How do you know if a launch is working?

Every number here comes from a synthetic dataset modeled on a real small-biotech workflow: one fictional injectable, 60 territories, calls and patient data from October 1, 2023, distributor sales from October 2, 2023, and 3,335 made-up patients, 1,612 of them already in the program when the data begins. Every feed is current through October 12, 2026, except one distributor's weekly file, which runs through October 9. There are no real patients, prescribers or accounts in it.

Before using an example to teach launch metrics, check that it looks like a launch. This one doesn't.

Sales start at full rate: 6,922 net units in the first month, a little above the recent average of 6,105. Fills start at full rate too, 1,634 in the first month against 1,543 lately, because 90% or more of the patients filling in those months were already on therapy when the data begins. Referrals are level from the start: 12 in the data's first full week, and 11 a week on average over the three years. The one column that falls, new prescribers, falls because the data starts: in its first months nearly every prescriber is new to the data.

So the example brand behaves like an established one from its first day of data. It doesn't model a launch curve, and nothing below is a launch result. The example shows how each metric is computed and read.

The first weeks: referrals, new prescribers and open cases

Count referrals in closed weeks, and don't call a trend on one week. In the last 13 closed weeks the example's referrals ran from 5 to 17 a week, around an average of 11, in a brand whose monthly referrals have been level for three years. Read a four-week average instead.

Count a prescriber as new on their first referral. In a launch every prescriber starts new, so track breadth (new prescribers) next to depth (prescribers who refer again). In the example's latest quarter, 89 of the 157 prescribers who referred were referring for the first time in the data.

Watch open cases from the first referral: count them by step, and flag the ones waiting longer than that step usually takes. On October 12 the example had 31 open cases, 17 of them waiting on a PA decision and 2 on appeal. How long does it take a patient to start therapy? goes through the steps one by one.

One to three months: compare referral cohorts at the same age

Time to therapy and conversion mean something only once a cohort's cases have had time to finish. Here are two cohorts side by side: the latest closed quarter, and the same quarter a year earlier.

Shipped to date says the latest quarter converts 65% of referrals against 75% a year earlier, a 10-point drop. It isn't one. The earlier cohort has had more than a year to finish, and some of the latest quarter's patients were referred 12 days ago. Read both at the same age and the gap shrinks to two or three points: 57% and 55% shipped by day 30, 71% and 68% by day 60.

Day 60 is a fair checkpoint here because 94% of the example's patients who ship do so within 60 days of referral (1,072 of the 1,141 referred by June 30, 2026 who have shipped). Take your own checkpoint from your own time-to-therapy curve. In the latest quarter only the 90 patients referred by August 13 have reached day 60; the rest join the count as they do.

The median has the same trap. Early in a cohort, only its fastest patients have shipped, so its median starts low. The latest quarter's 22 days isn't final until its 17 open cases close.

Three months and beyond: refills, persistence, demand and reach

These numbers need patients who have been on therapy for months, and each needs a rule about who is old enough to count.

Refills on time: a refill is on time when it ships within the previous fill's days of supply plus a 7-day grace. Count refills only, not first fills, so the rate means nothing until patients reach their second fill. Leave out a refill whose previous fill came before your data begins: there is nothing to time it against.

Persistence: count a patient only after a full six months: here, patients whose first shipment is at least 182 days old, with no discontinuation in those days. A patient who started last month hasn't had the chance to stop, and counting them flatters the rate.

Reach: count Tier 1 and 2 HCPs (healthcare professionals) with at least one call in the window, over every Tier 1 and 2 HCP. Count people, not calls. The example's field reached 96% of them in its first 60 days of data, another sign that it doesn't model a launch.

Demand: read sales by closed week or closed month. On October 12 the open week, ending October 16, holds three days: a weekend and a Monday federal holiday. One distributor's lines for those days arrive only with its weekly file the following Monday. The open week reads 59 net units against 1,372 in the last closed week. That is 96% lower, and it isn't a drop. Which time period should each sales report use? explains why. When the first goals arrive, set sales goals by territory and call goals by rep, measure pace against the goal prorated by weekdays to the last day of data, and never hold an open week or month to its full goal.