Sign in

Blog · Territory, capacity and quota planning · Pharma

Sample-to-script correlation per territory: what the samples actually moved

How a pharmaceutical commercial team relates sample drops to prescription change per account and per territory, from the sample log and prescription data: the account's own baseline, the change in the weeks after sampling against accounts of the same tier not sampled, the territories where sampling moves nothing, and the honest wording that keeps a correlation from being reported as a cause.

The short answerPer account, compare prescriptions in the eight weeks after a sample drop against the account's own prior eight weeks, and against accounts of the same tier and access status that were not sampled in the same period. The difference between the two differences is the sample-associated change. Rolled up per territory, it shows where samples are followed by prescription change and where they are not, and the allocation moves toward the first. It is a correlation from observational data, and the report says so on every line.

A field force drops samples according to a plan that was set from the last plan. The sample log and the prescription data can say, per territory, whether the drops are followed by any change in prescribing net of what similar unsampled accounts did. This guide sets out the account-level comparison, the territory roll-up, and the wording that keeps the finding honest.

The measures

Per sampled account:

Own change = scripts in the 8 weeks after the drop − scripts in the 8 weeks before Comparison change = the same, for unsampled accounts of the same tier and access status in the same weeks, as a median Sample-associated change = own change − comparison change

Per territory, per tier:

Median sample-associated change, and the share of sampled accounts with a positive one Samples per unit of associated change

The rows you need

  • Sample log: account, rep, date, quantity.
  • Prescription data: account, product, week, scripts.
  • Account master: account, territory, tier, access status.

Account identifiers only.

The assertion

sampled accounts + unsampled accounts = accounts in the tier and access cell, per period

Every account in one group per period. An account sampled in the comparison window is removed from the comparison group for that window.

A worked territory view

Territory Sampled accounts Median own change Median comparison change Sample-associated Share positive Samples per associated script
T-04 61 +3.1 +0.4 +2.7 68% 14
T-11 48 +2.8 +2.6 +0.2 51% 190
T-17 55 +0.9 +1.1 −0.2 44% not computable

Territory T-11's sampled accounts grew and so did everyone else's; the samples are associated with almost nothing. Territory T-04's sampled accounts grew more than the comparison group by a margin that repeats across tiers. The allocation moves from T-11 toward T-04, and the report shows the comparison that justifies it.

By tier within a territory

T-04, tier Sampled Sample-associated change
Tier 1 18 +4.2
Tier 2 27 +2.1
Tier 3 16 +0.3

Within the responsive territory, the response is in the top two tiers. Tier 3 samples are a cost.

Where it goes wrong

Own change only. Every sample looks effective in a growing market.

Reported as cause. The selection problem is real. Associated with, on every line.

Comparison group ignores access. Blocked accounts in the comparison group drag its change down and flatter the samples.

Samples per script read as ROI. It is a ratio for allocation, not a return.

Every cycle, per territory and tier

Mapped once, the sample log, the prescription data and the account master produce the account comparisons, the territory and tier views and the allocation signal every cycle. Covirage builds this from the exports as they are. The pharma page describes the setup, and the formulary status guide covers the access dimension the comparison group depends on.

Questions people ask

Why the unsampled comparison group?

Because prescriptions move for many reasons: season, formulary, a competitor's launch. Accounts of the same tier and access that were not sampled in the same weeks carry those reasons too, and subtracting their change removes most of the noise that would otherwise be credited to the sample.

Can this say samples cause prescriptions?

No. Reps sample the accounts they expect to respond, which is selection. The measure says which territories and account types show a change after sampling, net of the comparison group, and that is enough to allocate samples better. The wording on the report is associated with, never caused.

What data leaves the building?

Account identifiers, sample quantities and dates, prescription counts by period, tier and access status. No prescriber names. Prescription data is already aggregated to the account in most sources, and it stays that way.