How to Manage a Café Queue at Peak Hours

At a peak, the lever a café actually controls is the order things come out and who gets served first. Queue length is an output, not a setting. You can change the sequence of work behind the counter today, without hiring anyone, and that changes who stays.

Most advice goes the other way. Trim the menu, add a till, push people towards an app. Those are capacity decisions, and capacity is expensive. Sequence is free, and there is peer-reviewed work behind it.

The other half of the answer is duller and comes first. Before any of that is worth changing, you need to know when your own rush happens. The published UK figures will not tell you, and the reason is worth a section on its own.

Key Takeaways

• The morning is not automatically the peak. One UK publisher puts it at 32% of coffee shop visits (Meaningful Vision, 2025), on a page stating no sample and no method.
• Nothing published will tell you when your own peak falls. Two UK publishers report morning shares on windows neither defines, so the numbers cannot be compared.
• Sequence beats pace. A queue clearing a larger share of its length early, then slowing, holds people better than a steady one over the same total wait (Production and Operations Management, 2024). Laboratory and online experiments, not a real counter.
• Mobile ordering is a trade-off. Modelling motivated by coffee shops names settings where adding a self-ordering channel reduces throughput (Production and Operations Management, 2024).
• App ordering has reached the British high street. The Greggs app was scanned in 26.7% of company-managed shop transactions in the 52 weeks to 27 December 2025 (James Sharp, 2026).
• The customers who looked and left are missing from your data by design. A manager sees the queue, not the people who never joined it (arXiv, 2026).

Your café’s peak hour may not be the morning

Two UK data publishers now report that mornings carry less of a café’s trade than most operators expect. In a 2025 report, Meaningful Vision puts the morning at 32% of coffee shop visits, and 34% for bakeries (Meaningful Vision, 2025).

The rest of the day is bigger than that. Lunch, which the same page defines as 12pm to 3pm, runs at 36% of coffee shop visits, and the same share for bakeries (Meaningful Vision, 2025). Between 3pm and 6pm, coffee shops take 28% of visits and bakeries 26% (Meaningful Vision, 2025).

A second page from the same publisher points the same way. It calls 3pm to 6pm the fastest-growing daypart, ahead of lunch (Meaningful Vision, 2025).

Now the caution, and it is a big one. The afternoons page states no sample, no method and no data period. The words sample, methodology and fieldwork do not appear on it anywhere. So the shares are worth knowing about and not worth building a rota around.

A second publisher measures something else again. In January 2025, World Coffee Portal reported that early morning coffee shop activity rose from 22% to 24% in the UK (World Coffee Portal, 2025), on rising commuter footfall.

One is a share of visits, the other a measure of activity, so the two cannot be compared. Neither publisher defines what it counts as morning, and the second covers branded chains rather than every café.

Which leaves an operator where they started, with a question the sector data cannot answer. There is no substitute for finding when your own peak actually falls, on your own till records, across a fortnight that takes in the days you dread.

The constraint is a station, not the queue

Throughput at a peak is set by whichever station the work piles up behind. In most cafés that is one thing: the espresso machine, the steam wand, the till, or the spot where finished drinks land. The queue is what that station leaves behind it.

No in-date, measured figure for how long a UK café transaction takes was found while researching this piece. The drinks-per-hour numbers in circulation come from firms selling espresso machines, and they contradict one another on the same kit.

That absence is worth naming, because it makes a common café sum misleading. Take an invented saving per drink, multiply it by a day’s volume, and a large block of recovered time appears. The sum adds up and it means nothing on the floor. A café at a peak runs several people at once, so seconds trimmed at one station do not pool into a spare half hour anywhere.

Staffing works the same way. Two more people on shift change nothing if a single steam wand is the limit.

One thing a rota can change is not about headcount at all. A 2024 paper compared dedicated queues with pooled ones across two online experiments in a healthcare setting. In those experiments dedicated queues were faster “without sacrificing quality” (M&SOM, 2024).

Part of that gap ran through staff judging their own load more accurately. Part ran the other way, because owning a queue distracts from the customer being served. The participants were not baristas, and what travels is a staff member’s read of their own load, not a board showing it.

Which shape the queue itself should take is settled elsewhere, and this piece treats it as settled: how queue shape changes who waits longest works through the evidence. The narrower point here is that whoever is behind the counter should be able to see what they are carrying.

One buying consequence follows. Size a machine against your busiest stretch rather than the daily total, because the quiet hours either side are not the ones you are short of.

Does the order you serve in change how long people wait?

