Primary Research: Talking to People Who Actually Know
Primary research is asking somebody a question about their own dealings and recording what they said and who said it. Asking reaches facts about a field that no statement holds. No such fact was ever anybody's transaction to record. Four of them are published below, each with the route beside it, and the useful question is nearly always one the other person can answer without cost to themselves.
Why will an hour with a household's own electricity bill never say what the flat upstairs pays?
Consider the last electricity bill that arrived at a flat. The bill prints the units got through this month. The bill prints the rate they were charged at. The bill prints what is owed and the day it falls due. The bill usually prints the month before as well, in a small box put there for comparison. An hour of attention does not add a line to a document, so an hour spent with that sheet ends exactly where it began. Whatever that bill settles is already settled on its face.
Now change the question by a few words. What does the flat upstairs pay? Reading the sheet until midnight will not produce that figure, and the reason has nothing to do with the bill being a poor document. Nobody left the figure out. The figure was never in. The meter above the ceiling belongs to a different household, the payment left an account belonging to somebody else, and the paper in hand records one household's own dealings and nothing else whatever. The payment upstairs was never a transaction of that household, so none of its documents went anywhere near it. One route reaches the answer, and it runs up a flight of stairs and ends at a door.
The staircase matters more than the bill does. There is an entire class of fact that no amount of careful reading will produce, and carelessness has nothing to do with it. Such facts were never anybody's transaction to write down, so no document anywhere recorded them, and the only way to reach one is through a person who is in a position to say it. The route up the staircase reaches some of them, will never reach others however hard anybody works at it, and leaves something to be written down whenever it brings a fact back.
1. A household's own electricity bill prints what it used, the rate it was charged, what it owes and what it used last month. The question is what the flat upstairs pays. How much of that can careful rereading produce?
Where do the two kinds of fact actually come from?
Set the two side by side as physical objects and the difference stops sliding about. The first kind of fact arrives as a document: written up inside a business, gone over by somebody there, then put out for whoever wants it. Opening it is the whole of the work. The second kind arrives because somebody put a question to a person who draws no salary from that business, and that person answered. A price read out over a telephone. A rate card handed over a counter. A head teacher saying which of two names she knows and which she does not.
Neither of these two routes is more respectable than the other, and neither is a substitute for it. What a fact is worth is exactly what its route is worth, no more and no less. A fact carries the authority of the route it came down, and that is worth stating flatly rather than arguing over. The route goes beside the figure, or the figure is not finished. A figure with its route written beside it can be weighed, argued with and corrected next year. A figure with no route cannot be corrected at all. In a year nobody will remember where it came from.
Both routes already run at home without anybody thinking about them. A household paid to have a leaking tap seen to last month and the receipt is in a drawer, so it knows the cost to the rupee. The price the building next door pays for the same job is written nowhere within reach, and no amount of going through that drawer will produce it. The watchman knows. The neighbour who had it done in March knows. Neither of them is a lesser source than the receipt, and neither of them is a better one. The watchman and the neighbour are simply the only ones who hold that particular fact.
2. Facts a business publishes about itself, and facts about a field that no one business produces. How do the two compare in standing?
Two nearby words, two different things: which one is at issue here?
A reader may well arrive here holding a word that looks almost identical to the one in the title, met somewhere else in these notes and firmly attached to documents. The collision is real and it is worth thirty seconds rather than a chapter. One of those words describes a property of a document: the original record, made by the party accountable for making it. The other describes a property of an act: the asking of a question and the writing down of what came back. The act is the subject here, and nothing else.
The two are not a pair of words for one idea, and they do not nest one inside the other. A set of accounts, or the text of a call put out afterwards, is an original record reached by reading something somebody else produced. A head teacher's answer is first-hand and leaves behind nothing a second reader can go and open. The first-hand answer that leaves nothing behind to open is the harder of the two cases, and it is why the note written afterwards has to carry something other than a reference to a document. Original records, and who is accountable for making them, are covered separately under Primary Source: What Counts as One and Why It Ranks First. Ranking one kind of document against another, and sorting the two acts fully against each other cell by cell, are each covered separately, and the table at the foot names them. The act is what this guide works on.
What does a field fact look like once it is written down properly?
