Questions to ask before commissioning research
Commissioned research costs money and time, and a meaningful share of it produces a deck that is read once. The reason is nearly always that the commissioning conversation skipped a few questions that would have redirected or cancelled the project.
None of them are difficult. They are just uncomfortable to ask when a budget has already been mentioned.
What decision is this for?
The first and most important question. Research without a decision attached becomes a general study, general studies produce general findings, and general findings change nothing.
If the honest answer is that no decision depends on it, that is a legitimate outcome of the conversation. Curiosity is a fine reason to read and an expensive reason to commission.
Do we already have the answer?
Frequently yes, in existing data that nobody has looked at, or in a study commissioned two years ago that is sitting in a folder. An hour checking is cheaper than a project.
A searchable register of past research is one of the cheapest things a commercial team can maintain and one of the rarest.
Can the existing data answer it?
Transaction data answers what happened. It cannot answer why, or what someone would do under conditions that have not occurred. Knowing which kind of question you have determines whether research is needed at all.
A brand asking why sales fell should look at what fell — item, store, daypart, price — before commissioning a study on shopper attitudes.
What result would change the decision?
The standard question and it applies here with force, because commissioned research is expensive enough that the answer should be established before spending. If both possible answers lead to the same action, the project is documentation rather than research.
Who is the sample and does it match the question?
For channel-specific questions this is where most research goes wrong. A national shopper sample tells you little about the independent channel specifically, and a sample of that channel's shoppers is harder and more expensive to recruit.
Ask for the sample definition and the achieved composition rather than the target. Achieved is what you get.
What are the methodological limits?
Every method has them, and a provider who volunteers theirs is more trustworthy than one who does not. Stated response rates, sample sources, question wording and weighting all affect what the results support.
Ask to see the questionnaire before fieldwork. Wording effects are real and a leading question produces an expensive confirmation of what somebody already believed.
How will the output be used?
Specify the deliverable and the audience before work starts. A study delivered as a two-hundred-slide deck to an audience that needed three findings has failed at the last step regardless of the quality of the work.
Frequently asked questions
Should I run a pilot?
For anything large, yes, because a small pilot exposes questionnaire problems and sample difficulties while they are cheap to fix. Skipping the pilot to save time routinely costs more time than it saves when the main fieldwork produces unusable answers.
How do I compare agency proposals?
On method and sample rather than on price and presentation. Two proposals at different prices are usually proposing different samples or different depths, and comparing headline costs without that is comparing incomparable things.
Can I specify the sample myself?
You should specify what the sample needs to represent and let the agency propose how to achieve it. Specifying a method you are not expert in is how a project ends up technically compliant and practically useless.
What about reusing an old study?
Check when it was run and what has changed since. Shopper research ages faster than most commissioners expect, and a study spanning a period of unusual conditions may not describe normal behaviour at all.
Should the agency know what we expect?
They should know the hypothesis and not the desired answer. Sharing a prior helps design; sharing a required conclusion invites a study built to produce it.
How long should a project take?
Long enough to do the fieldwork properly, which varies by method and sample difficulty. A timeline compressed below what the sample requires produces either a thinner sample or a looser one, and the compression should be an explicit trade-off rather than a hope.
What should be in the contract?
The sample definition, the achieved-sample criteria, the deliverables, the data ownership and whether you receive the raw data. The last is frequently overlooked and it determines whether you can re-analyse later.
Who should attend the debrief?
Whoever makes the decision, which is not always whoever commissioned. A debrief to the commissioning team that never reaches the decision-maker is a recurring and expensive pattern.
What is the single best question?
What decision does this inform, and what result would change it. Asked early and honestly, it prevents most wasted research. Broader industry context is collected by trade bodies such as FMI.
Should research be commissioned to settle an internal disagreement?
Only if both sides agree in advance what result would change their view. Research commissioned into an unresolved argument without that agreement usually produces a result that both sides interpret in their own favour.
### What should a brand actually do with this?
Keep a register of what has been commissioned, when, and what it concluded. The most common waste in commissioned research is re-answering a question somebody answered two years ago, and a searchable register prevents it at almost no cost.
Review the register before every new brief. It takes minutes and it occasionally ends the project before it starts, which is the cheapest possible outcome.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.