Product mix (PMIX): read your bar's best and worst sellers in ten minutes
By Jack Hawkins · Updated
A product mix report is the ranked table of every item you sell against units, revenue and margin: the single export that tells you what your bar actually is, as opposed to what the menu launch deck said it would be. Bartenders argue from memory ("everyone has an espresso martini now"), owners argue from the top of the bestseller list, and the PMIX quietly shows both of them a more interesting story, because the ranking can change the moment contribution joins the table.
PMIX is commonly used as shorthand for product mix in hospitality reporting; in bar software it usually appears as "product mix", "PMIX" or "item performance". Whatever your POS calls it, the instrument is the same, and reading it properly takes one afternoon of setup and ten minutes a week forever.
The columns that matter, in order
A useful bar PMIX has exactly these, and you can build it from any item-level CSV export plus your costed specs:
- Units (period-fixed: 28 days, same window every time)
- Net revenue (ex-VAT — check your POS reports gross by default; confirm the exact field definition)
- Cost per unit from the spec, and contribution = net price − cost
- Total contribution = contribution × units, the column the list gets sorted on
- % of total units and % of total contribution, which is the two-column version of the whole business: most lists show a drink at 12% of units and 4% of contribution, and that sentence is the meeting.
Two reads do almost all the work. First, sort by total contribution and watch what leaves the top ten between months: those exits are your market shifting, faster than category trend pieces will tell you. Second, find the items with high units-share and low contribution-share, a group to investigate; confirm popularity and contribution per serve before calling an item a plowhorse, and their gentle repricing is the cheapest money in hospitality. The contribution matrix tool does the four-box sort from exactly these columns.
The three PMIX lies to check before believing it
The gross-price lie. If revenue isn't ex-VAT, the margin calculation uses the wrong revenue basis, with an error that varies by recipe cost and price and the "who's underwater" flag is lying. Same VAT problem the pour cost guide covers, wearing a report's clothes.
The combo lie. "Cocktails" as one line item is a category, not a product; a PMIX with categories in the product rows hides everything interesting. Item-level, every time; category rollups are the summary line, not the analysis.
The modifier lie. Upcharges (premium spirit, extra shot, zero-ABV swap) often report as separate $-less line items or get buried, and in an illustrative list where 15% of serves take an upcharge, that's a real revenue line going unattributed. Check where your POS parks them before deciding a drink's price was respected.
Ten-minute weekly ritual
Sort by total contribution, contribution, and units, once each, three screenshots, five minutes: the only reports where order is the analysis. Anything that moved more than five rank places since last month goes on the weekly review agenda; that's the whole ritual. Quarterly, the same three screenshots at 90-day windows tell the structural story, the tequila-vs-gin kind, which is a menu-planning input rather than a this-week decision.
The reason PMIX beats every fancier report for a solo operator is that it's the one artifact where money, popularity and cost share a table. Dashboards aggregate; PMIX ranks; and venue decisions are ranking decisions: what to promote, reprice, move, reorder, kill.
A small example of the quadrant rule
Suppose a ten-item category sells 1,000 serves. The popularity hurdle is 70 serves, while the contribution threshold comes from the sales-weighted average ingredient contribution per serve. An item selling 100 serves at £3 contribution is popular but below a £6 average: a plowhorse. An item selling 20 at £9 is unpopular and above that line: a puzzle. Their total contributions, £300 and £180, answer a different question: how much money each delivered in the period. Keep all three columns visible. Comparing shares of total contribution alone can obscure the distinction between popularity and cash earned per serve.
FAQ
What is a product mix report in a bar? An item-level table of units sold, net revenue, cost and margin for a fixed period, usually built from POS data. It ranks what actually sells against what actually earns, which are different lists in every venue.
How do I read PMIX? Sort by total contribution first. Compare each item's share of units against its share of contribution: high-popularity items below weighted-average contribution per serve are plowhorses; low-popularity items above that per-serve line are puzzles. Total contribution share is useful context, but is not the classic quadrant threshold. Watch rank changes month over month for market drift.
What's the difference between PMIX and menu engineering? PMIX is the report; menu engineering is the action framework built on it (the star/plowhorse/puzzle/dog boxes). Same data, different verbs: one describes, the other decides. The engineering guide has the full method.
Can I do this in a spreadsheet? Yes, easily: item-level POS CSV plus costed specs, a pivot, and four computed columns. Twenty minutes to build, using a repeatable import and matching process. Above 200 items or multiple sites, automation starts to pay for itself.
Method and assumptions
Conventions follow Kasavana-lineage mix analysis with UK VAT corrections (net = menu ÷ 1.20, reviewed October 2026); rank-change thresholds (five places) are working defaults, not findings. No venue performance claims are made.
Last reviewed 19 October 2026.