Retour
Explanatory Research: Definition, Methods and Examples

Matthieu Saussaye

Explanatory research explains why and how something happens. It starts from a pattern you have already observed, such as satisfaction falling in one segment, and tests explanations for it: which factors drive the outcome, through what mechanism, and under which conditions. It typically combines statistical analysis of structured data with follow-up questions that capture people's reasons.
The short version:
It starts with a known pattern. Unlike exploratory research, you already know what happened. You are trying to explain it.
It tests competing explanations. Is the drop in satisfaction about price, product quality, or support response times? Explanatory research weighs them against each other.
It looks for the mechanism. Not only 'price matters' but 'the price increase reduced perceived value mainly among customers who use few features'.
It usually mixes methods. Driver analysis or regression on survey data shows which factors matter most; open-ended follow-ups show how respondents experience them.
It is not the same as proof. Strong explanatory evidence makes a cause plausible. Confirming it with certainty takes causal research, usually an experiment.
It leads to action. Because it identifies drivers, it tells you which lever to pull first.
Explanatory research is one of the purpose-based types of research. This guide covers how it differs from the others, the methods that work, the questions to ask and a step-by-step example.
Explanatory research methods compared
No single method explains everything. Strong explanatory studies combine one method that measures how much each factor matters with one that shows how it works.
Method | What it explains | Data needed | Strength | Limitation |
|---|---|---|---|---|
Key driver analysis | Which attributes are most associated with an outcome such as satisfaction or NPS | Survey with an outcome and attribute ratings, ideally 200+ responses | Ranks levers by importance | Association, not proof; attributes must be on the list |
Regression with controls | Whether a factor still matters once other factors are held constant | Structured data with candidate explanatory variables | Separates overlapping explanations | Only controls for what you measured |
Mediation and moderation | Through what (mediator) and for whom (moderator) an effect works | Measures of the cause, the outcome and the mechanism | Gets at the mechanism | Needs a clear model and adequate sample |
Subgroup comparison | Where the pattern is concentrated | Segment variables in the data | Simple, easy to communicate | Many comparisons invite chance findings |
Explanatory sequential mixed methods | Why the numbers look the way they do, in respondents' words | Quantitative phase followed by targeted interviews or open questions | Connects numbers to reasons | Takes two phases unless built into one survey |
Comparative case study | Why similar cases ended differently | Detailed data on a few contrasting cases | Rich explanation of process | Hard to generalize |
Explanatory vs exploratory vs causal research
These three are often mixed up, partly because some textbooks use 'explanatory' and 'causal' as synonyms. A useful practical distinction:
Exploratory | Explanatory | Causal | |
|---|---|---|---|
Starting point | A vague problem | A measured pattern | A specific hypothesis |
Core question | What is going on? | Why and how does this pattern occur? | Does changing X change Y? |
Hypotheses | Produced as output | Several, tested against each other | One, tested directly |
Typical methods | Interviews, open survey, desk research | Driver analysis, regression, mixed methods | Randomized experiment, A/B test |
Strength of conclusion | Themes exist | Plausible, ranked explanations | Measured cause and effect |
In a typical sequence, exploratory research names possible explanations, descriptive research measures the pattern, explanatory research ranks and unpacks the explanations, and a causal test confirms the one you plan to bet on.
How to conduct explanatory research in 6 steps
Describe the pattern precisely. 'Satisfaction among enterprise accounts fell from 7.8 to 6.9 between the spring and autumn waves' is something you can explain. 'Customers are unhappy' is not.
List the rival explanations. Write every plausible reason, including uncomfortable ones. Draw on support data, sales notes and earlier exploratory work.
Translate each explanation into something measurable. If the explanation is slower support, you need a rating of support speed and, ideally, actual response times.
Collect structured data and the reasons together. Rate the outcome and each candidate driver, then ask an open question about the main reason for the score. Follow up on vague answers.
Analyze in two passes. First, which drivers are most strongly associated with the outcome, controlling for segment. Second, what respondents say about those drivers in their own words.
Write the explanation and its confidence. 'The decline is concentrated among accounts onboarded after the spring, and is best explained by slower admin setup' plus what evidence supports it and what would confirm it.
Mediators and moderators in plain language
A mediator is the step in between. A price increase (cause) lowers perceived value (mediator), which lowers renewal intent (outcome). A moderator changes the strength of a relationship. The price increase hurts renewal intent much more among light users than heavy users. Naming both turns a vague 'price is the problem' into a specific explanation you can act on.
Explanatory research questions you can use
Explanatory surveys pair an outcome with candidate drivers and a reason question. Grouped by role:
Outcome questions
Overall, how satisfied are you with [product]? (0-10)
How likely are you to renew your subscription when it comes up? (5-point scale)
How likely are you to recommend [product] to a colleague? (0-10)
Driver questions (rate each)
How satisfied are you with the speed of our support responses?
