The Portfolio of Work: Evidence of What You Can Do

A portfolio of work shows what somebody can do, as opposed to a list saying where they did it. Almost everything a finance professional produces is confidential, so a portfolio is built by redaction: strip the entity, the figures and anything supplied in confidence, and what survives is the structure. The structure is the skill. The structure is therefore the evidence.

The problem underneath that is this. Somebody has to decide whether an analyst can produce work they would not have to rebuild, and the usual thing handed to them is a sentence the analyst wrote about themselves. Nobody can check a sentence of that kind. Because everybody in the role says the same six sentences, a reader can only accept the claim, discount it, or set it aside, and most set it aside. Hand the same reader one sheet of a redacted workpaper and something different happens: they stop assessing the person and start reading, and two minutes later they have a view they arrived at themselves. A view somebody reaches on their own is worth more to them than any claim that could have been made, and it costs nothing except the twenty minutes it takes to prepare the sheet.

Take one assignment. On the Kavery research desk, four people inside Kavery Capital Services Private Limited, Sharada Iyer was seven months into the role and working alone on a piece of work for the third time. The subject was Meenakshi Tubes Private Limited, a company the desk has money out to. The work ran from 3 March to a committee decision on 12 March, and every sheet of it is confidential: the accounts, the amounts, the internal floor, the decision, the name. Three sheets from it now sit in a portfolio. Not one amount and not one entity appears on any of them, and a reader can still tell in about two minutes whether the person who built them knows what they are doing. Working out why that is true is the whole of the exercise.

What is a portfolio of work, and who is it for?

A portfolio of workA small set of redacted artefacts that shows what somebody can do, as distinct from a description of the assignments they were on. is a small set of concrete things somebody produced, prepared so they can be shown, and chosen so that each one demonstrates something the others do not. A portfolio is closer to a carpenter's sample joint than to anything written about a career. Nobody asks a carpenter which houses they worked on. A customer picks up the joint, turns it over, looks at where two pieces of wood meet, and decides. The joint is not a description of skill. The joint is a small part of the skill itself, detached and handed over.

The one question a portfolio answers is whether the work would have to be rebuilt by whoever receives it, and every other thing it might say about its author is decoration on top of that. Three kinds of reader want the answer. Somebody deciding how much of the file to open before they trust the number at the end of it. Somebody deciding whether a piece of work can go to one person alone rather than paired with a second. And the author, the reader people forget: assembling one forces a person to look at their own output the way a stranger would, and that look is uncomfortable in a useful way.

A portfolio is the portable form of the credibility built inside an organisation. Credibility inside an organisation is built where people can see the work every day, in small kept promises: the answer arriving first, the review finding accepted without argument, the follow-up landing on the day it was promised. Daily observation works beautifully and it does not travel. The credibility sits inside the heads of about six people who watched the work happen. A portfolio is what can be carried to a reader who was never in the room, and it is the only version of the same evidence that survives the journey.

Building one requires no seniority, no team, no history behind it and nobody's permission. The absence of those requirements is not a small detail. A portfolio is made out of work rather than out of standing, so a portfolio is the one instrument of credibility fully available to somebody with no track record inside an organisation yet. The three sheets described below were prepared in the seventh month of a first role, by somebody whose first two solo assignments had come back for rework, and they are no weaker for it. Nothing they show depends on rank.

Is confidentiality a real obstacle, or is it an excuse?

Confidentiality is a real obstacle, and it deserves to be taken more seriously than it usually is rather than less. Almost everything a finance professional produces belongs to somebody else. Much of it was supplied by a third party on terms, and some of it was said in a room on the understanding that it stayed in the room. How material is classified and what each class permits is set out under data classification. The honest position is that the content genuinely cannot leave. The conclusion almost everybody draws from that, that they therefore have nothing to show, is wrong.

The list of what confidentiality removes is worth looking at carefully. Confidentiality removes the content: which entity, how much, when, and anything a person said privately. Confidentiality does not remove the shape of the work: the order the steps ran in, the fields that were filled, the checks that were run and when they were run, the way the findings were grouped. Consider somebody who prepares the tax return for a household down the road. The preparer cannot show a single line of that return, ever. The preparer can show the checklist they work through, the way they lay out a working file, and the note they write when a figure has two possible treatments. A prospective client would learn a great deal about whether to use them, and nothing whatsoever about the household.

