The Decision Record and the Workpaper Behind It

A decision record is a single sheet stating the question, the decision, the date, who decided, what it rested on and what would change it. The workpaper is everything behind it: the documents, the computation file, the source log. The workpaper answers the demand to be shown the working. The decision record answers what was concluded, and on what basis.

Two documents, and almost every argument about record keeping comes from somebody trying to make one document do both jobs. The pair is not bureaucracy and it is not filing. The pair exists because of what happens to human beings roughly six months after any decision.

Six months after a decision, nobody remembers what was known on the day it was made. By then the outcome is in, and an outcome quietly rewrites memory in both directions. A decision that turned out badly gets remembered as obviously wrong at the time, with the warning signs standing out in a way nobody actually saw. A decision that turned out well gets remembered as confident and clear, when it was in fact narrow and argued about. The pair of documents exists to freeze the state of knowledgeWhat was actually known on the day, before anybody knew how it turned out. on the day. A decision can be judged fairly against nothing else.

The pattern is familiar from outside work. A household decides to take a loan for a repair, and a year later either the repair held and the choice was obvious, or the roof went again and everybody remembers being against it. If somebody had written four lines on the day, saying what the question was, what was chosen, what was known and what would have changed their minds, the argument a year later would be about the choosing rather than about the weather. A decision record is no more than those four lines. The paperwork is smaller than the idea.

One assignment carries the whole subject, the same one used under evidence in finance work and under the source log. The Kavery research desk, four people inside Kavery Capital Services Private Limited, was asked a question about Meenakshi Tubes Private Limited, a borrower the desk already lends to. Sharada Iyer produced the work, Prakash Nadar reviewed it, and Latha Menon commissioned it and made the decision at the end. The file was opened on 3 March, delivered on 11 March and decided on 12 March.

What the borrower makes never enters the reasoning. A bond, a fund, a warehouse or a single machine could be put in its place without changing a step. The subject is how a pile of evidence becomes one recorded decision, not whatever the evidence happens to describe. Evidence in finance work, how to build a source log and how to create an assumption register are each set out separately, and here the three are assembled rather than explained again.

Workpaper vs Decision Record: Which One Answers Which Question?

A workpaperEverything the conclusion was built from, kept in a form that can be shown to somebody else. is everything the conclusion was built from, kept so that it can be handed over. A decision recordOne sheet saying what was decided, when, by whom, and on what basis. is a single sheet saying what was decided and on what basis. The two are separated because they answer different questions, asked by different people at different times, and a document that tries to answer both answers neither well.

Every question anybody will ever ask about finished work is one of two questions, and each document takes one of them. Show me how that number was reached is a question about the arithmetic and the sources, and it is always answered by opening the workpaper. Asking what was concluded, and on what basis, is a question about the judgement, and it is always answered by opening the record. Once the two questions are heard as different, the split between the documents stops feeling like extra work and starts feeling like the only sane arrangement.

The merged version is easy to picture because most people have received one. A memo of forty printed sides arrives, the reasoning is spread through it, three exhibits contradict each other, and the actual decision is a sentence on side thirty one. The arithmetic was summarised into prose on the way in, so nobody can find the decision and nobody can rebuild a number either. One document that carries both jobs is worse than either document alone. The merged version is too long to be read and too summarised to be checked.

What Are the Six Fields of a Decision Record, and Which One Goes Missing?

Six fields, and the shape is fixed on purpose. A fixed shape means a missing field is visible instantly, in the way a blank line on a form is visible while a missing paragraph in an essay is not. Here are the six, with what each one carried on 12 March.

Five of the six get written without anybody being asked, and the sixth is the one that makes the record worth keeping. The sixth field is the only one that admits, in writing, on the day, that the decision could turn out to be wrong. Everybody finds the field uncomfortable to write, and that is exactly why it disappears. Notice what it does though. The field converts a decision from a verdict into something testable. Six months later a reviewer can check one thing: did the named condition occur, and if it did, was the decision revisited? Without that field the review has nothing to hold on to and collapses into an argument about whether the outcome was foreseeable.

