# Priority Ranking (MaxDiff)

Explains Priority Ranking (MaxDiff), which measures which feature stands out in the purchase decision by having personas rank features in sets of four, from most to least important.

Priority Ranking (MaxDiff, best–worst scaling) ranks a product's features by how much they matter in the purchase decision. It answers a question such as "Which feature of our coffee subscription matters most to customers, and which matters least?" It doesn't ask about price; it only compares the features with one another.

Each synthetic consumer (persona) sees the features in groups of four; these groups are called question sets. In each set, the persona ranks the four features from most to least important for its own purchase decision. The feature it puts at the top counts as its **Best** choice, and the one it puts at the bottom as its **Worst** choice.

These choices are added up across all sets and personas, and each feature gets a score and an importance rank. Because the persona has to put one feature ahead of the others in every set, the difference between "nice to have" and "must have" shows up more clearly than when you ask about each feature on its own.

> [!NOTE]
> This method appears under different names on different screens: **Priority Ranking** on the General Research card, **Feature Preference (MaxDiff)** on the design page, **Özellik Tercihi (MaxDiff)** (Feature Preference) on the Fashion card and Fashion's design page, **Feature Prioritization** on the preview badge and in the report's header, and **Priority** in the **Reports** filter. They're all the same method.

## When to use it

- You have four or more features and need to decide which ones to put first: product development priorities, marketing messages, packaging claims.
- You want to see how important the features are relative to one another, and you don't need to measure the effect on price.
- You want to see whether priorities differ between demographic groups. For example, do younger and middle-aged customers put the same feature first?

To measure how many lira a feature adds to or takes away from willingness to pay, use [Feature Price Impact](/en/docs/research/feature-price-impact). Feature Price Impact takes at most 5 features, so you can narrow a long list down with Priority Ranking first. To compare methods, see [Choosing the right method](/en/docs/research/choosing-a-method).

## Before you start

- **Population:** You can choose a saved population or create a new one during setup; see [Creating a population](/en/docs/populations/creating-a-population).
- **Credits:** To create a General Research study in the **New Research** flow, your balance must be at least 500 credits. Creating a new population in Fashion requires at least 50 credits.
- **Feature list:** Prepare the list in advance, following the principles below; the list is what shapes the result most.

### Preparing a good feature list

- **Number:** You need at least 4 features. The more features you add, the more question sets and credits the study needs; keep the list to 20 features at most.
- **Same level of detail:** A concrete feature such as "30-hour battery life" can't be compared fairly with a general one such as "good user experience". Write them all at a similar level of concreteness.
- **One idea, no overlap:** Each feature should describe one thing. If "Fast shipping" and "Free and fast shipping" are in the same list, personas assess the same idea twice. Merge features that are very close to each other, or remove one of them.
- **Description:** Personas also see each feature's description. Use the description field to clarify a name that's open to interpretation, such as "Premium materials".
- **Duplicates:** If you enter a name that's already in the list (ignoring case), the feature isn't added.

## Setting up the study

1. In the sidebar, click **New Research**.
2. Choose the **General Research** or **Fashion** card.
3. On the method screen, click the **Priority Ranking** card. In Fashion, this card is called **Özellik Tercihi (MaxDiff)**.
4. Choose or create a population. This step is covered on the [Starting a new study](/en/docs/research/new-study) page.
5. On the design page, titled **Feature Preference (MaxDiff)** (**Özellik Tercihi (MaxDiff)** in Fashion), enter your product's name in the **Product Name** field (required). You can also fill in the **Product Description (Optional)** field if you like.
6. Add features in one of the ways described below. You can write a short description in the **Optional description...** field under each feature, and delete a feature with the **x** button.
7. Check the status line. Once there are 4 or more features, the line turns green and reads, for example, **10 features · 8 question sets**. If there aren't enough features, a warning such as **Add 2 more features** appears.
8. Click **Next Step**. While the population is being prepared, the button reads **Waiting...** and can't be clicked. In General Research, click **Continue** in the **Are you sure you want to continue?** dialog that opens; your design is saved.
9. On the **Research Preview** page, check the **Person Count**, **Feature Count** and **Credit Usage** values on the **Experiment Summary** card.
10. Click **Start**, then **Continue**. When the study finishes, the **Your analysis is ready** card and the **Go to Reports** button appear; the report opens automatically within a few seconds.

