A broad subject like “supplier selection” or “permit delays” can’t be investigated until you turn it into a specific question, confirm that your audience makes decisions based on the answer, check what published work already covers, and make sure your evidence can answer it. This guide takes you through that selection in six steps. It ends with a one-paragraph idea statement you can judge before spending time or budget.
A research idea is a question, not a topic
A topic names a subject area. A research question defines what you will measure and how. Before you can investigate anything, a topic needs three parts: a defined audience, a clear scope, and an evidence route. If the study is numerical, it also needs a specific measure and a calendar period. If it’s qualitative, such as structured interviews, it needs a defined group of participants and the exact decisions or practices you will ask them about.
Topic lists built for students, such as Researchedd’s list of B2B marketing projects and A-Z Topics’ B2B marketing research topics, give academic researchers broad subjects to choose from. A B2B company has a harder filter: you have to pick a question your buyers care about and back it with evidence your team can reach.
Test your own draft idea against four checks before moving on: who the question is about, what scope it covers, what evidence could answer it, and, for numerical questions, what is counted and over what period.
Bare topic
“How contractors handle permits.”
Testable question
What share of commercial reroof permits issued in five metro areas between 1 January 2024 and 31 March 2026 were revised at least once within 180 days of issue?
Where your questions come from: your audience, your position, your data
Companies come into research from four different starting points. All four end in the same place: a list of specific questions to test.
- You already know what you want to say. You make a point regularly in sales calls or articles without numbers behind it, and you need a question that tests whether the evidence supports it.
- You already hold data. You have product, support or transaction logs (see proprietary data for content for how to take stock of them), and you need questions those fields can answer.
- You don’t know where to start. You have no internal records and no preset thesis, so you start from the practical problems your customers face and look for public data or a new study to answer them.
- You already publish research. You run an annual report or recurring survey and need your next question, one that goes beyond repeating last year’s questionnaire.
Whichever way you start, begin with the decisions your audience makes, such as what to buy, how to staff, or how to price a job, rather than what your marketing team wants to promote. Leave data availability for later. At this stage, write down every question worth asking.
Does it matter to your audience? Test relevance
Score every question on your list against one practical test: name the specific reader who would act on the answer, and the decision they would make differently once they have it. Search volume is not a relevance test, and a relevant question doesn’t guarantee media coverage or traffic.
A question can interest your internal team and still fail this test. Questions about how buyers view your own product features belong in product feedback, not public research. Keep only the questions that inform an outside buyer’s or practitioner’s decision.
Relevant
It names a reader who would act on the answer and the decision they would change, such as which job sizes a contractor bids on.
Not relevant
- How users rate one vendor’s feature: product feedback.
- “Is roofing a good industry to work in?”: too broad to change a decision.
What would your study add? Check existing knowledge
In The B2B Numbers Census, a website-content audit of 614 B2B companies observed in September 2026 and published in October 2026 by ClaimTheStory Research, 7 in 10 companies used someone else’s numbers, and 1 in 3 never published a number from their own data.
Originality in B2B research doesn’t mean nobody has ever studied the topic. It means your study adds something specific and verifiable to what is already published. A genuine contribution takes one of three forms:
- Updated evidence: newer figures for a more recent period, where older reports have gone stale.
- A different population: a specific industry, company size or region that broader studies lumped together or left out.
- A better-supported analysis: cleaner definitions, larger bases, or measured operational records where earlier studies relied on opinion or self-report.
Putting a new label on a known finding is not a contribution. To check what already exists, search published industry reports, trade associations, academic journals (A-Z Topics’ list above names several starting points) and government statistics. Record the publication date, population covered and method of every source you find.
Don’t rely on an AI chatbot to tell you whether a research idea is new. In ClaimTheStory’s Novelty Test (October 2026), given ideas that had already been done, ChatGPT and Claude couldn’t tell 54% of the time. Check primary sources yourself and log what you find.
Can you actually answer it? Feasibility and evidence
Every remaining question needs a realistic path to evidence. Your data will come from one of three routes:
- Records your company holds: operational, billing or usage logs, covered in proprietary data for content.
- Public or partner data: government, regulatory or association records, covered in how to get data for content.
- New collection: a structured questionnaire or a series of interviews, covered in survey research for content.
Match your question’s wording to what the records actually measure. A share of cases tells you how often something happened, not how long it took or why. When your planned evidence can’t answer a question, you have three choices: rewrite the question to fit the fields you have, switch to a different evidence route (such as interviews or a survey), or set the question aside until you confirm the missing data exists.
Choose one question and write it down
Compare your shortlisted questions side by side on relevance, contribution and feasibility, then pick one to develop. Write your choice down as a single paragraph covering seven items:
- The exact question
- Who it is for
- The decision it informs
- What is already known
- What the study adds
- The planned evidence and fields
- Its main limitation
Not chosen: row 1, bid win rates
Needs the company’s internal customer bid logs and a customer-data release approval.
Chosen: row 3, permit revisions
Public records, no customer-data approval, and limits the team can state plainly.
- Question What share of commercial reroof permits issued in five metro areas between 1 January 2024 and 31 March 2026 were revised at least once within 180 days of issue?
- For Independent commercial roofing contractors.
- Decision How much schedule and price allowance to set for permit risk on a reroof bid.
- Already known The cities publish permit records; the search found no analysis of revisions.
- Adds A better-supported analysis, every permit followed for the same 180 days.
- Evidence Permit records extracted 1 November 2026: permit ID, type, issue date, revision dates.
- Limit Shows that a permit was revised, not why or how long it delayed the job; five metro areas only.
Once your idea statement is written, move to the investment and cost checks in original research for content marketing. When you’re ready to run the study and publish the results, follow survey research for content, turning data into content and writing up survey and research findings. Our hub on data-driven content marketing shows how all the guides connect.
If you don’t know where to start, we find both: the idea, and where its data would come from. Book a demo