Primary vs Secondary Research: What Each One Can Settle
Asking somebody a question and writing down what they said is one route to a fact. Working from a record that somebody else made, checked and put out is the other. The two words cross rather than nest: the record of a call is an original record reached entirely second hand, and an answer from a head teacher is first hand and leaves nothing anybody can open.
Two facts, both useful. Which one is the primary one?
The definitions are the easy part and the question is where everybody comes unstuck, so the question comes first. Two facts sit on the desk. Both facts concern a small maker of hard-bound school registers called Anjani Stationers Private Limited, an invented business these notes use for teaching, and both are perfectly good facts.
The first: what the business said about its own year, taken from the written record of the meeting in which it said it. The second: that head teachers around the district have come across the name of one maker of school registers and have not come across the name of the other. Now answer the question in the heading. Which of those two is the primary one?
Both answers are defensible, and a great many careful people give one of them without noticing the other exists. Sit with both for a moment. The first fact came off a document that sits about as close to the thing itself as any document ever sits. Nobody summarised it, nobody paraphrased it, nobody selected from it. If closeness to the event is what the word means, that document wins and it wins easily. The second fact was got by asking. Nobody handed it over, nobody had printed it, and it exists in writing only because somebody went out and put a question to a person who could answer it. If doing the work yourself is what the word means, the second one wins and it wins just as easily.
Both readings are correct, and they are correct about two different questions that happen to have been given the same word. That collision is the whole of this guide, and everything after this point is either a way of seeing the collision clearly or a consequence of taking it seriously.
Before either side can be set against the other, each has to be said properly on its own. Each of the two is said in full and said short, because each is covered at length elsewhere in these notes.
What is it to ask somebody, said in full and on its own?
The analyst works out who would be in a position to know. A question goes to that person about dealings that are theirs, on a point they are at liberty to state without it costing them anything, and the answer is written down, along with who said it and when. Asking is the whole act, with no ceremony in it and no technique worth the name.
The asking exists for what it reaches. Some facts about a trade were never any one business's transaction to record, so no business writes them down about itself. Whether nine other paper mills within delivery range make the same weight and finish is not a line in anybody's accounts. Which of two names a head teacher has come across is not a line in anybody's accounts either. Both are perfectly ordinary facts. Both are things somebody knows. Printing either one was nobody's job, so neither was ever going to be printed. In these notes a fact of that kind is called a field factSomething true about a trade that no single business publishes about itself. Recording it was never any one business's job., and the only route to one runs through a person.
The one property this act carries, and it carries it alone, is that the answer exists nowhere at all until somebody writes it down. Nothing was waiting to be found. There was no document with the figure in it that went unlocated. The asking made the record, and without the asking there would be no record. The absence of any prior record turns into the sharpest practical difference between the two routes. The reach of a conversation, and why the useful question is nearly always an easy one for the other person to answer, is covered separately, and the closing table names where.
What is it to work from a record somebody else made?
Somebody inside a business wrote the thing up. Somebody else went over it. The record was put out where anybody who cares to look can reach it, and it is reached by opening it. On this route the whole view of that business runs through what the record chose to set down. The record settles what it recorded, exactly, with no argument and no ambiguity, and it settles nothing beyond that, however much a reader would like it to.
The first half of that is strong and the second half is firm. A published set of accounts was asked to present what a business spent on its people over a year, so it gives that figure to the rupee. What the maker down the road charges was never a transaction of the business writing, so the accounts do not give it. Rereading it for an hour adds nothing, in the same way that an hour with a household electricity bill leaves a reader knowing exactly what was known at the start. The document is not being coy. The figure was never in it.
The one property this act carries, and it is the exact opposite of the other's, is that a second reader can open the same thing and land on the same figure. The record sits still. The record does not remember its reader, does not change its mind, and gives the next person what it gave the first. Sitting still is the whole of its advantage, and an enormous one. Each kind of published document, and how far inside one a given figure sits, is covered separately, and the closing table names where.
1. Working from a record somebody else made carries one property that asking somebody never carries. Which one?
The same word names a document here and an act there. How can both be right?
Somewhere else in these notes the word primary is doing an entirely different job. There it describes a document, and it means two things about that document at once: that it is the original record of the thing rather than somebody's description of the thing, and that it was made by the accountable partyThe person or business answerable for the thing being recorded, as opposed to anybody merely reporting on it afterwards., the party answerable for what it says. On that reading a document is primary for precisely the fact it records and no further. The rule is a rule about paper.