It changes how many of them stay, which is a different result and a more useful one. A 2024 paper compared queues with identical total waits (Production and Operations Management, 2024). A queue that cleared a larger proportion of its original length early, then slowed, held people better than one moving at a steady rate.

The authors call it a relative progress effect, and the wording repays care. The effect is about the share of the queue already cleared, not about the queue being short. Five people dropping to two quickly and then stalling is a different experience from five people leaving one at a time, at the same overall pace.

The scope is narrow. These were laboratory and online experiments rather than observations of a real counter. What it offers a café is a hypothesis worth a Saturday: clear the front of the queue fast, rather than pacing the rush evenly.

One large chain has published an operating model built on the same instinct. Starbucks has set out a sequence with baristas “pulling espresso before steaming the milk when creating a beverage, instead of the other way around” (Nation’s Restaurant News, 2024).

The same announcement described a “Peak Play Caller” (Nation’s Restaurant News, 2024). The role is “a flexible position that acts as a ‘floater’ to help baristas where and when they need it during peak times”. It said the system would scale across the United States and Canada by the end of July 2024 (Nation’s Restaurant News, 2024). That was a stated plan, not a reported outcome.

A separate Starbucks disclosure attaches a number to a handoff target. In July 2025 the company reported “a double-digit improvement in cafe orders handed off in under four minutes” in stores running SmartQ, its order-sequencing system (The Motley Fool, 2025). Those stores were eight weeks into a 1,500-store test, and 80% of their in-cafe orders were then meeting the four-minute target (The Motley Fool, 2025). The call does not say which countries that test covered, beyond describing a North American portfolio.

That is a statement to investors, not an independent measurement. It is quoted here for what a large operator aims at, not as a target a twelve-seat café should adopt, and not as a result anyone should expect.

The claim	Where it comes from	What it does not establish
A queue clearing a larger share of its length early holds people better than a steady one	Laboratory and online experiments (Production and Operations Management, 2024)	Anything about a real café counter. The abstract reports no figures
Espresso pulled before the milk is steamed, plus a floating member of staff at peaks	A published operating model, United States and Canada (Nation’s Restaurant News, 2024)	Any throughput or wait outcome. None was published with it
A four-minute handoff target, met by 80% of in-cafe orders	One chain’s stores running SmartQ, inside a 1,500-store test whose countries the call does not state (The Motley Fool, 2025)	What a small independent café can expect. It is a statement to investors

None of this makes the wait shorter. It changes what the wait is like, which is the ground covered by the levers that change a wait without shortening it.

Mobile orders move the queue rather than removing it

Adding a mobile or self-ordering channel can reduce throughput rather than raise it. A 2024 paper on omnichannel services, motivated by fast-casual restaurants and coffee shops, names settings where that switch reduces it (Production and Operations Management, 2024).

Can, not does. The result turns on priority: who goes first when a counter customer and an app order arrive together. Serve app orders first and the counter queue absorbs the delay. Serve the counter first and the app’s promise stops meaning much.

That is analytical queueing work rather than field data, and its value here is structural. Nearly every café page recommending mobile ordering treats it as a clear gain. The gain depends on a choice most venues make by default.

There is a fairness cost too, and it is more specific than customers get annoyed. A 2025 experimental study compared two ways of skipping a queue (M&SOM, 2025). Customers were more satisfied when somebody took over an existing position than when an express lane inserted one at the front, and the insertion read as less fair.

An app order collected ahead of the queue is an insertion, not a swap. Nobody standing there gave up a place for it. That is not an argument against mobile ordering. It is an argument for making the collection point visibly separate from the queue.

How far has app ordering reached a British high-street queue? Greggs reported that its app was scanned in 26.7% of company-managed shop transactions in the 52 weeks to 27 December 2025 (James Sharp, 2026). The year before, the figure was 20.1% (James Sharp, 2026).

That is a food-to-go chain rather than an independent café. Still, a quarter of transactions arriving pre-paid changes what a counter is for. Ordering and collecting stop being one act.

What goes wrong there is already well covered in the handoff at a collection counter, which works through the four ways of telling somebody their order is up.

One tool for that point is worth knowing about, limits first. Channels are one-way and messages are capped at 40 characters. So the channel carries nothing back, and a customer who wants to change an order has to come to the counter.

Inside those limits, Upd8All works like this. Staff show a QR code at the till, the customer scans it, and short updates follow on the phone they already have. The customer does not need the app: scanning the QR code opens the updates in their browser through Upd8 Go. There are no accounts and no sign-up, for either side.

So the fit is narrow, and what Upd8All is and where it does not fit draws the boundary. It suits a paid-for order and a customer waiting nearby. It suits nothing that needs a reply.