Four such facts sit in these notes already, about the surroundings of Anjani Stationers Private LimitedA company form in India whose shares are not offered to the public, so a small number of people hold the business rather than it being traded., an invented business used throughout to make a school register. The table sets all four out word for word as they were written where they were published, together with the route each one came down, and that repetition is deliberate rather than lazy: an argument that the route is part of the fact would look ridiculous quietly rephrasing either half of somebody else's carefully worded line.
| The field fact, reproduced as published | The route it came down |
|---|---|
| Nine more mills close enough to deliver turn out paper of the same weight and the same finish, and any one of them will quoteA price a seller states for a stated thing, usually put in writing, and usually good only for a short while rather than indefinitely. inside a day | a quotation slip, got by ringing the mills |
| A first load from a mill not used before takes about two weeks to arrive | a delivery promise, got by asking the mill |
| A second maker of the same registers to the same specificationThe list a buyer writes out saying exactly what they want made, close enough that two different makers working from it turn out the same article. exists, Bhavani Register Works, promoter runRun by the people who started it and still hold it, rather than by hired managers answering to shareholders from outside., turning out 1,50,000 registers a year against Anjani Stationers' 2,50,000 | knowing the trade, in the way anybody who sells into a district knows who else sells into it |
| The name over Anjani Stationers' door is known to head teachers across the district and Bhavani Register Works' name is not | asking head teachers |
One sentence travels with that set wherever it is shown, and it too is reproduced rather than reworded: not one of those four appears in any statement, and every one of them is a fact. The sentence is two claims fastened together, and it is going to do a great deal of work in a moment. The route is doing something in each line as well. The route is not a footnote and it is not a courtesy. The route is the part of the fact that lets a later reader decide how much weight the line will carry, and it is the part that lets that reader go and get the same fact again for themselves. Take the routes away and what is left is four assertions of unknown parentage, all of which look exactly as confident as each other.
3. A note records that a mill nobody has dealt with before takes about two weeks over a first load. What has to sit beside that line before it is finished?
What breaks if either half of that sentence is dropped?
Hold both halves of that sentence at once: the half saying no statement carries these four, and the half saying all four are facts anyway. The trouble in this subject nearly always begins with somebody quietly letting go of one of them, and which half goes decides which way the work fails.
Dropping the first half sends the search into the wrong building. A count of mills was never a transaction of the business whose accounts are in hand, so it was never written down there. A reader opens the notes to the accountsThe supporting sheets that follow the main statements, where a single total is opened up into its parts and each part is explained. hunting for that count, reads them properly, and loses a morning to it. The reading is not careless. The reading is simply happening in a place where the answer has never been.
Dropping the second half costs more. Everything living outside a statement starts to look soft, or unserious, or optional. The four facts get hedged into vagueness or dropped out of the field sectionThe part of a written analysis dealing with the trade a business sells into, as against the part dealing with its own accounts. altogether, and what has been thrown away is precisely the material nothing else could have supplied. A hollowing out performed by somebody being careful is the hardest of all to spot and the hardest to argue with afterwards.
Both mistakes come from one root. The root is ranking the two routes against each other instead of simply writing each fact down with its route attached. Once the route is written down beside the line, nobody has to rank anything: a later reader looks at the line, sees where it came from, and decides for themselves what it will bear.
4. A reader accepts that none of the four field facts appears in any statement, but quietly drops the second half of the sentence, that every one of them is a fact. What is that reader likely to do next?
What do the four routes turn out to have in common?
Go back to the four and take the facts away, leaving only the asking. A mill was asked what it charges for its own paper. A mill was asked how soon it could get a first load out of its own door. Somebody who sells into the district was asked about the trade they sell into. A head teacher was asked which name over a door she recognises. Set out like that, one thing about the set becomes obvious, and it was not obvious at all while the facts were still attached.
In every case the person was asked about their own dealings, and in every case they were free to answer. Nobody was asked to hand over anything that was not theirs to hand over. Nobody was put in an awkward position by the question. The free answer is not a restriction placed on the work from outside and dressed up as good practice. Freedom to answer is simply what the four routes turn out to have in common on a proper look, and it is also most of the reason they worked at all. A question somebody can answer freely gets answered quickly, gets answered accurately because there is no reason to shade it, and can be asked again a year later by somebody else with the same telephone number.
The useful question is nearly always one the other person can answer without cost to themselves. The useful question is a finding about what works rather than a caution about what to avoid. A free answer is the reason a mill quotes inside a day: quoting is what a mill does all week, the answer costs it nothing, and it would rather be rung than have the caller ring somebody else.
5. Look at the four published routes together. What do they turn out to have in common?
What is this kind of asking for, and what are its three steps?