Here the word is doing something else. Here it describes an act. Somebody went and asked, and wrote down the answer. Asking is a rule about behaviour, and on this route there very often is no paper until somebody makes some, so the rule says nothing at all about paper.
Neither use is wrong, and nothing in the word itself settles which of the two a sentence means. That is the awkward part. A sentence carrying the word looks equally sensible under either reading, so a reader supplies whichever one they happen to hold, and two people can agree word for word while meaning entirely different things. The check worth carrying away is a small one and it takes a second: whether the sentence is about a sheet of paper or about something somebody did. If the writer cannot tell, neither can the reader.
2. A colleague writes that the work included first-hand research, and what they did was read the written record of a meeting a business held. What has gone wrong in that sentence?
What happens when both questions are asked at once?
Once two separate questions are seen hiding under one word, the sensible move is obvious: stop choosing between them and ask both. The two questions are short. Who brought this into existence, the analyst or somebody else? And what is it, the original record of the thing or somebody's account of that record?
Two questions with two answers each produce four combinations, and a classification is only honest if it carries as many cells as its tests actually produce. So four cells get drawn, and none of them is left empty for tidiness or shaded away as an odd case. Every one of the four turns out to be populated by items an analyst handles all the time.
Take them in turn. The analyst produced it and it is the original record: an answer from a head teacher about which of two makers she has come across, or a quotation slipThe written note of a price a supplier stated when asked, usually good for a short while rather than indefinitely. carrying what a paper mill said it would charge, written down after somebody rang and asked. Both are first hand, and neither leaves a document that a second reader can go and open. Somebody else produced it and it is the original record: an annual reportThe yearly account a business publishes of itself, carrying its audited figures alongside its own written description of the year., or a call transcriptThe written record of what was said on a call a business held, taken from the call itself rather than written up afterwards by a listener.. Both sit as close to the thing itself as a document ever sits, and both were reached entirely by working from something somebody else put out. The analyst produced it and it is an account of the record: the analyst's own note summarising what was read or what was told. Somebody else produced it and it is an account of the record: a broker note, a news report. Second hand on both counts.
The two facts that opened this guide now have addresses. The line from the written record of the meeting sits in the top right. The head teacher's answer sits in the top left. The two facts share a row and sit at opposite ends of it, and the opening question refused to settle for exactly that reason.
3. Apply the two tests to a slip carrying a paper price obtained by ringing a mill. Where does it land?
What falls out of the grid once every cell is filled?
Look at the top row and something awkward becomes visible that was invisible while the two ideas were being discussed one at a time. A call transcript is the original record, made by the party answerable for what it says, and it was reached by opening something that somebody else published and left lying about for anybody to read. The transcript is top rank on the paper question and thoroughly second hand on the act question, at the same moment, with no tension between those two readings whatever.
An answer from a head teacher is the mirror image. On the act question it is as first hand as anything ever gets: somebody asked, she answered, the answer was written down. On the paper question there is no paper, so the answer does not score at all. Nothing exists for a second reader to go and open, and no amount of care will make one exist.
So the two ideas cross rather than nest: neither is a subset of the other, and a single item can sit at the top of one and off the bottom of the other at the same time. That is the one sentence worth carrying away. The crossing is what makes the opening question unanswerable as asked, and what makes the two words dangerous when they are used loosely. A reader who believes one idea sits inside the other will read a claim about paper as a claim about work, or the reverse, and be completely confident either way.
4. The panel below opens on a call transcript. The control then moves to an answer from a head teacher. What will the marker on the grid do?
Move an item through the grid and watch what a later reader would need
Six items, one control. The panel puts each one through the same two questions, drops it into the cell those answers produce, and redraws the record a later reader would have to be given in order to get back to it. None of the four cells ranks above another. The panel opens on a call transcript, the item most people file in the wrong place.
A call transcript
Held fixed at every position: the panel sorts and does not rank, so no cell is better than another and no item is scored.
Educational illustration. Every item shown is either published in these notes for invented businesses or is a named kind of document. A landing in the top left means the fact is first hand and leaves nothing a second reader can open. The emptiness is a property of the route rather than a weakness in it.
