> ## Documentation Index
> Fetch the complete documentation index at: https://support.deepfield-ai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Choose a sample size

> Set a respondent target from the claims and subgroup bases the study needs to support.

Sample size should be determined by the smallest group you plan to report, the precision required, expected incidence, and the available budget. The feasibility bands under **Create → Quotas, panels & links → Sample size** provide a questionnaire-based reference, but they do not replace the sample design.

## Before you start

List the required comparisons and the smallest group that will support a reported claim. Bring the expected subgroup shares and their sources.

## Start with the planned analysis

List the results the study must support:

* A directional or exploratory read of the total sample.
* Estimates reported for the full sample.
* Comparisons across countries, segments, concepts, or quota cells.

Then identify the smallest subgroup used in a required comparison. A total of 300 responses provides a base of about 60 for a segment expected to represent 20% of the sample. If that subgroup carries an important claim, size and quota the study using that base rather than the total alone.

## Use the feasibility bands

The Sample size section places the saved respondent target within one of three bands:

| Band                          | Intended reading                                           |
| ----------------------------- | ---------------------------------------------------------- |
| **Exploratory**               | Directional patterns, themes, and supporting responses     |
| **Statistically significant** | Quantitative reporting primarily at the total-sample level |
| **Extended**                  | More subgroup analysis, crosstabs, and quota cells         |

The bands are calculated from the questionnaire. The displayed position changes when **Respondents** changes. Screening rules affect incidence and price rather than moving the target into another band.

## Check subgroup bases

DeepField's crosstab tests use the unweighted base:

* A column below 30 is not significance-tested.
* A column below 100 is marked for caution.

Work backward from the planned columns:

1. Estimate the share of each required subgroup.
2. Multiply that share by the proposed total.
3. Account for screening, quota limits, and expected dropout.
4. Increase the target or revise the analysis plan when a required group remains too small.

Use quota Min and Max values when natural incidence is unlikely to deliver the planned distribution.

## Balance precision, incidence, and cost

More completed responses can reduce sampling uncertainty, but they do not correct audience bias or questionnaire error. Narrow screening can also raise the price per response because a smaller share of entrants qualifies.

Review the sample plan with the Fieldwork price summary, where the target drives both the price per respondent in credits and, when DeepField recruits, the recruiting fee in euros. If the proposed target exceeds the budget, prioritize the groups and comparisons required by the decision, then revise optional cuts before reducing a required base.

## Record the decision

Keep the final target, expected subgroup shares, source of those expectations, and intended comparisons with the study brief. This makes quota changes and later weighting decisions easier to explain.

## Check the result

The saved panel targets match the documented sample decision. Record which subgroup bases the target is intended to support and review them again during fieldwork.

## Related

* [Configure your audience](/recruiting/panels-demographics)
* [Set quotas](/recruiting/quota-management)
* [Interpret significance](/analysis/reading-significance)
* [Decide whether to weight](/analysis/weighting-methodology)