The advice nobody has tested

Almost every guide to café rushes recommends moving demand out of the peak. A discount before 8am, a loyalty push at 3pm, an offer that only runs when the shop is quiet. No current published test was found for this article showing that any of it measurably shifts café load out of a peak.

That is a statement about the search behind this article, not about the world. Demand-side pricing is tested in other sectors, transport among them. What is missing is a café. Nothing found puts a number on how many of the 8:15 queue would have come at 7:45 for a discount.

There is a second reason for caution, and it is arithmetic rather than evidence. Shifting a customer from 8:15 to 7:45 only helps if 7:45 has room at the constraint. Otherwise the promotion has moved the queue rather than shortened it.

None of which makes off-peak promotions a bad idea. It makes them an untested one, which is not the same as a proven one, and worth knowing before the poster order goes in.

The customers you never see, and the number quoted about them

The people who looked at your queue and kept walking are absent from every record you hold. A queueing preprint on estimating service value from queue data, revised in March 2026, puts it in one line (arXiv, 2026):

“The manager observes the queue length process, but not the balking customers.”

That paper is theory and simulation rather than field data, and it is candid about the price of closing the gap. Accurate estimation, its authors write, “may require 10^7 or even 10^8 observations, depending strongly on the specific parameter configuration” (arXiv, 2026). That requirement belongs to one particular model setting rather than to queues in general. Either way, a café serving a few hundred people a day is not going to recover its lost trade from its own till data.

Which leaves one instrument: somebody with a view of the door. How far that loss can run, and what the published walkaway rates measure, is set out in the evidence on when a queue loses people.

There is a number the coffee trade quotes about exactly this, and it repays following back. In March 2026 a coffee trade publication reported (Perfect Daily Grind, 2026):

“Research from University College London found that the average customer is prepared to wait in line for 5 minutes and 54 seconds before considering leaving.”

The same figure has two earlier lives, and neither is about coffee. Nine years earlier, a café consultancy page credited the same 5 minutes 54 seconds to a professor of psychology at UCL, citing no study (Cafe Success Hub, 2017).

Three years before that, a European retail trade site had reported the same duration from a survey of 1,344 UK consumers, run by a retail technology provider (EuroShop, 2014).

Where it appeared	Date	What that page claimed
EuroShop, reporting a survey by a retail technology provider	28 April 2014	5 minutes 54 seconds as the average queue time UK shoppers tolerated before abandoning a basket, from research of 1,344 UK consumers
Cafe Success Hub, a café consultancy	First published 2017	The same 5 minutes 54 seconds, plus satisfaction falling from 95% to 85% after five minutes, attributed to a professor of psychology at UCL. No study, paper or dataset cited
Perfect Daily Grind, coffee trade press	19 March 2026	The same figure and the same 95% to 85% fall, attributed to research from University College London, in an article about coffee shops

Two things can be said about that chain, and one cannot. The same figure, to the second, appears on all three pages. Only the 2014 one says what was measured, or how many people were asked. No page in the chain cites a study, so the link between them cannot be shown, only noted. Whether the professor named in 2017 was involved cannot be settled from any of the three pages.

A basket abandoned in a shop queue and a person leaving a café queue to catch a train are not the same behaviour. A threshold from one has no claim on the other. The figure was twelve years old by the time the coffee trade printed it as university research.

Once you have watched one of your own peaks

Once the watching is done, the order of these levers tends to settle itself. The queue is the last thing to look at, because it is the output. The station where work piles up comes first, because everything else sits downstream of it.

Sequence comes next, since it costs nothing and can change on the shift after the one you watched. Mobile ordering belongs last, not because it is bad but because it is the only lever here with a published trade-off attached. Where the queue lands depends on a priority decision somebody has to take on purpose. Left to accident, the counter queue quietly pays for the app queue.

And the people who never joined the line stay invisible whatever you change, so the count only exists if somebody wrote it down at the time.

Frequently Asked Questions

When is a coffee shop busiest?

Not necessarily first thing, and no published figure will settle it for your shop. One UK publisher puts the morning at 32% of coffee shop visits and the 12pm to 3pm window at 36% (Meaningful Vision, 2025). The page behind those shares does not say who was counted or how, so your own till data is the better guide.

Should mobile orders be served first?

Treat it as a decision, not a default. Modelling motivated by coffee shops finds that adding a self-ordering channel can reduce throughput, and that which class goes first drives the outcome (Production and Operations Management, 2024). Serving app orders ahead of the counter also reads as an insertion rather than a swap, and an insertion is the arrangement people rate as least fair (M&SOM, 2025).

How many staff do you need for a café rush?

