Why "average price" is three different numbers
"Average price" is one of the most quoted and least defined figures in retail analysis. It can mean the average of the shelf prices observed, the total value divided by total units, or the total value divided by total volume. Those three can move in opposite directions in the same period, and each supports a different story.
An analysis that does not say which one it is using is not making a claim that can be checked.
What is price per unit?
Total value divided by total units sold. It is the simplest and the most affected by pack mix: a shift toward larger packs raises it without any price changing, and a shift toward singles lowers it.
It answers the question "what did a purchase cost on average," which is a real question and not the same as "what did the product cost."
What is price per volume?
Total value divided by equivalised volume. This one strips pack-mix effects and answers "what did a comparable quantity cost." It is the measure that moves when actual pricing moves.
For pricing analysis it is usually the right number, and it is the one most often missing from a summary because it requires size attributes to compute.
What is a weighted market average?
The average across items or stores, weighted by their sales. It is a market-level statement rather than a transaction-level one, and the weighting choice changes it substantially.
Weighting by sales gives a figure dominated by what sells; weighting by distribution gives a figure dominated by what is available. Neither is wrong and they answer different questions.
Which one moves with a promotion?
All three, by different amounts. A deep promotion on a large pack moves price per unit sharply, price per volume less, and the market average by an amount depending on that item's weight.
If a price analysis and a promotional analysis disagree, this is frequently the reason, and the resolution is to state which measure each is using rather than to re-run either.
How does mix create a false price movement?
The classic case is a category where a cheaper brand grows share. Every individual price is unchanged and the category average price falls, because the mix moved. A report describing that as "prices fell" is describing the market accurately and the cause misleadingly.
Separating mix from rate is a standard decomposition and it is worth asking for explicitly when a price movement is presented as a finding.
What about promotions and the base price?
The observed average price blends promoted and non-promoted sales. A brand can hold its base price perfectly and show a falling average price simply by promoting more often.
That is why base price and promoted price are usually reported separately. A single average price conceals the most commercially important thing about the period.
What should a report state?
The measure, the weighting, whether promoted sales are included, and the period. Four clauses, and their absence is the most common reason a price chart cannot be reconciled with anybody's experience of the market.
Frequently asked questions
Which measure should I use for elasticity?
Price per equivalised volume, generally, because it is the measure that reflects actual price changes rather than pack-mix effects. Elasticity estimated on price per unit will partly measure mix.
Why does my average price differ from the shelf price?
Because the average blends promoted and non-promoted sales, and blends items and stores. Shelf price is an observation about one item at one moment; average price is a summary of everything that sold.
Does tax affect these?
It can, depending on whether the data is captured before or after tax and on what is taxable in a given jurisdiction. This is a definitional question and it belongs in the data dictionary rather than in the analysis.
How do I compare against inflation?
Carefully, and with the measures aligned. Published price indices are constructed to hold quality and mix constant, which is precisely what a raw average price does not do, so the two are not directly comparable.
Is a rising average price good news?
It depends entirely on why. Rising because of a mix shift to premium is a different business outcome from rising because of a list price increase, and a third from rising because promotions were cut.
What is the quickest diagnostic?
Plot price per unit and price per volume together. Where they diverge, pack mix is doing the work, and that is the conversation to have.
Can a category average be meaningless?
It can be, where the category contains genuinely incomparable formats. An average across formats that no shopper would substitute is arithmetic rather than information.
Should the base price be modelled or observed?
Both approaches exist and both need stating. A modelled base price depends on the model's assumptions, and a report should say which was used rather than presenting the output as an observation.
### How should a price movement be investigated?
Start by decomposing it into rate and mix. If rates moved, the next question is whose and where; if mix moved, the next question is which items gained share and why.
That two-step takes minutes and it prevents the most common error in price reporting, which is describing a compositional shift as a pricing event. The measures themselves are standard and the Census Bureau publishes definitions that show how carefully they are usually stated.
Interested in NRS Insights data? See case studies or contact NRS Insights.