### Ways to add features

| Option | How it works |
| --- | --- |
| **Suggest with AI** | The AI suggests 10 features based on the product name and adds them to the end of the list. You can't click it while **Product Name** is empty. The gear icon next to it opens the **AI Instructions** field, where you can add guidance such as "Focus on sustainability features". |
| **Add feature...** field | Type the feature, then press Enter or click the **+** button. |
| Pasting | If you paste several lines into the **Add feature...** field, each line becomes a separate feature. If a line contains a Tab character, the part before the Tab becomes the name and the part after it the description, so you can copy the name and description columns from a spreadsheet together. |
| **Bulk Edit** | Opens the whole list in a text box: one feature per line, with the name and description separated by a Tab. Edit the text and click **Apply**. |
| **Select from Saved Surveys** | Loads a Priority Ranking design you saved to the **Surveys** library. The card only appears if you have a saved design. If you change a field in the loaded design, the selection is cleared. |

The **Clear** button deletes all the features in the list. AI suggestions come without descriptions; review the list and add descriptions where needed.

> [!WARNING]
> Once you've used **Suggest with AI**, a **Regenerate** button appears next to it. **Regenerate** deletes the whole list, including the features you added by hand, and fetches new suggestions. To get suggestions while keeping the features you added by hand, click **Suggest with AI** again; the new suggestions are added to the end of the existing list.

## Settings and limits

| Setting | Value |
| --- | --- |
| **Product Name** | Required, up to 255 characters |
| Number of features | At least 4; at most 20 recommended |
| Feature name | Up to 255 characters |
| Feature description | Optional, up to 500 characters |
| Features per question set | 4; can't be changed |
| Times each feature is shown | At least 3 per persona |

### How question sets are built

You don't enter the number of question sets; it's calculated from the number of features. The formula is in the credit calculation section below. The sets are balanced: each feature is shown roughly the same number of times and appears alongside different features. The features within each set are shown in shuffled order.

| Features | 4 | 5 | 6 | 8 | 10 | 12 | 16 | 20 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Question sets | 3 | 4 | 5 | 6 | 8 | 9 | 12 | 15 |

## Credit calculation

Credits = number of question sets × number of personas

Number of question sets = number of features × 3 ÷ 4, rounded up.

- 10 features, 100 personas: 10 × 3 ÷ 4 = 7.5, rounded up to 8 question sets; 8 × 100 = 800 credits.
- 16 features, 250 personas: 16 × 3 ÷ 4 = 12 question sets; 12 × 250 = 3,000 credits.

Every four extra features add three more question sets per persona. The estimated credits appear on the **Research Preview** page as **Credit Usage**, in the form "balance/estimate", for example **1.250/800** (a balance of 1,250 credits against an estimate of 800). If your balance doesn't cover the estimate, a red info icon appears next to the value, and you can go to **Buy Credits** from the icon.

Credits are deducted after the responses are generated and saved; a study that fails before any responses are generated doesn't spend credits. Your balance is checked again when the study starts; if it's insufficient, the study doesn't run and no credits are deducted. In General Research, the credits deducted are calculated from the number of saved responses (1 credit for each question set completed by each persona), so they may differ slightly from the estimate. For details, see [Credits](/en/docs/account/credits).

## Reading the results

Completed studies are listed on the **Reports** page; use the **Priority** filter to see only this method's reports. The **Refresh**, **Download PDF**, **Download Excel** and **Delete** buttons at the top of the report are covered on the [Reports](/en/docs/results/reports) page.

### Key measures

| Measure | Meaning |
| --- | --- |
| **Best** | Number of times the feature was chosen as most important in its sets (total across all personas) |
| **Worst** | Number of times the feature was chosen as least important in its sets (total across all personas) |
| **Net** / **Net Score** | Best minus Worst. If it's positive, the feature was put first more often; if it's negative, it was put last more often. |
| **Score (0-100)** | The net score converted to a 0–100 scale. The most important feature you tested sits at the top end of the scale, and the least important at the bottom end. |

Example: in a study with 100 personas and 10 features, each persona answers 8 sets, so 800 Best and 800 Worst choices are made in total. If a feature was chosen as Best 150 times and as Worst 30 times, its net score is 120.

