Research Question: How to Frame One That Can Actually Be Answered

A research question is answerable when it names a specific claim, a time frame, and the evidence that would settle it either way. "Is Kaveri Cold Chain a good business?" cannot be closed. "Will occupancy reach 80 per cent within twelve months on the signed contract plus spillover, and what evidence by month six would show it will not?" can. The framing decides what work gets done and what gets missed.

Work follows the question. A vague one produces vague work, weeks of it, and at the end it confirms whatever was hoped at the start. Nothing found along the way could have said no. A sharp one shrinks the work, points it at the few things that matter, and brings back an answer that can be wrong. A research questionThe single question an analysis is built to answer. Not the topic, and not the conclusion: the thing being asked, written down before the work begins. is worth writing down before anything else for that reason alone: not as a formality at the top of a report, but as the instruction that decides what the next three weeks look like. An interesting question and an answerableCapable of being closed: some evidence that could be gathered would settle it one way or the other. The opposite is a question that stays open no matter what is found. one can be told apart on sight, and the difference comes down to three parts: a claim, a time frame, and the evidence that would settle it.

What makes a research question answerable rather than merely interesting?

Start with something a household actually argues about. "Is this a good school for our daughter?" It is a real question, it matters enormously, and a month could go into it without ever being done: good for whom, at what, compared to which other school, and how would anyone know the answer was wrong? Compare: "Will she be able to reach the school by 8 in the morning on the bus route that exists today, and does the school publish its board results so we can compare them with the two others on our list?" That version is smaller and less grand, and it can be finished by Saturday. Both are questions. Only the second one can be closed.

A question is answerable when some evidence that could actually be gathered would settle it, and it is merely interesting when no evidence could. The test is not whether the question is important, or whether clever people are discussing it, or whether the analyst holds opinions about it. The test is this: describe the finding that would make the answer no. If that finding cannot be described, no amount of work will produce it, and the question will end exactly where the prior view started. Karl Popper's word for this property, in the philosophy of science, was falsifiableA claim is falsifiable when some possible observation would show it to be false. Karl Popper's test, from the philosophy of science, for whether a statement says anything at all.: a claim earns the right to be believed only if it could have been shown false and was not. Interesting questions feel like work. Answerable questions are work.

Ishaan Verma, the invented analyst whose Kaveri Cold Chain decision runs through this sequence, began in January with "is Kaveri a sound credit?" and answered it "probably". Notice that nothing he could have read in the next three weeks would have turned "probably" into "no": a weak year would have been "cyclical", a strong one "proof of demand", a competitor rumour "already priced". The question was interesting, and it was never at risk.

What are the three parts of one?

Think of a table with three legs. Take any one away and it does not wobble, it falls. An answerable question stands on three parts in exactly that way. The first is a claim: a specific statement about the world that is either true or false, not an adjective. "Occupancy reaches 80 per cent" is a claim, the thing a scientist would call a hypothesisA specific statement put forward to be tested: true or false once the evidence is in. In finance the word thesis is used the same way.; "the business is sound" is an adjective wearing a claim's clothes. The second is a time frameThe period within which the claim is meant to come true, stated as a date or a number of months. A time frame is what makes "not yet" eventually become "no".: by when. Without it, a claim that has not come true is never false, only early. The third is the evidenceThe observable readings, documents or counts that a claim would leave behind if true and fail to leave behind if false. Named in advance, so nobody can choose it after the fact. that would settle it, named before the work starts: which reading, from where, taken when, would show the claim is or is not coming true.

Claim, time frame, settling evidence: remove any one and the question cannot be closed, however hard anyone works on it. Each missing leg does its own damage. No claim, and there is nothing for evidence to bear on: the facts gathered about the company are each interesting and none of them decides anything. No time frame, and the claim can never fail, only wait: "occupancy will reach 80" is unfalsified in month twelve, month twenty four and month sixty. No settling evidence named in advance, and the analyst gets to choose the evidence after seeing it. The evidence that fits the hoped-for answer gets chosen, and the rest gets explained away. Settling evidence is the leg most often missing, and it is the leg that turns a decent question back into an interesting one.