The other subtlety is in the question field. The field holds the defined question, not the one that was first asked. The desk was asked whether it should be worried about Meenakshi Tubes Private Limited. A question in that form cannot be answered and cannot be finished. The recorded version carries a measure, a period and a threshold, and a record whose question has no answer cannot show that the work ever closed.

What Are the Four Parts of the Workpaper Behind It?

The workpaper on this assignment had four parts, and they were not four folders sitting side by side. The four parts nest, and the nesting is what lets a person walk from a printed figure back to a filed document without asking anybody for directions.

The computation file is the only part with a line to all three of the others, and that makes it the joint of the whole structure. The file reaches down into the documents through the log item numbers on its input cells, sideways into the assumption register wherever a cell holds a chosen figure rather than a filed one, and upwards into the decision record. The number printed in field five is the number this file produced. Removing the register removes the visibility of the choices. Removing the log removes the provenance. Removing the computation file leaves no path at all between the documents and the figure.

How Professional Finance Work Turns Evidence Into a Decision Record: Where Does the Rest of It Go?

Twenty one documents became one sheet. The narrowing sounds like loss, and most people producing their first serious file resist it. Everything in the folder felt necessary at the time it was collected. The compression is not loss and it is not summarising; it is the work itself, and each step throws away more than it keeps for a stated reason.

Stage by stage, the narrowing runs like this. The working file held 21 documents. The source log ran to 14 items, and two of those items were conversations rather than documents, so only 12 of the 21 documents made it into the log at all. The other nine were context, duplicates, and one superseded extract that disagreed with the filing. Of the 14 items, 11 were corroborated against a second independent item and 3 carry a visible mark saying they were not. Of the 14, only four turned out to be load-bearing, meaning the conclusion changes if any one of them is removed, and all four were statutory. The single sheet rests on those four items and three assumption register rows.

Each drop has a rule behind it, and the rules are worth naming because they are what stop compression turning into carelessness. A document leaves the log when nothing in the conclusion depends on it. An item stops being load-bearing when removing it leaves the answer where it was. Nothing is ever deleted from the working file, including the superseded extract that read Rs 20,90,00,000 against the filed Rs 21,20,00,000, a difference of Rs 30,00,000 or 1.4 per cent. The extract stays in the file, marked superseded, and a later reader can see the disagreement was caught rather than never noticed.

What Makes a Digital Workpaper Usable by Somebody Who Was Not There?

Everything above describes a structure. Whether the structure survives contact with a folder on a shared drive is a completely separate question, and the one that decides whether any of it was worth doing. Four habits carry almost all of it, and none of them is difficult.

The first is naming. A file name tells a stranger which file is current without them opening anything. Date first, then what it is, then a version number that only ever goes up. Version disciplineNaming files so that the current one is obvious without opening any of them. is not tidiness; it is the difference between one candidate file and four. The second is the input tabA separate sheet holding every raw number once, each tied to the source it came from.: every raw number lives once, on its own sheet, in its own cell, with the log item number beside it. Nothing raw is ever typed anywhere else in the file.

The third is that no number is ever typed inside a formula. A margin cell reads as a reference to two input cells and nothing else, and anybody clicking it lands on a cell rather than at a dead end. The fourth is cell-level tracingBeing able to click any figure and arrive at the item it came from, without asking anybody., and the first three habits add up to it: clicking any figure leads to the document it came from. The test for all four is the same. Can a person who was not there move backwards through the file without needing to talk to anybody?

Tracing the 10.8 per cent shows the chain. The figure sits in field five of the record. The output tab of the computation file produced it, and the formula in that tab points at two input cells. The two cells hold Rs 21,20,00,000 and Rs 2,30,00,000, each carrying a source log item number. Item 3 is the filed accounts for the year just reported, corroborated against item 1. And item 3 names the sheet and the line, so a stranger opens the filing and lands on the same two numbers. Five links, and a break at any one of them ends the trace.