Three tests need killing first. Each one is used every day and not one of them exists. The first is age: the assignment closed three years ago, so surely the figure is harmless now. The second is size: it is only a small number, and small numbers cannot hurt anybody. The third is obscurity: nobody outside the desk would recognise it anyway. No amount of elapsed time, no smallness and no obscurity converts a confidential figure into one that may be shown, and the third test is the worst of the three because it asks the one person who already knows the answer to judge whether a stranger could work it out. The rule is flat and it is easier to live with than a judgement: the figures come out. All of them. Every time.

How to Build a Professional Finance Portfolio of Work: What Goes In, and What Survives Redaction?

There is one test for whether an artefactOne concrete thing the work produced: a file, a record, a log, a sheet of findings. Something that exists, rather than a description of something that happened. belongs in a portfolio, and it is worth applying before any effort goes into it. With the content imagined away, is there anything left that a reader could form a view about? For some artefacts the answer is a great deal, and for others it is a title and some white space. The split runs along one line: whether the value of the artefact sits in its arrangement or in what it says.

Artefacts whose value is arrangement survive completely. A decision record is six labelled fields in a fixed order, and the labels carry the reasoning even when every value is gone. A workbook is six tabs in a fixed order with the last one holding the tests. A source log is a set of columns and a rule about when each cell gets written. An assumption register is six fields on every row and a marking that says which rows are load bearing. The arrangement was never the confidential part, so strip every value out of any of these and the arrangement is still fully legible.

Artefacts whose value is content do not survive, and here is the part that stings. The artefact an analyst is proudest of is almost certainly a content artefact. The proudest artefact is the memo with the sentence in it that changed the room, or the summary that made a difficult finding land in four lines. The memorable sentence is about one company's figures. With the figures gone the sentence is either empty or it is still the confidential finding wearing a coat, and neither of those goes in a portfolio. The artefacts that redact best are the dull ones, and being asked to show the dull ones is the first real surprise of building a portfolio.

There is one more entry on the list that people miss entirely: the handover note. One sheet, four headings, written for somebody who will pick the work up while its author is away. The handover note carries no finding and no figure at all. The absence of both makes it the single easiest thing on the desk to redact, and it demonstrates something a workbook cannot: whether work can be handed on cleanly.

What comes out, and in what order?

RedactionRemoving the entity, the figures and anything supplied in confidence from a piece of work, so that what remains can be shown. is four passes, and it is done in a fixed order for a reason that only becomes obvious the first time it is skipped. Each pass asks exactly one question. Run together, the four passes ask a blurred question. A blurred question produces a blurred answer, and the blur is where something confidential survives.

Pass one is the entity, and the name is the easy part of it. The hard part is the combination. A sector, plus a size, plus a location, plus one distinguishing fact is a name assembled out of pieces that each look innocent. A redacted sheet that needs to say the subject was a manufacturer of one component whose largest customer took a seventh of its revenue in a single quarter has not removed the entity, it has written a fingerprint of it. The discipline is to describe at the level the structure actually needs. For a decision record, a borrower is enough. For a workbook, a company that files annual accounts is enough. If the structure only makes sense with a sharper description, that artefact is not one that can be shown, and the right move is to pick a different artefact rather than a softer word.

Pass two is the figures, and the scope is wider than people expect. The scope is not just amounts. The scope reaches ratios, shares, growth rates, counts of the entity's own things, and dates that pin the work to a period. A margin is a figure. A percentage of revenue is a figure. The size of an order is a figure whether it is written in rupees or in words. One legitimate class of number stays, and without it the redacted sheet becomes unreadable: counts that describe the analyst's own process rather than anybody's business. Six tabs. Three findings. Four of fourteen logged items carried the conclusion. Process counts say something about how the work was done and nothing at all about the subject, and they are the numbers a reader most wants anyway.

Pass three is anything supplied in confidenceMaterial somebody supplied on the understanding that it would not travel beyond the work: a document handed over on terms, or something said in a room on that basis., and it is separated from the other two because it is the only pass that cannot be done by looking. The entity and the figures are on the sheet where they can be seen. Something said on the understanding that it stayed between two people is held in memory, and it may sit in the artefact as a conclusion with no visible source. The confidence comes out entirely. Not summarised, not paraphrased, not attributed to an unnamed party. A paraphrase of a confidential thing is still the confidential thing, and the paraphrase is worse than the original because it looks as though care was taken.

