Skip to content

The number that was only ever an assumption

In 2019 somebody typed 12 into a cell to see what would happen. Seven years later the household describes itself as being on a five-crore plan, and nobody in it can say where five crore came from. A spreadsheet renders a guess and a bank balance in the same font.

Risk & PsychologyIntermediate14 min read
Browse Risk & Psychology(105)

The sheet is built on a Saturday in 2019 and it is a reasonable piece of work. Column A holds the years, column B the monthly contribution of ₹40,000, and two cells at the top hold the assumptions: 12 for the annual return and 6 for inflation. The twelve is not a forecast and was never meant to be one — it was typed in to see what the shape looked like, with a vague intention of trying nine and fourteen afterwards, which never happened because dinner was ready. The bottom-right cell reads ₹5,13,22,785. Seven years later that cell is the only part of the exercise anybody remembers. The household refers to itself as being on a five-crore plan. A relative has been told the figure. A decision about how much equity to hold has been taken with it in mind. And nobody, if asked, can say why twelve and not nine, because the twelve is not stored anywhere as a choice. It is stored as a number, in a cell, next to other numbers that are facts.

Think of it like this
The quotation that became the bill

A contractor quotes ₹4,20,000 for the kitchen, on the back of an envelope, subject to the cabinet finish you eventually pick. Nine months later the family has planned around ₹4,20,000, argued about ₹4,20,000, and is genuinely surprised when the invoice differs. The envelope never said it was a bill. It also never said it was not.

In the market

A projection is a quotation with conditions attached. Once it is computed, the conditions live in two small cells at the top and the answer lives in bold at the bottom, and only one of those gets quoted at dinner.

What a spreadsheet destroys

The problem is not that the number is wrong. Twelve per cent may well turn out to be close. The problem is that a spreadsheet renders three completely different kinds of number in identical type, and after a few months nothing distinguishes them.

Kind of cellExample in this sheetWhat can be checked, and how
MeasuredThe ₹40,000 standing instruction, and the date it was set upVerifiable against a statement in under a minute. This is the only kind of cell that is a fact
ChosenThe 12, the 6, the retirement age of 58, the assumption that the contribution never risesNot checkable at all, because there is nothing to check them against. Each is a judgement someone made once, in a mood, with a reason that was not recorded
ComputedThe ₹5,13,22,785 at the bottomArithmetically correct and epistemically worthless on its own. It is exactly as reliable as the weakest chosen cell above it, and it does not look it

How much of the answer is the assumption

The way to find out what a projection is actually made of is to run it again with a different guess in the guess cell and nothing else changed. This is the whole of [[sensitivity analysis]], it takes about two minutes, and almost nobody does it — because the first run already produced an answer, and a second run can only make the answer less comfortable.

Worked example
The same ₹40,000 a month for 22 years, three assumptions
A monthly contribution of ₹40,000 for 264 months, paid at the end of each month, compounded at the annual rate divided by twelve
At 12 per cent a yearThe cell that was typed in first, and the only figure anybody has quoted since₹5,13,22,785
At 10 per cent a yearTwo percentage points lower. The answer falls by ₹1,31,95,813 — more than a crore, from a change nobody would notice in a cell₹3,81,26,972
At 8 per cent a yearFour points lower than the original guess, and not an unreasonable outcome for a 22-year period₹2,86,71,513
The spread across the threeThe distance between the top and bottom runs, all of it produced by the one cell that was never revisited₹2,26,51,272
Total contributed over the whole periodAdd these nominal rupees only to see the scale of the exercise — a sum of rupees paid across 22 years is not a today’s-money quantity and should never be compared with one40,000 × 264 = ₹1,05,60,000
What the middle run buys, in today’s moneyAt 6 per cent inflation, 1.06²² is about 3.60. The corpus that reads as ₹3.81 crore in the sheet has the purchasing power of a little over a crore now₹3,81,26,972 ÷ 1.06²² ≈ ₹1,05,80,000
The uncertainty introduced by one unexamined cell is ₹2.27 crore — 59 per cent of the middle run, from a choice made in about four seconds. Note that even the timing convention above is a chosen cell: shift the contributions to the start of each month and every figure rises by one month of interest — roughly 0.7 to 1 per cent — which nobody would call a change to the plan. That is not an argument for abandoning the projection. A range of ₹2.9 crore to ₹5.1 crore, in nominal rupees worth a little over a quarter of today’s, is genuinely useful information about what this plan can and cannot fund — and the same three runs show that the contribution and the number of years are yours to set while the return is not. It is an argument against ever quoting the single number, which is the only thing that has been quoted for seven years.