Who Writes Each of the Two, and When Does Each One Get Written?

The person doing the work writes the workpaper continuously, as the work happens. The workpaper is never assembled afterwards. Assembling a workpaper at the end is the same activity as reconstructing it, done by the one person in the world who can no longer tell the difference between remembering and inferring. Sharada Iyer kept the log as items arrived and filled the input tab as numbers came in, between 3 March and 11 March.

A decision record states a decision rather than a finding, and whoever decided writes it, at the moment of deciding. Latha Menon wrote it on 12 March. If the analyst writes it instead, the record captures what the analyst believes was decided, and the two versions diverge far more often than anybody expects, usually on the sixth field. The reviewer writes neither. Prakash Nadar ran a check on the work, and a review pass is a check rather than a decision.

The effort split says something the structure alone does not.

Forty minutes against nine hours forty is under 7 per cent of the total, and that 7 per cent caught the one error that would have reversed the conclusion. The transcription had divided current year profit by prior year revenue. A review pass, and how an analyst runs one on their own work, is set out under errors and review. Only the ordering matters here. A record written first quietly becomes something the review is reluctant to contradict, so the record is written after the review pass and never before it.

How Does Anybody Know the Pair Actually Works?

Everything above is a claim about a file. The reproduction testSomebody who was not there rebuilding one figure from the file alone, with no conversations. is the only way to find out whether the claim is true, and it is deliberately cheap and deliberately late. Six months after the decision, a different person on the desk was handed the workpaper and asked to rebuild the 10.8 per cent from it. No conversation with anybody who did the original work. The rebuild took 25 minutes.

The 25 minutes converts an argument about good practice into a measurement, and it is the number to remember. Nine hours and forty minutes went into producing and checking the work. Getting one figure back out of it, half a year later, through somebody who was not in the room, cost 25 minutes. The ratio of nine hours forty to twenty five minutes is the entire case for naming files properly, keeping an input tab and writing item numbers beside cells, and no amount of argument about discipline makes the case as well as one measured rebuild.

The test also shows where the file is weak. Rereading the file never does. Whichever link the rebuilder stalls at is the link to fix, and it is almost never the one the author expects. The test says nothing about whether the decision was any good, only about whether the basis for it can be found. The quality of a decision and the traceability of its basis are separate properties, and only the second is settled by a workpaper.

Who Actually Opens One of These Later, and What Do They Do With It?

A lender revisiting an exposure does not want the note again. The note argued a case at a moment and the moment has passed. The lender wants the record instead, and specifically field six. The only question that matters at a review is whether the named condition happened. If the December order did not repeat, the review is short and structured: the condition occurred, the recorded consequence was a 5.0 per cent margin, and the decision needs revisiting. If field six is blank, the same review turns into a discussion of whether anybody could have seen it coming. Nobody wins that discussion.

An analyst uses the pair in the opposite direction, producing rather than reviewing. Filling field five forces the analyst to work out which items actually carry the conclusion, so the habit worth building is writing the record before anybody asks for it, even where nobody requires one. Most people discover, the first time they try, that they cannot name four items out of fourteen without going back and checking. The discovery is the useful part.

Somebody working alone gets more from this than a desk of four does, not less. There is no colleague to hand the file to and no reviewer to catch a transcription, so the file has to do the reviewing. A household making a decision it will revisit, whether to take a loan or to move a child to a different school, can run the same six fields: what was the question, what was chosen, on what date, who chose it, what was known at the time, and what would have made the other choice the right one. Six fields on one sheet, written on the day, is enough to make a decision reviewable a year later, and there is no smaller thing that achieves it.

The Kavery research desk, Kavery Capital Services Private Limited, Meenakshi Tubes Private Limited, Sharada Iyer, Prakash Nadar and Latha Menon are invented.
Educational material. Not advice on any investment, tax, budget or market position.