5. An analyst has two facts to write up. One came out of a published document and one came out of a conversation. What goes beside each?
So what can each of them actually settle?
The grid sorts items. The grid does not yet settle what to do with any of them. For that, one question goes to both routes, and the two answers come back very differently. The question is the plainest test there is for whether a run of work can be stood up a second time: can a second reader get back to the same input?
Put it to a fact taken off a record. The record settles what it recorded, and a second reader goes to the same document, turns to the same place and finds the same figure sitting there. The answer to the test is yes, cleanly and completely, and this is what the whole apparatus of referencing exists to make possible.
Put to a fact taken off a conversation, the same question answers otherwise. The fact settles what the person asked could see from where they stand, and that is often the only thing anybody can see about it. And there is no document to go back to, so a second reader cannot get back to it. The answer is not a failure and it is not a shrug. A second reader can instead reach the same person and ask again, a different answer rather than no answer at all.
The asymmetry between the two answers produces every practical difference between the routes. The asymmetry decides what has to be written down, how much of the work survives being handed on, and what it costs to have been wrong. Rebuilding somebody else's figure out of the inputs they used is a discipline in its own right, covered elsewhere in these notes, and only that single test is borrowed from it.
What gets written beside each line?
Here the whole subject turns into something done on a Tuesday afternoon. Beside a fact that came off a record, write a citationA reference precise enough that a second reader can open the same thing and land on the same figure, rather than merely on the same document. exact enough that somebody can open the same thing and land on the same figure, not merely on the same document. Beside a fact that came out of a conversation, write who said it and when, precisely enough that somebody could reach that same person and put the question again.
The common ground between those two records is easy to miss under how different they look. Look at it anyway. Both of them are the route written down. Both of them give somebody picking the work up afterwards a way back, and a way back is the one thing that counts. The two records look nothing alike because the ways back to them are nothing alike, and a reader who believes only one of the two counts as a source will write one of them and leave the other blank.
Writing the route beside the figure is the whole of the change asked for, small and with a long reach. Both of those records amount, taken together, to provenanceThe record of where something came from and how it reached the analyst, kept alongside the thing itself rather than in somebody's memory.: the trail from the figure back to wherever it came from, kept in writing rather than in somebody's head. Grading information by how sensitive it is, and keeping hold of where each fact came from, is a subject in its own right, and the table below names where it is covered.
What does this look like when somebody is actually working?
Take a household job first. Everybody has done the household version. Somebody is collecting quotations for a wedding. The caterer hands over a printed rate card, and it goes into a drawer; six months later it comes out with the same rates still printed on it. A cousin also says, over the telephone, what she paid for the same number of guests last winter. The cousin's figure is what somebody actually paid rather than what somebody is asking, so it is worth as much as the rate card and possibly more. But written on the back of an envelope with no name and no month against it, in six months it has quietly become a rumour, and nobody can work out whether it was the cousin or the neighbour who said it, or which winter she meant.
Somebody assembling a working file on a business runs the identical problem at a larger scale, with the same fix. Every figure that came off a published document goes into the file with a reference against it. Every fact that came out of a conversation goes into the file with the person and the day against it. When the file gets picked up nine months later by somebody who was not there, the second sort is exactly as checkable as the first, and no other test matters. The half of a file that carries no route is not a smaller half. The bare half is the half that stops being evidence.
The note where half the facts carry a reference and half carry nothing, and nobody decided to do it
The person who does this is not careless. The person is careful, and follows the only convention anybody ever taught them: a source goes in a footnote. Watch what that convention does to work drawn from both routes at once.
There is an obvious place to point, so every figure that came off a document gets a reference, precise and checkable. A conversation offers nothing to point at, and the convention has no other slot to put anything in, so every fact that came out of one gets nothing at all. Nobody made a decision. The format made it for them. In the finished note the two halves of the work now look completely unlike each other: one half appears sourced and the other half appears merely asserted, and a reader has no way of telling that the second half was obtained rather than assumed.
The cost of that is exact rather than vague. On the day those field facts were obtained, every one of them was re-askable. Somebody knew the mill, the head teacher and the week. The name and the date were never written anywhere, so six months later not one of them is. A fact that could have been re-opened for the price of one telephone call has been made permanently unopenable by a formatting habit. Nothing was invented. Nothing was wrong. The work is now unrepeatable in precisely the half where repeating it would have been cheapest.