Strip the four routes back to their shape and there is very little left to learn. Ask a party to a tradeSomebody who buys, sells or supplies in it, and who can therefore speak about their own dealings in it. about their own dealings. Ask about something that party is free to state. Record what was said, who said it and when. Three steps, and no more than three.
What is the asking actually for? Working along a selling line and asking each party about their own dealings goes by different names in different trades, and a channel check is one of them. The name matters far less than the purpose, and the purpose is narrow and worth stating narrowly. The check establishes what each party to a trade can see from where they stand, and it establishes nothing else. A mill can see its own price and its own despatch. Somebody selling into a district can see who else sells into it. A head teacher can see which names reach her desk when it is time to place an order. Put those together and the result is a picture of a trade assembled out of first-hand pieces, every one of them attributable to a person who was in a position to say it.
And here is the line, a hard line and not a matter of tone. Everything described here is a question somebody is free to answer. Where a question would instead put the person answering in difficulty, that ground carries its own rules; those rules turn on the place and on the house the work is done for; and none of them is stated here. The block immediately below says what to do before going anywhere near it, and the table further down names where those rules are taught in these notes.
What travels across a border here, and what does not?
Asking a party to a trade about their own dealings, and writing down who answered and on what day, behaves identically everywhere anybody sells anything. Nothing in the three steps above turns on the place. India supplies the currency, the digit grouping, the legal form Private Limited, the district the head teachers work in and the school year that sets the buying rhythm behind these invented businesses, and not one of those carries any weight in the argument. In any country the three steps read the same.
Rules do change with the place, and with the house the work is done for: rules about who may be approached, what may be asked of somebody who works inside a business, and what a person is required to do with an answer once they are holding it. A half-stated rule is more dangerous than no rule at all, so those rules are read from the source that sets them, on the day, by anybody whose work looks like going near that line, and the date goes beside whatever is taken away. The table lower down names where that ground is actually taught in these notes.
What can a conversation settle, and what will it never settle?
A conversation settles what the person asked can see from where they stand. Their own price. Their own promise. The trade they are in every day. The names that reach their desk. The list is real and useful, and it is also complete. A conversation settles nothing that no single person is placed to see, and putting the same question to more people does not change that. Adding up a hundred views from ground level never produces a view from above.
The contrast works on the material already given. Ringing the mills settled how many mills within reach turn out that weight and finish, and settled how quickly a first load arrives, the delay the trade calls the lead timeHow long a first delivery takes to arrive after it has been ordered, counted from the order rather than from despatch.. Both of those worked because a mill can see its own trade and its own despatch dock. Now a different question, put to the same district. How many makers in this city turn out that register to that specification? Nothing written anywhere states the number. One further maker is named in these notes, and one further maker sets a floor rather than giving a count. Nobody was asked the question, and more to the point there is no single person standing anywhere they could answer it.
So that line stays exactly as it is, and it reads not established. Not established is the written outcome of going to look for a fact that would meet the condition and coming back without one. Not established is a different thing from a hedge and a different thing from a gap in the work. A hedge softens a claim still quietly being made. Not established withdraws the claim and says why it was withdrawn. Not established is a line anybody can fill the moment somebody gets the fact, and until then it tells a later reader precisely which facts the note does not hold. A conversation is not a substitute for a census.
6. Ringing the mills settled how many mills close enough to deliver turn out that paper. A reader reasons that the same approach will settle how many register makers the city holds. What is different about the second question?
There is no document to open, so what goes in the note?
A citationA pointer telling a later reader exactly where to open something for themselves: the document, and the place inside it where the figure sits. works because a second reader can go to the same document, turn to the same place, and find the same figure sitting there waiting. Here there is nothing of the kind. Nothing was published, nothing was lodged, and the fact exists in one person's knowledge of their own dealings and in the note made of what they said. Half the reason field facts get treated as second class is this and only this: there is no printed reference to put after them, and a line with no reference after it looks naked beside one that has.
So ask the honest question instead: what could a later reader actually re-open? Not a document. There is not one. A person, and a moment. A later reader can ring the same mill. A later reader can ask the same head teacher. The note carries, in place of a reference, who answered and when they answered, and those two items do exactly the work a printed reference would have done. The general discipline of leaving work behind so that a second person can rebuild it is covered separately under Audit Trail: Making Analytical Work Reproducible, and the only thing added here is that one specific pair, for the one kind of fact with nothing behind it to open.
The difference shows on a street corner. A note reading that the stall outside one particular office building sells out by two o'clock is checkable next Tuesday by anybody willing to walk past at two. A note reading that stalls generally sell out by the middle of the afternoon is checkable by nobody, ever. The second sentence is the one that sounds more authoritative, and it is the one that can never be corrected. There is nobody it belongs to.