How is a vague question narrowed without being trivialised?

Here is where people go wrong in the other direction. Told that "is Kaveri a sound credit?" is too vague, an analyst can retreat to "did Kaveri sign the pharma contract?" That is perfectly closable: the contract is on the desk, signed, dated. The answer is also worthless. Everyone already knows it, and the loan decision does not turn on it. Narrowing is not the same as shrinking. A street vendor asking "should I move my cart to the station road?" narrows well to "will the station road sell forty more plates a day within the first month, counted on the pad I already keep?", and narrows badly to "does the station road have a pavement?"

Narrowing works when each step keeps the question attached to the decision while making the evidence more specific, and it fails the moment the question becomes closable but no longer decides anything. Follow the sequence for Kaveri Cold Chain. "Is it a good business?" becomes "is it a sound credit for Rs 25,00,00,000?" The second version at least names the decision. The credit question becomes "will occupancy reach 80 per cent within twelve months?" The forecast profit before tax of Rs 4,20,00,000 needs 80 and break-even sits near 62, so occupancy is the claim the whole thesis rests on. The occupancy question becomes "will the 14 contracted points plus 4 points of spillover arrive within twelve months?" The rewrite separates the part with a signature from the part with a hope. And the spillover question becomes "what by month six would show the spillover is not coming?" The last version would have put the weakest input under a lamp. Each step is smaller than the last and each step still decides the loan. One step further, to "did the contract get signed?", falls off the edge: closable, and irrelevant.

A useful way to see the edge is to plot questions on two axes at once: how closable each is, and how much of the decision turns on it. The vague question sits high on relevance and hopeless on closability. The trivial question sits the other way round. The scopeThe width of what a question takes in. Too wide and no evidence can settle it; too narrow and the answer, though certain, decides nothing about the matter at hand. that works is the top right corner: closable, and still carrying the decision. Every narrowing step is a way of walking a question into that corner.

Why does the framing decide the work that follows?

Consider what an analyst actually does on Monday morning after being handed a question. The question is the instruction. Handed "is Kaveri a good business?", the sensible response is to read everything: the sector, the promoters, the history, the accounts for five years, the competitors, the regulations on cold storage, the pharma client, the tariff card. Three weeks later there are forty sides of typed note, every one of them relevant to the question as asked, and not one of them can say no. The question never said what no would look like. Handed the reframed question, the same analyst does something quite different: gets the occupancy build and checks it adds up, 62 plus 14 plus 4; verifies the contract, the part with a signature; goes looking for a source for the 4 points of spillover and finds there is none; puts a month-six occupancy check in the diary; and writes down, in advance, that a reading under 68 or a spillover count of zero means the thesis is failing. The reframed plan is about a week of work, and the answer can be no.

The framing decides the work because every task in a plan is an attempt to answer the question, so a question with no way to be failed produces a plan with no task that could fail it. This is why the framing is the most consequential sentence the analyst writes, more than any conclusion. A wedding budget makes the same point at household scale. "Can we afford a good wedding?" sends the couple touring venues for a month. "Can we hold it for under Rs 12,00,000 with 300 guests, and which three quotes by the end of this month would tell us we cannot?" sends them to three phone calls. The second set of tasks is smaller, and it can come back with a no. Coming back with a no is exactly what makes it worth doing.

When should a question be reframed mid-way?

A question is not a vow. The question was framed on the facts available in January, and facts arrive after January. The rule for when to reframeTo rewrite the research question because new information has changed what would settle it: adding a part, moving the time frame, or naming new evidence. Not the same as changing the answer. is simple to state: reframe whenever new information changes what would settle the question, and never merely because the answer is looking uncomfortable. New information of the first kind adds a part, moves the clock, or names a new piece of settling evidence. A competitor's plan does exactly that: it introduces a claim the original question did not carry, that spillover survives the entry, and it names a new check, whether the competitor's site opens and whether any of Kaveri's spillover pallets move to it.