The second failure is the mirror of it, and it is a failure of the word rather than of the format. Somebody reads that a call transcript ranks near the top among documents, fastens that ranking to the act, and writes that the work included first-hand research. It did not. Somebody sat and read a document that a business had published. Reading one is an entirely respectable thing to have done, and it is not what the sentence says. A reader taking the sentence at face value believes a person was asked something, and no person was asked anything. Both failures come off the same root: two different properties wearing one word, and a reader ranking the two routes instead of recording each one.
What does neither of them settle?
Setting two routes against each other invites a comfortable conclusion: between them they cover the ground. They do not.
A ledger for the register trade sits elsewhere in these notes, built by running the field through in a fixed order, one line at a time. Five lines went in. Two of them came back carrying a fact. Three of them came back carrying the words not establishedThe written outcome for a line that somebody actually chased and could not close, as against a line nobody reached., written out in full and at the same size as everything else on the sheet. A blank left blank reads as an oversight, and a blank written out reads as a result.
The two that stand rest on published figures. On the buying side, one named school group takes a large share of the year's sales, holds an even larger share of the money still owed, and pays a good deal slower than everybody else does. On the supply side of it, everything awkward gathers at one small counterparty: a machine servicing contractor that takes 4.0 per cent of what goes out to parties outside the business. Exactly one alternative to it is known of, and swapping it over runs to twenty six weeks. Set that beside the paper mill. The mill takes 46.0 per cent, comes with nine alternatives, and delivers inside two weeks. Both readings came off records somebody else made.
Now look at the three that did not stand, and at what each route could have done about them. No published fact describes how the two makers conduct themselves when they meet in the same district. The cost of building or buying a register works has never been put in print, and no maker has been recorded arriving or leaving. And nothing published anywhere names a thing beyond the register trade that a school would put that money into instead. Neither route settles any of those three, and saying so plainly is the honest floor of this whole comparison. Those lines are not an unfinished run. The three lines are the result of the run, and each can be filled the day somebody gets the fact. A line quietly stuffed with an average can never be filled that way.
6. In the published ledger for this trade, three of the five lines are written out as not established after both routes have been considered. What does that show?
Why does the cheap route quietly become the serious one?
The least flattering part of the subject explains far more bad work than carelessness ever will. The two routes do not cost the same, and the gap between them is enormous. One fact is already sitting in a file that opens in the next ten seconds and asks nothing but attention. The other requires working out who would know, finding a way of reaching them, catching them at a moment when they are willing to talk, and then writing down carefully what came back.
Now watch what that gap in effort turns into. Nobody ever sat down and decided that the cheaper route was the more serious one, and yet that is the ranking the effort quietly produces, in both directions at once. The route that is always within reach starts to look like the solid one purely because it is always within reach. The route that costs something gets left out of the write-up altogether, or, worse, gets filled in with a plausible figure that nobody actually supplied. Both of those are effort dressed up as judgement.
The same pattern is familiar from ordinary life, where it has nothing to do with evidence at all. The shop at the end of the lane is not better than the market two bus rides away. The shop is nearer. Without once asking which of those two things is doing the work in a decision about where to buy rice, a person ends up with a firm opinion about quality that was actually an opinion about distance. Neither of the two research routes sits above the other, and nothing about what either one costs to get changes that.
7. A field section gets written from the desk far more often than the accounts work does. What is the best explanation for that pattern?
Does any of this change depending on where it is read?
Which cell an item falls into, and what a later reader needs in order to get back to it, behave identically everywhere. Neither route turns on a currency, a legal form or a place. The register maker carries the words Private Limited after its name and the head teachers work in a district, and neither of those details is holding up a single claim made here.
The one thing that genuinely varies from place to place is what gets published and when it appears, and that varies too much for any general statement to hold. A reader comparing two ways of getting a fact reaches for a calendar almost by reflex, and no calendar can be given. Without a calendar this much holds: one route waits on somebody else publishing and the other waits on somebody answering, so the two arrive on different clocks and neither clock is under the analyst's control. When any particular document becomes available is read at the source on the day it is needed, and the day of reading is written down beside whatever is taken from it. The rule stated above about every other fact applies here to itself.
What is worth carrying out of all of this?
Two things.
