table-plan

---
name: table-plan
description: Rebuilds your floor plan around the party sizes you actually get, so the same room seats more people without hurrying anyone out. Use when tables sit half empty while people wait at the door, when you are buying furniture, or when you want to know how many covers the room can really do.
---

# Turn the covers you already have, without rushing anyone

You give this a rough sketch of the room with every table on it, eight weeks of bookings or till receipts, and the widths of your exits. You get back a new table layout with the number of two-seaters, four-seaters and six-seaters written down, a seating rule for each party size, the occupancy figure your fire risk assessment has to live inside, and a printed plan for the door. It works from the party sizes you really get, not the ones you remember.

## What it does

1. **Count the room before you change anything.** List every table you own with three facts against each one: how many people sit at it, whether it is fixed to the wall or can be moved, and which other table it can be pushed against. Measure the clear floor area of each dining area in square metres, excluding the bar serving area, the toilets, the stairway enclosure, stores and fixed furniture. Measure the clear width of every exit door with the door fully open, taking off anything that projects into the opening. Write the totals at the top: total seats, total tables, and seats in two-tops, four-tops and six-tops. Most rooms fail here before any maths starts, because nobody has ever written down what they own. The Chevys restaurant studied by Cornell had 56 tables in the main dining room and 53 of them were four-tops, which is the fact that explained everything that followed.

2. **Get the real party-size distribution out of your own records, never out of memory.** Pull eight weeks of covers from the booking diary, the till or the reservation platform's export, and count how many parties were of one, two, three, four, five, six, seven, eight and nine or more. Do it separately for your busy hours and your quiet ones, because they are different businesses. Express each as a percentage of all parties, and also as a percentage of all covers, since a handful of parties of eight can be a fifth of your covers while being a twentieth of your bookings. Kimes and Thompson found the Chevys seat occupancy "rarely exceeded 50%" and gave the cause in one sentence: "This is primarily due to the poor table mix (most of the tables were tables for four and over half of the parties had only one or two customers)." If half your parties are ones and twos and most of your tables seat four, you are seating one person per two seats and paying rent on the rest.

3. **Work out the naive ideal mix, then treat it as a starting point and not an answer.** For each party size, assign the table it ought to sit at: ones and twos to a two-top, threes and fours to a four-top, fives and sixes to a six-top, sevens and eights to a two-top and a six-top pushed together, nines to a four-top and a six-top. Weight each table size by the share of parties that need it, multiply by your total seat count, and convert seats back into tables. In the Cornell worked example, 230 seats produced "a recommended mix of 43 2-tops, 27 4-tops, and six 6-tops" against an existing floor of almost nothing but four-tops. Run the same arithmetic on your own numbers and write the answer down in tables, not seats. Then stop, because the paper is explicit that the real answer is messier: "identifying the best mix of tables cannot be done so simplistically", since arrival times, party sizes and how long people stay all vary night to night.

4. **Measure how long a table is actually occupied, by party size, and count the tail.** Take the time the party sat down and the time the table was clean and ready for the next party, not the time the bill was paid. The Cornell simulation is careful about this: "Parts of the service duration, most notably seating the party and bussing the table, occur when the customers are not, in fact, at the table." Record a mean and a spread for each party size at each service. A table of two at 19:00 on a Saturday is not the same length of job as a table of six at 13:00 on a Sunday, and a plan built on one average will be wrong at both ends.

5. **Name your peak hours and size everything for those.** Count covers by hour by day across the same eight weeks and mark the hours where you regularly turn people away or hold them at the bar. The Chevys restaurant had eleven such hours a week out of a full trading week, and produced "over one-third (34%) of its annual revenues" inside ten of them. Everything in this plan is aimed at those hours. Off-peak, the layout barely matters. If you cannot identify a genuine peak from the data, say so plainly and stop, because rearranging a room that is never full will not add a penny.

