Spend per head: the bar metric that tells you more than revenue
By Jack Hawkins · Updated
Spend per head is net revenue divided by covers over a period: £9,400 of wet sales across 1,100 covers is £8.55 each. It's the one number that tells you whether a good week was real, because revenue alone can't distinguish "we got busier" from "we got better", and only one of those two compounds.
Most venues report revenue weekly and covers occasionally, which means the ratio that actually drives the business is computed by accident, usually in the landlord's head, usually late. This is how to make it a habit, what to do with it, and the traps that make it lie.
The three numbers, not one
Spend per head in isolation is a scoreboard. The useful instrument is the triple:
- Average spend per head (SPH): net revenue ÷ covers, weekly.
- Drinks per head: units poured ÷ covers.
- Average unit price: net revenue ÷ units, which is your menu's realised price once discounts, happy hour and the free-pour culture have had their way.
SPH is the product of the other two, and every movement in it has a cause you can find. SPH up 6% with unit price flat means people drank more, which in October means the footfall was genuinely trading, not just arriving. Unit price down with SPH up means your offer is cannibalising itself gently. This is the arithmetic behind the weekly review workspace's top row.
What's normal, honestly
A national spend-per-head figure needs its sample and format definition alongside it, because the number is a function of format: neighbourhood pubs with covers data (from door counters, table bookings or ordering systems, usually the ordering system) commonly run £6-£10 net; cocktail venues with a list to the £13-£15 tier run £13-£18; hotel bars trade higher per head on lower volume. These are trade-estimate bands, not statistics.
The number worth obsessing over is your own trend, in your own format, same-weekday compared. A venue at £8.40 in June and £7.90 in October has lower spend per guest; total revenue depends on whether cover volume grew enough to offset that fall, and the two components will tell you which half of the business changed its mind.
Covers: the dirty prerequisite
Spend per head is only as honest as your cover count, and most bars don't actually count covers. The three working proxies, in rough order of reliability: a door counter or booking system; table-turn data from the EPOS if you run sections; and the fallback, tables seated × average party size, which inherits every lie table sizes tell. Whatever you pick, pick it forever and write it down. A changing cover definition turns SPH trends into noise with a decimal point, and the first thing staff will (rightly) say about a new metric is "that's not a real number of people".
Four moves that actually raise it
The second-drink window. Slow service can limit the number of orders a room completes, but it does not translate into a fixed lost-round count. Track order-to-delivery time alongside drinks per guest and shift mix. Compare like periods before deciding whether more staff, a shorter build or a different ordering flow would produce the best return.
Ladder design, not pushiness. A £11.50 list with three drinks at £14 and a visible premium tier earns unit price without anyone being upsold like a car warranty. A flat £11.50 list has nowhere to go up. Our menu contribution matrix is what this ladder should be ranked on, because you want the premium tier to be premium and profitable, not premium and apologetic.
Food that fits. Even a two-item snack menu moves SPH without touching the drinks line, and food may change dwell, spend and demand for drinks; measure the effect at your venue. The margin question belongs on the break-even page; check total contribution as well as any increase in spend per head.
Fewer, sharper discounts. Every happy-hour hour is an SPH haircut at the exact hours when the bar can least afford one if the room would have filled anyway. The full arithmetic of when discounts pay is in happy hour ROI; the summary is that venues that track SPH during offers tend to run fewer of them.
Choose a cover definition that survives a busy night
Count guests rather than transactions: one person can pay twice, and one transaction can cover six people. Door counts include non-buyers unless adjusted; bookings miss walk-ins and may include no-shows. Put the limitation beside the metric. If the numerator includes food, use total spend per head and total items per head, or split food and wet revenue first. A drinks-only average price multiplied by drinks per guest reproduces wet spend per head, not the whole bill. Where covers are unreliable, report revenue per transaction under its own name rather than presenting an estimated guest count as exact.
FAQ
What is spend per head in a bar? Net revenue divided by covers over the same period, usually weekly and ex-VAT. It splits cleanly into drinks per head and average unit price, which is why it's more diagnostic than revenue.
How do I calculate spend per head without a door counter? Use booking/table data, or covers = tables seated × average party size from reservation notes. Write the method down and keep it consistent; label the approximation and keep its method stable so the trend can be interpreted.
What's a good spend per head for a UK bar? Format-dependent and honestly unbenchmarkable nationally: trade-estimate bands run £6-£10 net for neighbourhood pubs and £13-£18 for cocktail venues. Your own same-weekday trend is the number that means something.
Why is spend per head dropping when we're busy? Usually the unit price leg: more discounts, a cheaper category mix (spritz weather), or a new cohort ordering differently. Split SPH into drinks-per-head × unit price; the half that moved names the problem.
Method and assumptions
Band ranges are trade estimates labelled indicative, not audited statistics; no single national SPH figure can be transferred to every format without checking its cover definition and sample. The component decomposition (SPH = drinks/head × unit price) is arithmetic. Method standards: data standards.
Last reviewed 19 October 2026.