How satisfied are you with the value you get for the price you pay?
How easy is it for your team to set up and administer [product]?
How well does [product] cover the tasks you need it for?
How reliable has [product] been over the past three months?
Reason and mechanism questions
What is the main reason for the score you just gave?
What has changed, if anything, in how you use [product] this year?
You rated value lower than last time. What made it feel less worth the price?
Which one change would most improve your rating?
Context questions (potential moderators)
How many people in your organization use [product]?
How long have you been using [product]?
Which plan are you on?
The reason questions only explain anything if the answers are specific. 'Support is slow' is a start; 'it takes two days to get an answer to an admin question' is an explanation.
A worked example: a satisfaction drop in one segment
A software company's twice-yearly survey shows satisfaction falling among enterprise accounts while other segments hold steady. Illustrative numbers throughout.
Step | Finding |
|---|---|
Pattern | Enterprise satisfaction 7.8 to 6.9; mid-market unchanged at 7.5 |
Rival explanations | Price increase, new admin console, slower support, competitor offer |
Driver analysis | Ease of administration shows the strongest association with satisfaction among enterprise accounts; price is second |
Moderator check | The drop is concentrated in accounts onboarded after the console change |
Open answers, coded | Most common theme: permissions setup takes too long for large teams |
Explanation | The new console made multi-team setup slower, which lowered satisfaction mainly for new enterprise accounts |
Next step | Causal test: a guided setup for half of new enterprise accounts, compare satisfaction and time to setup |
Without the open answers, the team would know 'ease of administration' matters but not what to fix. Without the driver analysis, they would have a pile of complaints with no ranking.
Advantages and limitations of explanatory research
Advantages | Limitations |
|---|---|
Turns a measured problem into a ranked list of causes to address | Does not prove cause; confounders may remain |
Uses data you often already collect, plus a few added questions | Only explains with the drivers you thought to measure |
Combines numbers with reasons, which is persuasive internally | Driver analysis can mislead when drivers are highly correlated |
Identifies for whom a problem is worst | Self-reported reasons can be post-hoc rationalizations |
Feeds directly into a focused experiment | Interpretation depends on analyst judgment |
Two habits keep explanatory work honest: report the rival explanations you rejected and why, and state which conclusion would need an experiment to confirm. For more on combining the two kinds of evidence, see qualitative vs quantitative research.
Common mistakes in explanatory research
Explaining a pattern that is just noise. A two-point dip in a subgroup of 40 respondents may not be a pattern. Check the base size and the margin of error before you look for causes.
Stopping at the first plausible story. The explanation that fits the team's existing view is the one most likely to go untested. Write the rivals down and test them all.
Taking stated reasons at face value. 'Too expensive' is often shorthand for 'not worth it for how little I use it'. Probe until the reason is specific.
Ignoring correlated drivers. When support speed and ease of setup move together, driver analysis can split their importance arbitrarily. Look at them together and in the open answers.
Treating the explanation as proven. Present it as the best-supported explanation, then test the lever you plan to invest in.
Collect the numbers and the reasons in one pass
Explanatory research often takes two rounds of fieldwork: a survey to find the drivers, then interviews to understand them. The second round is where projects stall.
SmartInterview combines both in a single survey. Respondents answer your outcome and driver ratings, then an AI interviewer asks why, by voice or text, and follows up on vague answers in the respondent's language. Open answers are coded automatically into themes, so you can put each driver's score next to the reasons respondents gave for it.
Start free and run your next explanatory study.
Frequently Asked Questions
What is explanatory research in simple terms?
Explanatory research explains why something you have already observed happens. It takes a known pattern, such as falling sales in one region, tests possible explanations against each other and identifies the factors and mechanism most likely behind it.
What is an example of explanatory research?
A company notices lower satisfaction among enterprise customers, surveys them on several drivers such as support speed, price and ease of setup, runs a driver analysis and asks open follow-up questions. The study concludes that slower admin setup best explains the drop. That is explanatory research.
What is the difference between exploratory and explanatory research?
Exploratory research investigates a problem that is not yet defined and produces hypotheses. Explanatory research starts from a defined, measured pattern and tests explanations for it. Exploratory usually comes first, explanatory later.
Is explanatory research the same as causal research?
They overlap and some sources treat them as synonyms. In practice, explanatory research explains why and how a pattern occurs using analysis of observed data and people's reasons, while causal research confirms a specific cause by deliberately changing it, usually in an experiment.
Is explanatory research qualitative or quantitative?
It can be either and is strongest when it uses both. Quantitative analysis such as regression shows which factors matter most, and qualitative follow-up shows how they work. This combination is often called an explanatory sequential mixed-methods design.
What are the limitations of explanatory research?
It cannot fully prove causation, it can only test explanations you thought to measure, and people's stated reasons may not match their real motivations. Confirm the most important explanation with an experiment before making a major investment.