Why the tail of the number does so much damage

The cell reads ₹5,13,22,785. Every digit after the first two is manufactured. The inputs support, at best, a statement of the form somewhere in the region of three to five crore, and the output is displayed to the nearest ten rupees — because that is what a spreadsheet does, not because anything in the calculation justifies it. Precision is read by human beings as a signal of care, so the tail actively increases confidence in the number in exact proportion to how meaningless it is.

  • A precise number invites planning around it. ₹5,13,22,785 sounds like something somebody worked out. "Three to five crore" sounds like a guess, which is what both of them are.
  • A precise number [[anchoring|anchors]] every later conversation. Once five crore is in the room, a subsequent honest estimate of three crore is experienced as a setback rather than as an equally valid draw from the same distribution.
  • A precise number is hard to argue with. Disagreeing with it feels like disputing arithmetic, when the arithmetic is not in dispute and never was.
  • A precise number hides its own age. The sheet does not display the date the 12 was chosen, so a guess made before a change of job, a change of income and two changes in market conditions is still presented as current.

Repairing a sheet you have already built

Four things that restore what the file threw away
  1. 1
    Mark every cell as measured, chosen or computed

    A background colour is enough. The purpose is not tidiness — it is that a chosen cell should never again be readable as a fact by somebody glancing at the sheet in two years, including you.

  2. 2
    Put the date and the reason next to every chosen cell

    One short line: what the number is based on and when it was picked. "12% — long-run Indian equity, chosen Mar 2019, not revisited" is a complete entry and it does the entire job. Without the date, a stale guess and a fresh one look identical.

  3. 3
    Replace each point estimate with three runs

    A low, a central and a high, and report all three or none. The discipline is that the sheet must never again be able to produce a single number, because a single number is the form in which this error travels.

  4. 4
    Add one line converting the answer to today’s money

    Divide the corpus by one plus inflation raised to the number of years. It is one formula and it prevents the most common misreading of any long projection, which is treating a distant rupee as though it were a present one.

◆ Your call

Somebody finally opens the 2019 sheet

The bottom cell says ₹5,13,22,785. The return cell says 12 and has never been changed. The household has been describing itself as on a five-crore plan for four years, and a decision about equity allocation was taken partly on that basis.

Check yourself

A 22-year projection of ₹40,000 a month gives ₹5,13,22,785 at 12 per cent and ₹2,86,71,513 at 8 per cent. What does that comparison establish?

Simple bhasha mein
Kachcha andaaza, pakka number

2019 ke ek Saturday ko sheet banayi: ₹40,000 mahina, 22 saal, aur upar do cell — 12 aur 6. Baarah koi hisaab nahi tha, bas "dekhte hain kaisa lagta hai" tha; nau aur chaudah try karna reh gaya kyunki khaana lag gaya. Neeche wale cell mein aaya ₹5,13,22,785, aur saat saal se ghar mein bas yahi ek line chalti hai — "paanch crore wala plan". Aaj wahi sheet dobara chalao, sirf ek cell badal kar: 10% pe ₹3,81,26,972, 8% pe ₹2,86,71,513. Ek cell ne jawab ₹2.27 crore hila diya — yaani beech wale jawab ka 59%. Jo aapne khud tay kiya (kitna daalna hai, kitne saal chalana hai) usse zyada asar us andaaze ka hai jo kisi ne socha bhi nahi tha. Aur mehngai 6% maano toh ₹3.81 crore 22 saal baad aaj ke kareeb ₹1.06 crore jitna hi khareedega. Sheet mein bank balance aur andaaza ek hi font mein chhapta hai — isliye har andaaze wale cell ke saath tareekh aur wajah likho, ek number ki jagah teen chalao, aur ek line jodo jo jawab ko aaj ke paise mein badal de.

What to remember
  • A spreadsheet stores values and discards provenance, so a guess and a bank balance become indistinguishable within months.
  • Re-run every projection at three assumptions and report the range — a single number is the form in which this error travels.
  • The displayed precision comes from the software, not the inputs, and it raises confidence in exact proportion to its meaninglessness.
  • Label each chosen cell with its date and its basis; an undated assumption cannot be told from a current one.
  • Add one line converting the answer to today’s money, and never compare a sum of nominal rupees paid across decades with anything.
Finished this lesson?

Mark it done to track your progress through the curriculum.