7. A fact has come from a conversation and there is no document anybody can open. What goes in the note?
Why do good field sections get written badly?
Here is the least flattering fact about this whole subject, and it explains far more of the bad work than carelessness ever will. A field fact is bought with a telephone call. An accounts fact is bought with a download. The gap in price explains more bad field writing than any failure of care ever will, and it has almost nothing to do with anybody's opinion about which kind of evidence is worth more.
Name the moment precisely. A reader who has ever written anything under a deadline will recognise it immediately. The clock says seven in the evening. Three days of accounts work has gone well and is sitting finished. The accounts are still open on the screen, so anything they can answer is four seconds away. One section is left, it needs a fact nobody has published, and the call that would get it takes ten minutes and cannot be made until the morning. Adding a day to work that is otherwise complete feels absurd at that hour.
So notice the two things the moment is not. The moment is not laziness and it is not indifference. The moment arrives when the work is nearly finished, and that is precisely when a person is least willing to reopen it. The person standing there has already done everything else properly. The remedy cannot be more care. The care is already there. The remedy is knowing that the moment arrives, seeing it coming, and knowing in advance what to write when it does. The panel below shows what each of those calls actually bought.
8. The panel below starts with all four routes held and gives up every one of them in turn, down to none. What happens to the block standing for the accounts drawn beside the four cards?
Give the four routes up, one at a time, and watch what the field section can still say
The panel opens holding everything these notes hold: all four routes, all four lines filled. Moving the control down gives one route up, then another. The panel only ever takes a route away, and no setting adds one.
Educational illustration. All four facts and all four routes are published in these notes for invented businesses, and the panel generates nothing at any setting. Giving up a route removes what that route obtained and nothing else, so no other line changes and the accounts do not change at all. The order of removal is the order these notes print the four in, and is not a sequence anybody worked through. A line reading not established means nobody asked and nobody holds it, and that is a result rather than a gap in the work. No count of makers, no field total and no price for the second maker exists at any setting, and none is computed here.
What gets written when nobody has the fact?
Sooner or later a proper question put to the field gets nothing back, and this is the part of the method most treatments leave out. The instruction runs to three lines, and each of the three is doing a job of its own rather than repeating the one before it.
| The line that gets written | What that line is doing |
|---|---|
| Write the question down | Fixes exactly what was asked, so a later reader is not left guessing at the shape of it |
| Write what would settle it | Names the fact that would close the line, which is what makes the line findable next year |
| Write that nobody has it | Records the outcome, so the blank reads as something somebody went and checked |
The three lines are the whole of it, and they take about as long as making something up would have taken. Then comes the asymmetry that turns three lines from a preference into an argument. A line reading not established gets filled the day somebody finally obtains the fact, while a line already carrying a trade average nobody published never gets corrected at all. The reason it never gets corrected is not that anybody objects to fixing it. The reason is that in a year nobody will remember the line was made up. By then it is just another sentence in a document, set in the same typeface as the ones beside it and quoted by people who were nowhere near the room it was made up in. The phrase an invention wearing the clothes of research is pointing at exactly that, and the clothes are the whole problem.
So the closing claim here is a small one and it is worth stating plainly. Four telephone calls, or something very like four telephone calls, produced four facts that no statement anywhere holds. Every one of them travels with the route it came down. Every one of them can be put to the same mill or the same head teacher again by anybody holding the number, and if the answer has changed since, the change is itself worth knowing. Four re-askable lines are what the act buys, and they are why the act is worth the ten minutes at seven in the evening.
Who actually runs this, and what does it buy them?
Take a lender looking at a working line against a small maker's stock. The file arrives complete and costs nothing: what was bought, what is owed, what is sitting in the shed and what it was carried in at. Now look at the questions that actually decide the lending. If the mill currently supplying the paper stopped next month, how long before a first load arrives from somewhere else, and is there a somewhere else? Do the people who place the orders across that district recognise the name over the door, or is it the other name they reach for? Not one of those sits in the file, and every one of them is answerable by asking a party to the trade about their own dealings. A lender who reads only the file has read the half of the position that somebody happened to write down.
The same shape runs at household scale, and seeing it makes the act ordinary rather than professional. Somebody choosing between two schools reads both prospectuses and learns exactly what each school chose to print about itself. How quickly either one answers its telephone in July is a fact of precisely the kind at issue here: first-hand, obtainable in ten minutes at the gate, written down nowhere at all, and lost the moment nobody notes who said it and on what day.