Nothing found for this article puts a defensible number on it. Extra hours only help if the extra people can work at the station where the queue is forming. If one machine is the limit, a third person on the till moves the bottleneck rather than clearing it. Count against the constraint, not the queue.

How long will people queue for a coffee?

There is no in-date measurement for cafés. The figure quoted most often, 5 minutes 54 seconds, traces back to a 2014 survey of 1,344 UK consumers about abandoning a basket in a shop queue (EuroShop, 2014). Later pages restated it as university research about coffee shops. Measured on shop queues twelve years ago, it is not a café threshold.

Do off-peak discounts reduce the morning rush?

No current published test was found showing that off-peak promotions shift café load out of a peak. Your own till records are the one place an answer exists for your shop. Run the offer for a fortnight, then compare those hours against the same weekdays without it. Watch the total, not just the discounted sales, because a promotion that only moves existing customers earlier has not added trade.

Sources

• arXiv, Podorojnyi and Ravner, “Estimation of service value parameters for a queue with unobserved balking”, version 2 posted 24 March 2026, retrieved 2026-08-31, https://arxiv.org/html/2409.04090
• Cafe Success Hub, “Why the Wait”, first published on CoffeeBi 1 March 2017, retrieved 2026-08-31, https://www.cafesuccesshub.com/why-the-wait-queue/
• EuroShop, “Gone in 6 minutes: average queuing time UK shoppers are willing to wait”, dated 28 April 2014, retrieved 2026-08-31, https://www.euroshop-tradefair.com/en/Home/Archive/Gone_in_6_minutes_average_queuing_time_UK_shoppers_are_willing_to_wait
• Greggs plc, Preliminary Results for the 52 weeks ended 27 December 2025, RNS announcement published 3 March 2026, retrieved 2026-08-31, https://www.jamessharp.co.uk/market-news/greggs-preliminary-results/
• Manufacturing & Service Operations Management, Althenayyan, Ülkü, Yang and Cui, “Not All Lines Are Skipped Equally: An Experimental Investigation of Line-Sitting and Express Lines”, volume 27 issue 1, pages 287 to 304, published January 2025, retrieved 2026-08-31, https://doi.org/10.1287/msom.2022.0338
• Manufacturing & Service Operations Management, Song, Armony and Roels, “Queue Configurations and Operational Performance: An Interplay Between Customer Ownership and Queue Length Awareness”, volume 26 issue 6, pages 2284 to 2304, published November 2024, retrieved 2026-08-31, https://doi.org/10.1287/msom.2023.0202
• Meaningful Vision, “Why Afternoons are Now Driving Café Traffic”, published 25 November 2025, retrieved 2026-08-31, https://meaningfulvision.com/why-afternoons-are-now-driving-cafe-traffic/
• Meaningful Vision, “How Traffic Shifts are Redefining Café Culture”, published 12 November 2025, retrieved 2026-08-31, https://meaningfulvision.com/how-traffic-shifts-are-redefining-cafe-culture/
• Nation’s Restaurant News, “Starbucks unveils new ‘Siren Craft System’ to help baristas meet growing customer demand”, published 2 July 2024, retrieved 2026-08-31, https://www.nrn.com/quick-service/starbucks-unveils-new-siren-craft-system-to-help-baristas-meet-growing-customer-demand
• Perfect Daily Grind, “Should coffee shops batch matcha?”, published 19 March 2026, retrieved 2026-08-31, https://perfectdailygrind.com/2026/03/should-coffee-shops-batch-matcha/
• Production and Operations Management, Aksin, Gencer and Gunes, “How Observed Queue Length and Service Times Drive Reneging Behavior in Queues”, published online 19 September 2024, retrieved 2026-08-31, https://doi.org/10.1177/10591478241286504
• Production and Operations Management, Kang, Doroudi and Delasay, “Prioritization in the Presence of Self-ordering Opportunities in Omnichannel Services”, published 5 February 2024, retrieved 2026-08-31, https://doi.org/10.1177/10591478241231874
• The Motley Fool, “Starbucks (SBUX) Q3 2025 Earnings Call Transcript”, call held 29 July 2025, transcript published 5 August 2025, retrieved 2026-08-31, https://www.fool.com/earnings/call-transcripts/2025/08/05/starbucks-sbux-q3-2025-earnings-call-transcript/
• World Coffee Portal, “Value in the spotlight as competition heats up in £6.1bn UK branded coffee shop market”, published 30 January 2025, retrieved 2026-08-31, https://www.worldcoffeeportal.com/news/value-in-the-spotlight-as-competition-heats-up-in-6-1bn-uk-branded-coffee-shop-market/