Pass four is the read-through by a stranger to the case, and it is the pass people skip. A self-read cannot work, and the reason is not carelessness. The reason is that the redactor knows what was removed, so the eye slides over the gaps and refills them from memory without announcing that it did. Somebody who has never seen the assignment reads the sheet cold. The stranger does not have to be a specialist. Because they cannot fill a single gap and will say so, a person from outside finance entirely is in some ways better. One question is enough: is there anything here that points at anybody.

What is kept, and why is the structure the part worth showing?

Set side by side, the two halves look like a mistake. Almost the entire visible substance of the work is gone. The remainder is a set of labels, an order, a list of tabs and three lines of findings, the part anybody would have discarded first if asked to save something. And yet that remainder is the evidence. The ledger below covers the three sheets that came out of one assignment.

Why is the surviving part the part worth showing? Because anybody who reviews files reads structureThe shape of a piece of work: the order it runs in, the fields it carries, the checks it holds and the way findings are grouped. far faster than they read content, and they know it predicts more. A tab order tells a reviewer within seconds whether the author builds files a second person can open. A separation between numbers somebody filed and numbers somebody chose tells them whether the author knows the difference. A checks tabThe last tab of a workbook, holding what was tested, on what date, and what the test found. tells them whether the file was tested before it reached them. A checks tab that records something having been found reads better than a blank one. A blank tab cannot tell anybody whether the file was clean or the check was never run.

The surviving part turns out to be exactly the part that is hardest to fake and quickest to read, a lucky accident rather than a design. A fluent sentence about a company can be written on a good day. A file with a fixed tab order, a checks tab with a dated test in it and findings grouped by kind cannot be produced without actually working that way. Each of those is a habit, and a habit leaves marks across a whole file rather than in one place. The reader is not admiring the structure. The reader is reading the structure as a trace of behaviour, in the same way that a tidy toolbox tells a customer something about a plumber before any work has started.

What did three artefacts from one assignment look like after redaction?

Three sheets came out of the March assignment. Each was redacted the same way and each shows something the other two cannot. There are three rather than one or six for exactly that reason.

The decision record is the most surprising of the three. In its live form the record is one sheet carrying the question, the decision, the date, who decided, what it rested on and what would change it. Redacted, every value comes out and every label stays. The question field reads that the question carried a measure, a period and a threshold, rather than being the loose one first asked. The decision field reads that a limit changed and a review date was attached to the change. The fifth field reads that the conclusion rested on four of fourteen logged items, all of them statutory, plus three named assumption rows. And the sixth field, the one that matters, reads that a condition was written down on the day which would make the decision wrong. The sixth field is the one most people leave off a record entirely, so its presence on a redacted sheet tells a reader more about the person than the other five put together.

The third sheet is the review findings, and it needs one clarification that saves people from a real error. Prakash Nadar ran the review pass on this file and wrote the findings; Sharada Iyer built the file and received them. A finding somebody writes about another person's work is evidence about the reviewer, not about the builder. Evidence about the builder is what was done with the finding. So the sheet that goes in is the closing side: the three findings restated by kind, the instances behind each, the change made for each and the date it closed. The three kinds were a number typed inside a formula instead of referenced, covering eleven cells; a cell reading the prior year where the label said current year, covering two; and an input cell carrying no source log item number, covering one. Findings by kindReview findings grouped by the class of problem they belong to, rather than listed by where in the file each instance sits. only work if the classes account for every instance, so the arithmetic is itself part of what the sheet demonstrates: eleven plus two plus one is fourteen, with nothing left over.

Three sheets, and between them a reader sees how the author reasons a decision, how they organise a build, and how they behave when somebody finds something wrong in their own file. The missing content matters as much. No conclusion about any company. No number. No sector. A reader who wanted to know what the desk decided in March would get absolutely nothing from all three sheets together, and a reader who wanted to know whether to hand that person a piece of work would be finished in two minutes.

Why does a decision record survive redaction better than a research note?

Two artefacts, the same redaction, opposite results. The contrast is the clearest illustration of where value sits, and it explains why people who try to build a portfolio out of their best writing give up and conclude the whole idea is impossible.

The mechanism is simple once it is named. A note is an argument about particular numbers, so the numbers are load bearing and removing them collapses the argument. A decision record is a shape into which any decision can be poured, so removing the contents leaves the shape standing and the shape is what was skilled about it. Anything whose value is structural redacts well, and anything whose value is a specific claim about a specific entity redacts to nothing, and that single distinction decides the entire selection.

How many artefacts, and how are they chosen?