6. **Check the new plan against the fire occupancy figure before you order a single table.** Leicestershire Fire and Rescue Service, setting out the calculation for licensed premises, states that "The occupancy figure for any building is based on the lesser of the two following calculations; the number of persons who can safely reside in the premises ... and the width and capacity of the exit routes to allow them to escape safely." Their floor space factors are 1.0 square metres per person for "Dining rooms, Seated Lounge/Bar, Restaurants", 0.5 for assembly and dance areas and 0.3 for "Standing areas, Bars without seating". On exits, a route between 750mm and 1050mm takes 100 people at medium risk, and "Over 1050mm" adds "+15 persons" for every extra 75mm. Then apply the rule that catches people out: "you must remember to discount the largest exit or exits if they are close together ... as they may be unavailable due to the fire." Note also that "An occupancy limit of 60 persons will apply where only one exit exists". Print both numbers, floor space and discounted exit capacity, and take the lower one as the ceiling the new layout must sit under.

7. **Keep the gangways and the accessible seats in the plan, not in the apology.** The same guidance gives a minimum escape route width that "should ideally be 1050mm but in any case not less than 750mm and where wheelchair users are likely to use it not less than 900mm", and for moveable seating asks for "adequate gangway widths (1.05m) and seat way widths (305mm)". Mark those corridors on the plan as no-table zones in a different colour so nobody quietly fills them on a busy Saturday. Then place at least one table on the main circulation route that a wheelchair user can use without anything being moved, and record which table it is. Section 20 of the Equality Act 2010 sets a duty where "a physical feature puts a disabled person at a substantial disadvantage", requiring "such steps as it is reasonable to have to take to avoid the disadvantage", which it says includes "removing the physical feature in question", "altering it", or "providing a reasonable means of avoiding it". A layout that squeezes the last four covers in by blocking the route to the accessible table has not gained four covers, it has created a problem.

8. **Write a seating rule for every party size, including which tables may be combined and which may not.** For each size, name the first-choice table, the second choice, and the combination to build if both are gone. The Cornell work on combinable tables is worth knowing before you make every table combinable: Thompson "found that combinable tables worked better only for smaller restaurants with small mean party sizes and that dedicated tables worked better in other situations", and the warning attached is that combinable layouts lose money when "tables are held out of service until adjacent tables become available". So write an explicit rule that a table is never held empty waiting for its neighbour unless a booked party is due within a stated number of minutes, and write that number. Add the rule that no party is moved once seated and no party is told when the table is needed back, because the gain in this plan comes from the shape of the furniture, not from hurrying the guest.

9. **Produce the four things that go on the wall.** A before and after floor plan with every table numbered. A table matrix card for the door, listing party size against first choice, second choice and combination. A one-page occupancy sheet showing the floor space figure, the exit capacity figure with the largest exit discounted, and the lower of the two as the house limit. And a change list with the cost: how many tables to buy, how many to sell or store, which fixed seating would have to be moved and which cannot be. Chevys did not build the mix the model produced, they built "39 2-tops, 35 4-tops, and 2 6-tops" for design reasons, and the model put that within "about 1.3%" of the ideal. Expect the same. The plan is a target to move towards over a refit, not a demand to empty the room on Monday.

## Then it checks

1. Every table in the after plan carries a number, a seat count and a note saying whether it is fixed or movable, and the after plan's total seat count is stated next to the before plan's.
2. The party-size percentages are derived from a stated number of parties across a stated date range, and the source is named as the booking diary, the till or the platform export, with no figure taken from recollection.
3. The occupancy sheet shows both calculations, floor space and exit capacity with the largest exit discounted, and the house limit equals the lower of the two.
4. No table in the after plan stands inside a marked gangway, and at least one table is identified by number as reachable and usable without moving furniture.
5. Every party size from one to nine or more has a seating rule with a first choice, a second choice and a combination, and every combination names the two specific table numbers that make it.
6. The change list gives a count of tables to add and remove, and no item reads "some", "a few" or "as needed".

Any check fails: name it, redo that step once. Failed twice: say what is wrong and stop.

## Rules
- Public information only.
- Never invent a fact, a number or a quote.
- Anything sent in someone's name says whose name it is.
- Never state an occupancy figure as approved. This plan produces a working calculation for the fire risk assessment. The responsible person signs the assessment and the licensing authority may attach its own condition, so the number here is a proposal to be checked, never a permission.
- Never gain covers by shortening the time a guest is given. Every extra cover in this plan comes from matching table size to party size and from clearing tables faster after the guest has left, never from a turn time imposed on someone mid-meal.
- Never remove the last accessible table or narrow a marked gangway to fit another two-top in, whatever the arithmetic says the room could hold.
- This output is a working document prepared for the owner's fire risk assessor, licensing adviser or solicitor to check before the layout changes or the occupancy figure is used. It performs arithmetic on measurements you supplied. It is not advice on fire safety law, licensing conditions or the Equality Act.

