to Predict
Be clear about what the analog is for. For the shape of the uptake curve, you care most about how the market adopts new products: prescriber behavior, referral patterns, diagnosis rates. For peak share, you care about competitive position and clinical differentiation. For persistence and compliance, you care about the route of administration, tolerability, and patient support. For time to reimbursement, you care about pricing and the payer environment.
One analog rarely serves all four purposes. Many forecasts improve simply by using different analog sets for different assumptions.
Building the Set
First, list every product that entered a market with a roughly similar structure. At this stage you want options, not precision.
Second, score each candidate against the factors that matter for your product. Be honest about where the match is weak. A product that matches on population but not competitive context should be flagged, not quietly accepted.
Third, look at the spread, not just the average. Say, for illustration, that your analogs produced peak shares between 9% and 34%. That range is itself a finding. Averaging it into a tidy 20% hides the uncertainty the team needs to understand.
Finally, write down why each analog was chosen and why others were left out. This takes an hour and saves a great deal of argument later, especially when a stakeholder asks why a well-known launch isn’t in the set.
When There are Only a Few Analogs
In rare diseases, the problem is often the opposite: too few comparable launches, not too many. A handful of orphan products may be all that exists, and for a first-in-class gene therapy there may be none. Borrow carefully: look at adjacent diseases with similar diagnosis pathways and prescriber concentration, or products with a similar modality, and be explicit about which factors transfer. A thin set makes the written rationale even more important.
Where Forecasts go Wrong
Choosing analogs that tell the story you want: A team is excited about a product, and the success stories come to mind first. The slower launches get forgotten. Ask what the set would look like if a skeptic had built it.
Ignoring survivorship: Products still on the market are easy to find. Those that were withdrawn or never found their footing are harder to track down, and they carry some of the most useful lessons. Take Pfizer’s inhaled insulin, Exubera. It was widely expected to become a blockbuster, with projections of roughly $2 billion a year by 2010. Reported worldwide sales in the first nine months of 2007 were about $12 million. In October 2007, Pfizer announced it would exit the product and recorded $2.8 billion in pre-tax charges in the third quarter. Its filing cited a lack of acceptance by patients, physicians, and payers. Exubera is a reminder that a launch can fail on acceptance rather than on efficacy, and a set made up only of successes will never show you that. It will push every forecast upward.
Treating old launches as current ones: The market a drug launched into ten or fifteen years ago may look nothing like today’s. Payer controls have tightened and specialty pharmacy has grown. An uptake curve from an older launch may be optimistic or simply wrong today.
Copying the curve without understanding it: It is tempting to overlay an analog’s year-by-year sales on your product. But the curve is an outcome. What produced it was a particular set of circumstances: a specific label, a specific competitor exit, perhaps a guideline update in year two. If you don’t know what caused the shape, you can’t judge whether it will repeat.
Underestimating access: PCSK9 inhibitors are the standard example. Repatha and Praluent launched in 2015 at list prices of more than $14,000 a year, with forecasts of multibillion-dollar annual sales. Payers responded with strict eligibility rules, and Repatha’s 2018 sales were around $550 million. In October 2018, Amgen cut Repatha’s US list price by 60%, to $5,850. That cut largely reset the list price to the net price Amgen was already receiving after rebates, and it was aimed mainly at patients’ out-of-pocket costs, especially in Medicare. Praluent had been repriced earlier in 2018 through a deal with Express Scripts. The clinical case was strong. The forecasts had not given enough weight to access.
Getting it Right: Habits Worth Building
Use several analogs and weigh them: In our experience, three to six well-chosen ones serve better than one “perfect” match, which rarely exists.
Adjust, don’t just copy: If your product has a convenience advantage the analog lacked, say so and make an explicit adjustment. If the analog launched into an easier payer landscape, adjust downward. The adjustments are where the real thinking happens, and they should be visible to anyone reading the model. Access also differs by market: reimbursement under NICE, G-BA, or HAS is a different problem from US prior authorization.
Talk to people who were there: Commercial and market access teams who worked on the analog launches know what the numbers don’t show. A thirty-minute conversation about why a launch stalled in year two can be worth more than another round of data pulls.
Learn about our Market Access and Reimbursement Insights Reports
Revisit the choice as new data arrives: Phase 3 results, label negotiations, and competitor news can all change which products make a fair comparison. A set chosen at Phase 2 may not be the right one at filing.
A Short Test for any Analog
Before accepting an analog, ask four questions. Would a physician treating this condition see the two products as similar choices? Would a payer treat them the same way? Were market conditions at launch broadly comparable? And do we understand why the analog performed the way it did? If the answer to any of these is no, the analog needs either an explicit adjustment or a place in the discard pile.
The Takeaway
Analogs are an argument, and like any argument, they should be open to challenge. The best forecasts come from teams that can explain plainly why they believe their product will behave the way it will, and what would change their mind. The cleverness of the model matters less.
Thelansis supports forecasting teams with analog selection and launch curve analysis. Get in touch to discuss your next forecast.