# From Survey Results to Business Strategy: How to Turn Findings Into a Decision

How to turn survey results into a business decision: frame the choice, judge effect size not just significance, size segments and rank actions on a matrix.

Published 2026-10-30 · Updated 2026-10-30
Canonical: https://hercules.works/blogs/survey-results-to-business-strategy

Start with the decision and the threshold that would change it, written before the data arrives. Then judge each finding by effect size with its interval, not by significance alone. Size the segments where the effect sits, and rank the possible actions by impact and effort. A significant result below your threshold does not justify full commitment.

## The Report Is Not the Decision

Many survey reports end with a list of significant differences. A business still has to decide what to launch, change or stop. The gap between those two things is where studies stall.

This guide is about the stage after fieldwork. It does not help you choose a study type (see [which research study fits a decision](/blogs/research-for-business-decisions-india/)) or write the decision rule in advance (see [data-driven decision making](/blogs/data-driven-decision-making-research-india/)). It starts when the numbers are in and shows how to read them for action, with one worked example built on invented figures.

## What Decision Is the Result Feeding?

Write four lines before you open the tables:

| Line | Example |
|---|---|
| Decision | Launch the new pack design nationally, in part, or not at all |
| Metric | Top-2-box purchase intent from the same question for both designs |
| Threshold | Launch widely only if the lift is at least 5 points |
| Evidence that would change the plan | A segment with a much bigger lift than the rest |

The threshold is a business judgement. It comes from what the change costs and what a gain is worth, not from statistics. Fixing it before the data arrive stops the team from adjusting the bar to fit the result.

## Why Is a Significant Result Not Enough?

Sullivan and Feinn, in the *Journal of Graduate Medical Education*, define effect size as "the magnitude of the difference between groups" and note that "the P value will not reveal the size of the effect". They add that "with a sufficiently large sample, a statistical test will almost always demonstrate a significant difference", and that "very small differences, even if significant, are often meaningless" ([Sullivan and Feinn, 2012](https://pmc.ncbi.nlm.nih.gov/articles/PMC3444174/)).

For survey percentages, the effect size is usually the difference in points, and its 95% interval says how large it could plausibly be. Ask two questions of every finding:

1. Is the interval clear of zero? (Is there probably any difference?)
2. Is the interval, or at least its lower end, clear of your threshold? (Is the difference big enough to act on?)

## Worked Example: A New Pack Design (Illustrative)

Everything below is invented to show the method. Two groups of 600 category buyers each see one design and answer the same purchase-intent question. The metric is top-2-box intent, as in our [top 2 box guide](/blogs/top-2-box-score/).

| | Current design | New design | Lift | 95% interval for the lift |
|---|---|---|---|---|
| All respondents | 204 of 600 = 34.0% | 246 of 600 = 41.0% | +7.0 points | +1.5 to +12.5 |

The two-proportion z is 2.50, so the lift is significant at 5%. Against the rule set earlier, though, the interval's lower end (1.5 points) is below the 5-point threshold. The data say there is probably a gain. They do not say the gain reaches 5 points.

Now split by a segment defined before fieldwork, say metro buyers aged 18 to 34:

| Segment | Share of category buyers | Current | New | Lift | 95% interval |
|---|---|---|---|---|---|
| Metro, 18-34 (180 per design) | 30% | 54 of 180 = 30.0% | 90 of 180 = 50.0% | +20.0 | +10.1 to +29.9 |
| Everyone else (420 per design) | 70% | 150 of 420 = 35.7% | 156 of 420 = 37.1% | +1.4 | −5.1 to +7.9 |

Segment sizing is share multiplied by lift. Per 1,000 category buyers, the first segment adds 0.30 × 0.200 × 1,000 = 60 more people with high intent, and the second adds 0.70 × 0.0143 × 1,000 = 10. The two sum to 70, the same as the 7.0-point total. Almost all the gain sits in 30% of the market.

Two cautions apply. A segment cut has fewer people, so its interval is wider. And a split chosen after seeing the data is exploratory; every extra cut raises the chance of a false positive, as our [cross-tabulation guide](/blogs/cross-tabulation-survey-analysis/) explains. Here the segment was set in advance, which makes the finding stronger but still worth confirming. Stated intent is also not sales, so treat these as evidence of appeal, not a revenue forecast.

## How Do You Rank the Possible Actions?

An impact and effort matrix sorts options into four groups. Minitab's guide names them Quick Wins (little effort, high impact), Major Projects (much effort, high impact), Fill Ins (little effort, low impact) and Time Wasters (much effort, low impact), and advises doing Quick Wins first because they "yield the best return on investment" ([Minitab](https://support.minitab.com/en-us/workspace/help-and-how-to/forms/types-of-forms/project-management/impact-vs-effort-matrix)). The page does not say how to score impact or effort, so use the survey for impact and your own team's estimate for effort.

| Action | Impact from the survey | Effort (a team estimate, assumed) | Group |
|---|---|---|---|
| Launch new design to metro 18-34 only | About 60 per 1,000 buyers | Low: one channel | Quick Win |
| Launch new design to all buyers | About 70 per 1,000 | High: new print runs everywhere | Major Project |
| Re-test a revised design with the other 70% | About 10 per 1,000 at the current lift | Low: one more survey | Fill In |
| Redesign the pack for everyone | Unknown | High | Not rankable until tested |

The matrix does not make the decision, but it shows the trade. In this example the segment launch captures 60 of the 70 points of gain at lower effort, which supports a targeted launch followed by a second wave to confirm the 20-point lift.

## Mistakes That Turn Good Data Into Bad Decisions

- **Letting p < 0.05 settle the question.** It answers a narrower question than the business is asking.
- **Moving the threshold after the result.** If you change it, say so in the report.
- **Reporting only the winning segment.** Say how many segments you looked at.
- **Ignoring who was not surveyed.** A result describes the people sampled, so say who they were and who was missing.
- **Treating one study as the last word.** Plan the confirming step into the decision, as the example does.

## Next Step

Before your next study, write the four-line frame from the first section and share it with whoever owns the decision. For the analysis steps used above, see our guides on [top 2 box scores](/blogs/top-2-box-score/), [cross-tabulation](/blogs/cross-tabulation-survey-analysis/) and [statistical significance testing](/blogs/statistical-significance-testing-india/). The Free plan includes 30 responses, enough to test a questionnaire before a full study.

## Frequently asked questions

### How do I turn survey results into business decisions?

Write the decision and a threshold before the data arrive. Then read each finding as an effect size with an interval, size the segments where it sits, and rank the actions by impact and effort.

### What is the difference between statistical and practical significance?

Statistical significance says a difference is unlikely to be chance. Practical significance asks whether it is large enough to matter. With a big sample, tiny differences can be significant but meaningless.

### What is effect size in a survey?

Effect size is the magnitude of the difference between groups. For percentages it is usually the gap in points, shown with its confidence interval.

### How do I size a segment from survey results?

Multiply the segment's share of your target market by the lift or gap measured in it. In the example, 30% of buyers times a 20-point lift gives 6 points of overall gain.

### Is stated purchase intent the same as sales?

No. Intent measures appeal under survey conditions. Treat it as evidence for a decision, and confirm with an in-market test or sales data where possible.

### How big a sample do I need before acting on a segment?

Look at the segment's own interval, not the total sample. If it spans both a trivial and a large effect, treat the segment result as a lead and confirm it with a second wave.