Price band analysis splits a category into price ranges and measures each one separately — volume, value, brand composition, growth. It is the standard corrective to the category average, which in most consumer categories describes a price at which almost nothing is actually bought.
Why the average fails#
Consumer categories are rarely distributed around one centre. They are typically multi-peaked: demand clusters at distinct price points, separated by ranges where very little sells. Those clusters correspond to real consumer decisions — an entry purchase, a considered upgrade, a premium choice — and the gaps between them are where products fail to find a buyer.
An average falls somewhere between the peaks. That produces two failures at once:
- It names a price nobody pays. A category with clusters around ¥89 and ¥399 might average near ¥210, where the assortment is thinnest.
- It is stable under changes that matter. Volume can move substantially from a low cluster into a high one while the average barely moves, because other movement offsets it. A metric that does not move when the market restructures is not measuring the market.
Reading the shape of the distribution rather than its centre is the entire point.
Defining the bands#
Bands should come from the category's own distribution, not from an external scale.
- Plot actual selling prices at SKU level. This has to be SKU level — an SPU spanning three sizes has three prices and no single one represents it. See SPU vs SKU.
- Find the clusters. Where does volume concentrate?
- Cut in the sparse ranges between clusters, so each band holds products consumers treat as substitutable.
- Fix the boundaries and hold them. Bands that move between periods make the series uncomparable, and a band boundary shifted mid-analysis can manufacture any trend you like.
Bands imposed from outside — round numbers, a global tier scheme, last year's boundaries applied to a category that has moved — routinely cut through the middle of a cluster and split one competitive set across two bands. The analysis then describes the band definition rather than the market.
List price or selling price#
Selling price, in nearly every case.
In categories with heavy promotional activity the two diverge far enough that a list-price analysis describes a market that does not exist: products nominally positioned in one band transact consistently in another, sometimes permanently. On platforms where promotional pricing is close to continuous, list price is best understood as a marketing artefact rather than a commercial fact.
The exception is when the question is specifically about positioning intent — how a competitor wishes to be perceived, or where they will return to after a promotional period. That is a legitimate question and list price is the right input for it. What is never legitimate is mixing the two inside one analysis.
What the bands reveal#
Where value sits versus where volume sits. These are frequently different bands. A category can take most of its units at the entry point and most of its value two bands up. Which one a brand should care about depends entirely on its economics, and the aggregate figure answers neither question.
Whether a brand's position is broad or concentrated. A brand with a modest overall position may hold a strong one inside a single band. That is a materially different commercial situation from being uniformly modest across the category: the first is a base to expand from, the second is not. Only band-level measurement distinguishes them.
Gaps. Ranges where demand exists but few products are positioned. These are the openings a challenger can enter without meeting an entrenched competitor head-on, and they are invisible in any aggregate view.
Migration. Comparing the same bands across periods shows whether demand is moving up or down the price ladder. This is among the most useful signals in category work, and it is one an average will actively hide.
Where band analysis goes wrong#
- Bands redefined between periods. Any trend can be produced this way. Fix them once.
- Mixing list and selling price. Covered above; it is the most common error.
- Ignoring the platform split. The same product often transacts at different prices across platforms, so a single blended distribution can show a cluster that exists on no individual platform. Where the platform mix differs materially, band the platforms separately — see Tmall vs JD vs Douyin.
- Reading bands without units. Value per band and units per band tell different stories, and a band can grow in value purely on price movement.
- Too many bands. Beyond five or six, each band holds too little volume to be reliable and the analysis stops being readable.
Where to look next#
For the sales value the bands aggregate, see GMV. For why price work must be done at variant level, see SPU vs SKU. For how price position interacts with repeat purchase, see Category Penetration and Repurchase. For platform-level differences in price behaviour, see Tmall vs JD vs Douyin.
Common questions#
Why is a category average price misleading?#
Because most consumer categories are not distributed around a single centre. They are usually multi-peaked, with clusters of demand at distinct price points separated by ranges where very little sells. An average falls between those clusters, so it frequently names a price at which almost nothing is bought. Worse, the average is stable under changes that matter enormously: a category can shift substantial volume from a low cluster to a high one while the average barely moves, because the movement is offset elsewhere. Price band analysis exists because the shape of the distribution, not its centre, is what carries the commercial information.
How should price bands be defined?#
From the category's own distribution rather than from a fixed scale. The practical method is to plot selling prices, find where volume actually clusters, and set band boundaries in the sparse ranges between clusters — so each band contains a group of products consumers treat as substitutable. Bands imposed from outside, such as round numbers or a global tier scheme, will cut through the middle of a cluster and split one competitive set across two bands, which makes the analysis describe the band definition rather than the market. Bands should also be held constant over time, or a series cannot be compared.
Should bands use list price or actual selling price?#
Actual selling price, in nearly every case. In categories with heavy promotional activity the two diverge so far that a list-price analysis describes a market that does not exist — products nominally positioned in one band transact consistently in another. The exception is when the question is specifically about positioning intent rather than market reality, for example assessing how a competitor wants to be perceived. Whichever is used, it must be stated, and the two must never be mixed within one analysis.
What does price band analysis reveal that a share figure does not?#
Where a brand can realistically compete, and what it would cost. A brand with a small overall position may hold a strong one inside a single band, which is a completely different commercial situation from being uniformly small — the first is a defensible base to expand from, the second is not. Band analysis also exposes gaps: ranges where demand exists but few products are positioned, which are the openings a challenger can enter without confronting an entrenched competitor directly. An aggregate share figure shows none of this.