Quick answer: Plan your data cuts before the survey goes live, and tie each one to a business question you need to answer. List the segments you will compare, check that each will have a large enough base, and make sure the questionnaire captures every variable those cuts need. Cuts requested after fieldwork often reveal that a key variable was never asked or a group is too small to read.
Most survey teams agree the questionnaire carefully and then leave the analysis to sort itself out. The data arrives, someone asks for "a cut by region, age and weekday versus weekend", then a second wave of requests follows a week later, and a third after the first presentation. Each round adds time, and some requests turn out to be impossible because the variable was never collected.
Data cuts are the breakdowns of your results by group: users versus non-users, regions, age bands, ethnic groups, purchase channels and so on. They are where most insight lives, because averages hide the differences that drive decisions. Planning them before fieldwork is one of the simplest ways to make a study more useful.
Why plan cuts before the survey goes live?
There are three practical reasons.
- The questionnaire must capture the variable. If you want to compare weekend and weekday shoppers, you need a question that tells you which is which. If it is missing, no amount of analysis can create it.
- The sample must support the cut. A cut is only readable if each group has enough respondents. Knowing your cuts in advance lets you set quotas or boost small groups.
- Analysis gets faster. When the cut list is agreed, tables and dashboards can be prepared as data arrives, instead of waiting for requests after fieldwork closes.
How do you decide which cuts matter?
Start from decisions, not from demographics. For each business question, ask which groups you need to compare to answer it.
| Business question | Useful cuts | Why it matters |
|---|---|---|
| Which pack design should we launch? | Current users vs non-users, age group, main shopping channel | Shows whether the winner attracts new buyers or only pleases existing ones |
| Why are sales falling in one region? | Region, user status, recent switchers | Separates a regional problem from a national one |
| Is our campaign reaching the right people? | Ad recall vs no recall, target segment vs others | Links exposure to changes in brand perception |
| Which occasions should we target? | Usage occasion, time of day, weekday vs weekend | Shows where the brand is strong and where it is missing |
If a cut does not connect to a business question, it can usually wait. Every extra cut adds pages to the report and makes the important differences harder to spot.
How granular should cuts be?
Simple splits are a good start, but they can hide the pattern that matters. A weekday versus weekend split, for example, may show little difference, while a split by time of day and occasion reveals that the brand wins breakfast but loses late-night snacking.
The trade-off is base size. The more granular the cut, the fewer respondents in each group. Before agreeing a detailed cut, estimate how many respondents will fall into each group. As a general norm, many researchers avoid reading percentages from groups much smaller than about 50 respondents, and treat groups under 100 with caution. Our guide on whether a difference in your survey results is real explains how base size affects what you can conclude.
What goes into an analysis plan?
An analysis plan is a short document agreed alongside the questionnaire. It does not need to be long.
- The business questions, numbered, in plain language.
- The cuts for each question, with the variable that defines each group.
- Expected base sizes for each group, with any quotas or boosts needed.
- Key measures to compare across cuts, such as awareness, consideration or purchase intent.
- Significance testing rules, so everyone knows which differences will be flagged.
- Output format: dashboard filters, data tables, or a written report.
Share it with every stakeholder before fieldwork and ask them to add any cut they expect to need. One consolidated request at the start replaces several rounds of requests at the end.
What if new cut requests arrive after fieldwork?
They will, and some will be good ideas. Handle them in three steps. First, check that the variable exists in the data. Second, check base sizes before promising a result. Third, ask what decision the new cut will inform. If nobody can name one, it can go into an appendix rather than the main story.
It also helps to look at what you already have. Past studies often contain variables that can answer new questions, as our guide on getting everything out of your last study shows.
What does this look like in practice?
Imagine a beverage brand planning a quarterly survey on drinking occasions in Malaysia. The first brief asks for results "by demographics". The analysis plan sharpens this into three business questions: which occasions are growing, which groups are switching to competitors, and whether the latest campaign is reaching younger drinkers.
That plan reveals two gaps in the draft questionnaire: there is no question on time of day, and campaign recall is not linked to the brands recalled. Both are added before launch. It also shows that the youngest age group will be too small to read each quarter, so the team agrees to combine two waves before reading it. When the data arrives, the dashboard filters are already set up, and the first presentation focuses on decisions rather than on new table requests.
Where does Vase.ai fit?
We build Vase.ai so teams can set up quotas, screeners and dashboard filters around the cuts they need before launch, with a verified panel of Southeast Asian consumers and AI-validated responses. Our in-house research experts can also help write the analysis plan, or deliver the full research report built around your business questions.
Tell us the business question behind your next study, and we will talk through which comparisons will answer it and what the questionnaire needs to capture to make those comparisons possible.
Frequently asked questions
What is a data cut in survey research?
A data cut is a breakdown of survey results by a group, such as region, age, user status or purchase channel, so you can compare how different groups answered the same question.
How small can a group be before results are unreliable?
As a general norm, many researchers avoid reading percentages from groups much smaller than about 50 respondents and treat groups under 100 with caution. Significance testing helps judge each difference.
What is an analysis plan?
An analysis plan is a short document, agreed with the questionnaire, that lists the business questions, the cuts needed for each, expected base sizes, key measures and how results will be reported.