Reframing changes the question and therefore the work, and it is not the same thing as changing the answer. In the Kaveri case, a report reached Ishaan Verma in February noting that a competitor planned a facility nearby. He filed it. Under the vague question there was nothing to reframe. "Is Kaveri a sound credit?" had no part a competitor could touch, and the report was just one more interesting fact about the sector. Under the reframed question, the February report would have hit a specific leg. The 4 points of spillover were the only part of the thesis with no signature behind them, a competitor nearby takes spillover away, and the month-six check would have gained a second line: has the competitor opened, and has spillover arrived at all? Twelve months on, occupancy reached 71, spillover delivered 0, the competitor was open. The question as framed in January could not have caught that. The question as it should have been framed would have caught it in February, and again in June.

How was the Kaveri question framed, and how should it have been?

Put the two versions side by side and score each on the three parts. Ishaan Verma's January question was "is Kaveri a sound credit?" Its claim is an adjective. Its time frame is absent. Its settling evidence is whatever the analyst finds persuasive at the end, chosen after the fact. Zero of three, and the answer it received, "probably", is the natural output of a question that cannot be failed. The reframed version reads: "will occupancy reach 80 per cent within twelve months, from the 14 contracted points plus 4 points of spillover, and what would show by month six that the spillover is not coming?" Its claim is a number that is either reached or not. Its clock is twelve months with a check at six. Its settling evidence is a month-six occupancy reading, with an illustrative trigger under 68 per cent, and a spillover pallet count of zero on the site log. Three of three.

The reframed question would have put the 4 points of spillover, the only input in the register with no source behind it, under a lamp in January. The month-six check it named would have caught the competitor. Look at what it would not have done: it would not have changed the January recommendation on its own, because in January the spillover was still possible and the contract was real. The reframed question would have changed what happened next. The February report would have had a leg to hit. The June check would have existed. And in December, when occupancy read 71 rather than 80 and profit before tax read Rs 1,10,00,000 rather than Rs 4,20,00,000, the team would have known whether the loan was a good decision with an unlucky spillover, or a lucky decision on a question that could never have said no. The distinction between a good decision and a lucky one is what the assumption register, the decision log and the post-mortem all exist to preserve, and it is unreadable without a question that could be failed.

How do lenders, analysts and households actually use this?

A credit committee uses the research question as the sentence it will hold the analyst to. When Ishaan Verma's team meets, the strongest chair does not ask "is it a good credit?" but "what is the one thing that has to be true for this to work, by when, and what would tell us early that it is not?" Written into the approval, that sentence becomes the covenant test and the monitoring calendar: occupancy reported monthly, a review at month six, an automatic reopening if the reading is under the line. The lender's question is the loan's alarm clock, and a loan approved on a vague question has no alarm.

A sell-side or buy-side analyst uses it as the first line of the note, above the conclusion. The best notes state, near the top, "the thesis rests on X by date Y; we would be wrong if Z", and the whole document is organised as an attempt to fail Z. The Securities and Exchange Board of India (SEBI) research analyst regulations, on sebi.gov.in, ask by name that a recommendation carry a documented basis. A stated research question with its settling evidence is the most defensible basis there is. It says what the analyst looked for and what would have changed the view. Practitioners who write the no line first spend less time and produce work that can be checked, and that is why the sentence goes at the top rather than the end.

A household uses it without the vocabulary. "Should we buy this flat?" becomes "can we carry an equated monthly instalment (EMI) of Rs 42,000 a month if one salary stops for six months, and does our current saving rate, checked against the last twelve bank statements, say yes?" That is a claim, a period and named evidence, and it can come back with a no before any money moves. Every scale, from a wedding budget to a Rs 25,00,00,000 loan, uses the same three legs.

Kaveri Cold Chain Private Limited and Ishaan Verma are invented.
Educational material. Not advice on any investment, tax, budget or market position.