Three is a good number and five is usually too many. The constraint is not modesty. The reader's attention is finite, and it is spent whether or not anything is bought with it. SelectionChoosing artefacts so that each one shows something the others do not, rather than choosing the artefacts that happen to be favourites. is therefore a coverage problem rather than a quality problem. The question for each slot is not whether the artefact is good. The question is what this artefact shows that no other artefact in the set shows.

A second artefact showing the same skill does not add half an artefact or a quarter of one. The second artefact adds nothing, and it costs the attention needed for the slot it took. The reason is that the reader is not accumulating evidence of how good the author is at building files; they formed that view on the first one and the second cannot move it much. The reader is still missing everything the first file does not touch. The right question at each slot is which column is still empty. The wrong question, the one everybody actually asks, is which of my files is best.

One practical note on ordering. Put the artefact with the least explanation attached to it first. A reader who has to be told what they are looking at before they can look at it has already been handed a claim rather than a thing, and the whole point of the exercise was to stop doing that.

How to Build a Finance Portfolio of Work: When Is It Actually Done?

The timing of the work matters as much as the contents, and getting the timing wrong is the most common way the whole thing fails. The rule has three parts and the first one carries most of the weight: one artefact is redacted at the moment an assignment closes, not at the moment something is needed to show.

The reason is not tidiness. The reason is that at closing time the analyst is already inside the material with a legitimate working reason to be there, still remembers which parts were supplied in confidence, and the artefact is fresh enough for its shape to be visible. Twenty minutes, once, and the sheet is done. Six months later none of that is true. Nobody remembers what was said privately. There is no current reason to be in the file, and going back into confidential material to look for something flattering is a bad reason dressed as an errand. Going back in is also, on any honest reading, exactly the behaviour the confidentiality undertaking exists to prevent.

The second part of the rule is that the redacted sheet is rebuilt, never edited. Opening the real file, deleting things and saving a copy does not work. Files carry more than what is visible on the surface: hidden rows, comments, tracked changes, cached values behind a deleted formula, and a properties panel with a name in it. The structure is typed into a blank file from scratch. Because the structure is short by definition, retyping takes longer to describe than to do, and it removes an entire class of accident that no amount of careful deleting can remove.

The third part is that artefacts are replaced rather than accumulated. Three slots are kept. When a new assignment produces something that covers a column better than what is currently in that slot, it is swapped in and the old sheet deleted. A portfolio that grows year by year turns back into a list, slowly, and nobody notices it happening because each individual addition seemed reasonable. A short private index of what is in each slot and which skill it covers keeps the swap decision to a minute rather than a re-read.

How is a portfolio different from a list of what somebody has done?

Both make the same claim. Only one of them lets the reader check it, and that single difference decides everything else. A list of experienceA description of the assignments somebody was on, which the reader has no way of verifying and therefore has to either accept or discount. asks to be believed. A redacted artefact asks to be read. Asking to be believed can only work if the reader already trusts the writer. In precisely that situation the document was not needed.

The line and the artefact carry identical information about what was done. The difference is in what the reader can do next, and that difference is the only one that has ever mattered. There is a second effect worth naming. The list invites the reader to compare its author against everybody else who wrote the same line, and everybody writes the same line. The artefact invites them to look at one specific thing that nobody else has, and comparison stops being the mode they are in.

Who actually reads one, and what do they do with it in two minutes?

Watching a reader work is worth the trouble. The reading is faster and more mechanical than people imagine, and it does not proceed in the order anybody would expect. Nobody starts at the beginning. A person who reviews files for a living opens the checks first. The checks answer whether anybody tested the work before it arrived. Then the arrangement. The arrangement answers whether they can find anything. Only then, and often not at all, whatever content is left.

The last row is not a decoration. The same mechanism runs at household scale, and it is where most people first meet it without noticing. Two people offer to keep a small set of books. One says they have done this for many households for many years. The other shows a blank version of the file they keep, with the tabs labelled and a sheet at the back listing what they check each quarter. The second person has said nothing about any client and has shown everything about how they work. The second person is not more impressive; something is simply known about them that was not known before, and knowing beats being impressed every time.

There is one more reader, and it is the one that pays off soonest. Building the set forces a person to look at their own output through a stranger's eyes, and the first attempt usually turns up one of the three columns with nothing worth showing in it. The empty column is useful information about what to volunteer for next, and it arrives free, as a side effect of an afternoon spent redacting three sheets.

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