## Built from
- Sheryl E. Kimes and Gary M. Thompson, Cornell University, "Restaurant Revenue Management at Chevys: Determining the Best Table Mix", https://ecommons.cornell.edu/server/api/core/bitstreams/f812c462-85df-4ce8-83d4-898868776c81/content, no publication date shown on the page, read 14 September 2026: the party-size-to-table-mix method in step 3, the seat occupancy diagnosis in step 2, the service-tail warning in step 4, the peak-hour concentration in step 5, and the combinable-table caution in step 8.
- Leicestershire Fire and Rescue Service, "Calculating Occupancy for Licensed Premises and Other Places of Assembly", https://leics-fire.gov.uk/uploads/calculating-occupancy-figures-for-licensed-premises-(2).pdf, no publication date shown on the page, read 14 September 2026: the floor space factors, exit width capacities, the discount-the-largest-exit rule, the single-exit limit of 60 and the gangway widths, which together set steps 6 and 7 and the house limit in step 9.
- Equality Act 2010, section 20, "Duty to make adjustments", https://www.legislation.gov.uk/ukpga/2010/15/section/20, latest available revised version as shown on 14 September 2026, read 14 September 2026: the second requirement on physical features, which is why step 7 places the accessible table on the plan before the last two-top rather than after it.

# Rebuild a restaurant floor plan around the party sizes it actually gets

## You are given
- A list of current tables: table number, seats, fixed or movable, and which table numbers it can be pushed against.
- The clear floor area in square metres of each dining area, excluding toilets, stairway enclosures, bar serving areas, stores and fixed furniture.
- The clear width in millimetres of every exit door measured with the door fully open, and how many exits the room has.
- A booking or till export of at least eight weeks, one row per party: date, time seated, time the table was free again, number of people, total spend.

Anything missing is reported, never filled in.

## Produce
Write four files into a folder named `table-plan-<venue>-<YYYY-MM-DD>`.

1. `party-mix.csv` - columns: `party_size, parties, pct_of_parties, covers, pct_of_covers, mean_minutes_occupied, median_minutes_occupied, mean_spend_gbp, peak_parties, offpeak_parties`. Rows for party sizes 1 to 8 plus a final `9+`.
2. `table-mix.csv` - columns: `table_size_seats, tables_now, seats_now, tables_recommended, seats_recommended, change`. `change` is signed. A `TOTAL` row must show `seats_now` equal to `seats_recommended`.
3. `occupancy.csv` - columns: `area_name, area_m2, floor_space_factor, persons_by_floor_area, exit_id, exit_width_mm, exit_capacity_persons, largest_exit_discounted, exit_capacity_after_discount, house_limit`.
4. `seating-rules.csv` - columns: `party_size, first_choice_tables, second_choice_tables, combination_tables, notes`.

Then `README.md`: venue name, date range, number of parties counted, the house limit with both calculations shown, and a numbered list of tables to add and remove.

## Rules
- Floor space factors: 1.0 square metres per person for dining rooms, seated lounges and restaurants; 0.5 for assembly and dance areas; 0.3 for standing areas and bars without seating.
- Exit capacity: a route 750mm to 1050mm carries 100 persons at medium risk; over 1050mm add 15 persons per extra 75mm; no single door counts as wider than 2m. Discount the largest exit entirely before totalling. One exit only means a limit of 60 persons.
- `house_limit` is the lower of the floor area total and the discounted exit total.
- The recommended mix must use the same total seat count as the current mix.
- Time occupied runs from seated to table free again. If the export only has seated and paid, say so in `README.md` and label the column as excluding clearing time.
- Never output a figure calculated from fewer than 20 parties for that party size. Write `insufficient data`.
- Percentages to one decimal place, money to two. No cell may read approximately, circa or tbc.

## Return
The absolute path of the folder, the four filenames with row counts, the house limit with the two calculations behind it, the current and recommended table mix as two short lists, and every input that was missing or produced `insufficient data`.