Ask It Twice: Why the Same Evening Comes Back With Five Different Answers

Take one evening. A hundred guests, seated, a room you have already walked. Send it to five people and ask a single question: how many tables.

You will get back twelve, thirteen, ten, thirteen, and “depends on the linen.”

Nobody lied to you. Every one of those numbers has a reason behind it, and if you ask, you will get one. The rounds are 1.8 metres so that is ten. They always plan eight to a round. The venue diagram says twelve.

Five defensible answers to a question that has one correct answer. That is the whole problem, and it is not a small one, because the table count is not a small number.

Linen is priced per table. Centrepiece vessels are counted per table. Pin spots are hung per table. Stem counts, service pass timings and load-in minutes all sit downstream of it. Get the table count wrong by three and you have three tables set, dressed, lit and bare.

Variance has three sources and only one of them is honest

The first is different inputs. You told one person the room was fourteen metres and another forty-six feet, and one of them rounded. This is a data problem, it is boring, and it is fixable.

The second is different rules. Everyone is computing correctly from a different rule. Hotel banqueting plans 0.55 metres of table edge per diner. I would plan 0.65. Both are arithmetic. They are arithmetic on top of two different opinions about how much room a person needs to eat without apologising to the person beside them.

This second kind is not a bug. It becomes one the moment nobody says which opinion is in play.

The third is no rules at all. Somebody looked at the room and said a number. This is the kind that gets defended hardest and costs the most.

The fix for any of it is not “be more accurate.” The fix is to say which rule you used, in public, before anyone asks.

The kind that hides inside the arithmetic

The second kind and the third kind are hard to tell apart from outside, because both arrive as a whole number and neither one shows its working.

Take the division itself. Two people can hold the same rule and still disagree, because one of them rounds the result and the other floors it. Rounding a table count means that at some headcounts something will, with total confidence, seat 8.375 people at a table. There is no such person. There is no such table.

Downstream, that one decision separates and nothing pulls it back together. The florist quotes a hundred and fifty stems against thirteen vessels. The lighting plan hangs pin spots for ten. Three tables, dressed and unlit and unflowered, in a room where somebody has paid for all of it.

Nobody in that chain did arithmetic wrong. They did correct arithmetic on a number that had already moved.

One number, and a stated order of precedence

So decide the order once, write it down, and make everything downstream read the same number.

A declared table count wins over everything. If you say there are twelve tables, there are twelve tables. A declared seats-per-table comes next. Table geometry comes third: perimeter divided by 0.65 metres a cover, floored and never rounded, because a fraction of a chair is not a chair. A default table comes last, and it says out loud that it is a default.

Four steps, in that order, every time. The point of writing them down is not that they are clever. It is that a florist and a lighting designer can be handed the same four steps and arrive at the same number without a phone call.

A 1.8 metre round seats eight. Not ten. Ten is the banqueting squeeze and you are entitled to it, but you have to declare it — and then the linen, the vessels, the pin spots and the stems all move together, because they are finally reading the same number.

Three classes, and the third one is the honest one

Every number in a plan belongs to exactly one of three classes, and the discipline worth adopting is to print the class next to the number.

Computed means arithmetic on a named external standard. Amps from watts is Ohm’s law, and none of it is mine. The eighty per cent circuit headroom is NEC 210.19(A) and 210.20(A), which sizes a continuous load at 125 per cent — the same statement turned over. Two metres of clear exit width for four hundred guests is five millimetres a person, the convention behind IBC 1005 and Approved Document B. You can check every one of these against a document that existed long before the plan did.

Stated means mine, and labelled as mine. The 0.65 metre cover is the example. There is no standard, I picked a number, and the honest version says so in the same sentence it gives you the number. A stated rule is still deterministic. It simply does not get to hide behind somebody else’s letterhead.

Judged is the class most planning documents would never admit to having.

The energy curve that shapes an evening — the peak, the flat middles, the fault line I would call The Turn — is a point of view about how parties feel, written out as numbers. Finding the sharpest change in a curve is defensible mathematics. The curve it runs on is taste.

So it does not belong in the same list as the other two, and neither do scenario fit scores, readiness rubrics, stem ranking or moodboard confidence. Every one of those is professional opinion in numeric costume. They are useful. They are still opinions. Putting them next to Ohm’s law would make Ohm’s law look like an opinion too.

A system that computes everything is lying about something.

What to do with this

Next time a number arrives — from a venue, from a vendor, from a planning tool, from me, from anywhere — ask one question. Which rule, and where does it come from?

There are only three possible answers. A named standard, which you can go and read. A stated position, so you now know whose position it is and can decide whether you share it. Or a shrug, which is the most valuable thing you will learn in that conversation.

A rule worth trusting arrives with four things: the formula, the inputs it needs, the basis it rests on, and the limits it will not cross. An exit-width check should say plainly that it is an order-of-magnitude check and not a code sign-off, because only the authority having jurisdiction can give you one of those. A queue estimate should say that above full utilisation it stops estimating a wait and reports the backlog instead, because there is no steady state left to estimate.

A rule that will not tell you where it stops is not a rule. It is a guess with good posture.

Ask it twice

Here is the test, and you can run it on anything.

Put the same evening in twice. Change nothing. If the answer moves, something in there is judgement wearing a lab coat, and you are entitled to know which part.

Answering the same is not a boast about being clever. It is the least a thing that produces numbers can do. What you should be able to get back, from any of it, is the number, the rule that made it, and a straight account of which parts are taste and always will be.

Pretty is not a plan. And a plan that gives you a different answer on Tuesday was never one.