The field section written at the desk, and it is the careful reader who writes it
Who commits this is not who anybody would guess. Somebody has spent three days on a set of accounts and done it properly. The reconciliations close. The margins are worked and stated. One section is left. Seven in the evening has come round and the accounts are still open on the screen.
The section needs a line on what the second maker charges. There is no published figure and there is no call anybody has made. The call is ten minutes and cannot happen until the morning. So the line goes in reading broadly in line with what makers of this size charge, and the whole decision takes about forty seconds. Notice that it is not a lie. Nobody has asserted a number. Nobody has invented a price. The line is simply not checkable, and it never will be.
Here is what it costs, and the cost is specific rather than general. The hedged sentence now sits alongside four facts that were genuinely obtained, in the same typeface, at the same weight, in the same hand as the rest. Nothing in the note tells them apart. Four of the five lines can be re-asked by anybody holding the number. The fifth cannot be checked by anybody at all. Nobody was rung, so there is nobody to ring. An invention wearing the clothes of research means this in practice: the problem is not the invention, and somebody would catch that. The problem is the clothes.
And there is a mirror failure, discussed far less and about as expensive. The same person on a different evening does the opposite. The same person holds four perfectly good field facts, obtained properly, each with its route beside it. Then they notice that not one of the four appears in a statement, decide the section looks unsupported next to the accounts work, and either drop the four or bury them under so much hedging that nothing survives. Four facts nothing else could have supplied get thrown away by somebody being careful, and the note that results is thin in exactly the place notes are usually thin. Both failures come out of one mistake: ranking the two routes against each other instead of writing each fact down.
Where this guide stops. Its subject is asking somebody a question about their own dealings, what such a question can settle, and how to write the answer down when there is no document behind it. The thirteen things a reader is most likely to want next are covered separately, and the table below pairs each one with where it is held.
| The question a reader arrives with | Read instead |
|---|---|
| What makes a document an original record, and who the accountable party for it is | Primary Source: What Counts as One and Why It Ranks First |
| How one kind of document ranks against another as evidence | Source Hierarchy: Ranking Evidence From Filing to Commentary |
| What the text of a call put out afterwards is evidence of | Earnings Call: Structure, Signals and What the Transcript Reveals |
| Leaving work behind so that a second person can rebuild it from the same inputs | Audit Trail: Making Analytical Work Reproducible |
| Framing a question so that it can be answered either way rather than only one way | Research Question: How to Frame One That Can Actually Be Answered |
| Guarding against hearing only the answers that agree | Confirmation Bias in Financial Analysis: How It Shows Up and How to Counter It |
| Where a question would put the person answering in difficulty, and the safeguards around that | Research Independence: The Conflicts and the Safeguards |
| A dealing that somebody inside a business has disclosed, and how to read it | Insider Transactions: Reading the Disclosure Carefully |
| Sorting information by how sensitive it is and where it came from | Handling Information: Classification by Sensitivity, and Provenance |
| What a published set of accounts actually gives a reader, document by document | Company Filings as a Research Source: What Each Document Offers |
| 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 else has written down | The Variant View: Disagreeing With Consensus, With Reasons |
| Sorting the two ways of getting a fact against each other, cell by cell | Primary vs Secondary Research: What Each One Can Settle |
What here can a reader check, and what was made up for teaching?
No institution is named below, and that is a finding rather than a thinness. Nobody publishes what a conversation can settle, no disclosure body has anything to say about it, and reaching for an authority to lend weight would be the precise fault of putting a borrowed name where a route belongs. The table names instead where each part of the material was written down, so a reader can go back and read it rather than take it from here.
| Named for | Where that was read | Site | Read on |
|---|---|---|---|
| That the four field facts worked through above, and the route written beside each one, were set down beforehand. | Fin Maverick, where one business is set beside the field it sells into | finmaverick.com | 27 August 2026 |
| That two of the questions left open here were already recorded open, in writing, rather than quietly filled with a plausible figure. | Fin Maverick, where a field is run in a fixed order | finmaverick.com | 27 August 2026 |
| Every name and every count printed above. Both businesses were made up so that a method could be shown working, and the paper, the registers, the district and the schools were made up along with them. No lodged document, no survey and no trade study stands behind any of it. | Fin Maverick | finmaverick.com | 27 August 2026 |
Anjani Stationers Private Limited and Bhavani Register Works are invented.
Educational material. Not advice on any investment, tax, budget or market position.