The first is one line, repeated word for word rather than reworded. The wording is the point. Write the route beside the figure, or the figure is not finished. Every reading of how a business works, every drawing of the field around it and every judgement about what its numbers mean depends on somebody having recorded where each fact came from. A figure carrying its route can be questioned, disputed and put right a year later. A year from now nobody will recall that a figure without one was made up, so it cannot be put right at all.
The second is a matter of what a demonstration can carry. Anjani Stationers Private Limited and Bhavani Register Works were made up so that a method could be shown on something small, along with the schools, the mills, the registers and the district. A method shown on a made-up case travels out of the case, and not one of its numbers does.
Which part of all this was the method and which part the furniture is worth separating. The two tests, the four cells, the question about whether a second reader can get back to the input, and the ledger with three lines written out rather than filled in: those are the method, and they hold for any business anywhere, of any size, in any trade. The register maker, the second maker and the school group are the furniture. The three were put there to show the method working on something small enough to hold in mind all at once. The distinction between method and furniture holds wherever a method is taught on an invented case.
8. Every business used across these notes was made up for teaching. What follows for a reader who has just finished this guide?
Where this guide stops. Its subject is setting the two ways of getting a fact side by side, working out what each can settle, and deciding what gets written beside each. Ranking documents against one another, working either route at length, and valuing or rating anything are separate subjects. Fifteen more questions that bring somebody to this subject are settled elsewhere in these notes, and the table names where.
| What a reader may have come for | Where it is covered |
|---|---|
| Putting one kind of evidence in order against another, from a published document down to somebody's passing commentary | Source Hierarchy: Ranking Evidence From Filing to Commentary |
| What makes a record the original one, and who is answerable for what it says | Primary Source: What Counts as One and Why It Ranks First |
| What a published set of accounts gives a reader, document by document, and how far inside one a figure sits | Company Filings as a Research Source: What Each Document Offers |
| What a conversation reaches, what it will never reach, and how to put the question | Primary Research: Talking to People Who Actually Know |
| Reading a yearly account properly, and the order worth reading it in | Annual Report: What Is Inside It and How to Read It in Order |
| What the written record of a call is actually evidence of | Earnings Call: Structure, Signals and What the Transcript Reveals |
| How a business's own slide deck differs from what it lodges | Investor Presentation: What It Is For and How It Differs From the Filing |
| Rebuilding a number from the inputs somebody else used, so the work can be stood up again | Audit Trail: Making Analytical Work Reproducible |
| Framing a question so that it can honestly come back either way | Research Question: How to Frame One That Can Actually Be Answered |
| Recording what was assumed, separately from what was found | Assumption Register: Recording What Was Assumed and Why |
| Drawing the field a business sells into, before judging where it stands in it | Competitor Mapping: Drawing the Field Before Judging Position |
| Building a disagreement with a reading somebody has written down | The Variant View: Disagreeing With Consensus, With Reasons |
| Understanding how a business works before any price is put on it | Business Analysis: Understanding a Company Before Valuing It |
| Where a question would put the person answering it in difficulty | Research Independence: The Conflicts and the Safeguards |
| Grading information by how sensitive it is, and keeping hold of where it came from | Handling Information: Classification by Sensitivity, and Provenance |
Where were the borrowed items first written down?
Three rows below and not an institution among them. Nobody publishes a rule about how asking somebody compares with reading something, so there is no authority to point at. The rows carry instead the place each borrowed item was first written down, and anybody who wants to argue with a cell of the grid can go and read the material that cell was populated from.
| Named for | Where that was read | Site |
|---|---|---|
| The two ways a fact reaches an analyst at all, the field facts and the route written against each, and the observation that the two do not cost the same to obtain. Two of those facts are lifted into the grid above as items to sort. | These notes, on setting one business against the field it sells into | finmaverick.com |
| The five-line ledger reproduced above, both lines that stand and all three written out as not established. It carries the count of five, the reading on the buying side, the terms sitting at one small counterparty, and the statement that nothing published settles the other three by either route. | These notes, on working a field through in a fixed order | finmaverick.com |
| The single clause saying that a figure can sit at one of several distances inside a published document. The working of it is covered under the source named beside it. | These notes, on what each company document offers a reader | finmaverick.com |
Anjani Stationers Private Limited and Bhavani Register Works are invented.
Educational material. Not advice on any investment, tax, budget or market position.