NRS Insights Blog · Topic

Measurement & Testing

Lift, incrementality, experiments and the statistics that decide whether a result is real.

All topics

Measurement & Testing

Newest first

Measurement & Testing
Choosing test stores for a pilot

Test market selection for a retail pilot: how to choose test stores that represent the rollout, match them to controls, and avoid common traps.

Sep 11, 20264 min read
Measurement & Testing
Confidence intervals: reading a range instead of a point

Confidence intervals give a range of plausible values instead of one number. How to read one in retail data and why the width matters.

Aug 17, 20264 min read
Measurement & Testing
Statistical significance for brand managers

Statistical significance for brand managers: what a p-value says, what it doesn't, and why a significant result isn't always an important one.

Aug 15, 20264 min read
Measurement & Testing
Product cannibalization: when a new item takes your own sales

Product cannibalization happens when a new item takes sales from your existing line. How to measure it in scan data and decide if it's acceptable.

Jul 3, 20264 min read
Measurement & Testing
Incrementality: measuring sales that wouldn't have happened

Incrementality measures the sales a promotion or campaign actually added. Here's how to set a baseline, pick control stores and read results.

Jul 1, 20264 min read
Measurement & Testing
Reading a regional rollout

A rollout produces data from markets at different stages. Reading it as one trend confuses stage with performance.

May 16, 20264 min read
Measurement & Testing
Forecast error: measuring it honestly

Every forecast is wrong; the question is how wrong and in which direction. The error measures worth using and the ones that flatter.

Apr 25, 20264 min read
Measurement & Testing
Significance, confidence, and what a p-value does not say

Significance testing answers a narrow question. What it means, what it does not, and how to present uncertainty in a commercial setting.

Apr 24, 20265 min read
Measurement & Testing
Test and control design for retail

Store-level tests are feasible in independent retail and usually badly designed. The decisions that determine whether a test can answer anything.

Apr 22, 20265 min read
Measurement & Testing
Measuring incrementality without a holdout

A proper holdout is the cleanest test and it is often impossible. What the alternatives are and how much confidence each earns.

Apr 20, 20265 min read
Measurement & Testing
Cohort analysis on transaction data

Cohorts group by when something started, not by what it is. How the idea applies to stores and items without shopper identifiers.

Apr 18, 20265 min read
Measurement & Testing
Repeat rate and the repeat window

Repeat purchase is the clearest signal a new product works. Measuring it without shopper identifiers, and choosing the window honestly.

Apr 17, 20265 min read
Measurement & Testing
New item ramp curves: what "normal" looks like

A launch's early weeks blend distribution build with trial. Reading a ramp correctly means separating the two before judging it.

Apr 6, 20265 min read
Measurement & Testing
Halo and pantry effects around a promotion

A promotion moves more than the promoted item. Halo, pantry loading and the post-period dip all belong in the measurement window.

Mar 6, 20265 min read
Measurement & Testing
Cannibalisation: when a launch moves its own volume

A new item that sells is not necessarily a new item that added anything. How cannibalisation is measured and why launches overstate.

Mar 4, 20265 min read
Measurement & Testing
Base and incremental volume: separating the two

A promoted week's sales include what would have sold anyway. Separating base from incremental is what makes a promotion measurable.

Mar 2, 20265 min read

Measurement & Testing

All topics

Measurement & Testing

Choosing test stores for a pilot

Confidence intervals: reading a range instead of a point

Statistical significance for brand managers

Product cannibalization: when a new item takes your own sales

Incrementality: measuring sales that wouldn't have happened

Reading a regional rollout

Forecast error: measuring it honestly

Significance, confidence, and what a p-value does not say

Test and control design for retail

Measuring incrementality without a holdout

Cohort analysis on transaction data

Repeat rate and the repeat window

New item ramp curves: what "normal" looks like

Halo and pantry effects around a promotion

Cannibalisation: when a launch moves its own volume

Base and incremental volume: separating the two