> [!IMPORTANT]
> Scores are relative. The **Not Important** label doesn't mean the feature has no value to consumers; it means the feature came last among the features you tested. Adding features to the list or removing them changes all the scores; don't compare the scores of two studies run with different lists directly.

### Summary cards

The report has five summary cards:

- **Participants:** The number of personas included in the analysis.
- **Total Features:** The number of features tested.
- **Best Feature** and **Worst Feature:** The first and last features in the ranking, with their scores.
- **Most Differentiated Feature:** The feature whose net score is closest to zero; it was either chosen as Best and Worst about equally often, or rarely chosen at either end. It doesn't mean "most liked"; it shows the feature on which personas were split or indifferent.

### Best / Worst tab

In this tab's chart, each feature's Best choices are shown as a green bar and its Worst choices as a red bar, with the net score on the right. A feature with both a long green bar and a long red bar is very important to some personas and unimportant to others; to see who it matters to, check the **Segment Analysis** tab.

### Importance Ranking tab

The **Feature Importance Ranking** section has a card for each feature, showing its rank (#1, #2...), its score out of 100, its importance label and its Best, Worst and Net counts. Labels are assigned by score:

| Score | Label |
| --- | --- |
| 75 and above | **Very Important** |
| 60–75 | **Important** |
| 40–60 | **Medium** |
| 25–40 | **Low Importance** |
| Below 25 | **Not Important** |

The **MaxDiff Feature Details** table below the cards has **Rank**, **Feature**, **Score (0-100)**, **Net Score**, **Best** and **Worst** columns. You can sort the table by clicking a column heading, search for a feature with the **Filter...** box and download the table with **Download CSV**. At the bottom of the page, the **Most Important Features** and **Least Important Features** boxes summarise the top three and bottom three features.

Results are language model output; they show direction and magnitude. For high-stakes decisions, see [Accuracy and limitations](/en/docs/getting-started/accuracy-and-limitations).

## Analysing results by segment

The **Segment Analysis** tab shows how priorities change across demographic groups.

1. Choose a dimension from the **Select Demographic** list, for example **Age**, **Gender**, **Education**, **Income Group** or **SES** (socio-economic status). If you added a custom feature to your population, **Custom Feature** is listed too. When the page opens, the first dimension in the list is selected.
2. In the **Segment Comparison Chart**, compare the top 8 features across segments. Choose **Score (0-100)**, **Net Score** or **Ranking** from the **Metric** list.
3. In the **Analysis by …** section below the chart (e.g. **Analysis by Age**), read each segment's number of personas, its share and its **Most important:** feature.
4. In the **Feature Comparison** table, compare the rank and score of up to 10 features side by side across the five largest segments.

The report has no free-form filter: breakdowns are limited to the dimensions in the **Select Demographic** list, and you can't filter by two dimensions together (e.g. "women aged 25–34"). To look at a specific group, filter the **Segment Impact** and **Raw** sheets in the Excel file, or set up a new study with a population made up only of that group ([Filters and attributes](/en/docs/populations/filters-and-attributes)). For details of the dimensions, see [Segmentation](/en/docs/results/segmentation).

## Exporting

- **Download PDF:** If the report includes segment dimensions, the **PDF segment selection** dialog opens first. Tick the dimensions to include in the PDF and click **Generate PDF**. The PDF report includes importance ranking, best/worst, breakdown and findings sections.
- **Download Excel:** An Excel file containing all the data is generated and downloads when it's ready. The file is in your interface language, but the sheet names stay in English.

| Excel sheet | Contents |
| --- | --- |
| Meta | General information about the study |
| Executive Summary | **Feature MaxDiff Summary**: **Personas**, **Features**, **Sets**, **Top Feature**, **Bottom Feature**, **Response Completion** and the **Feature Importance Scores** chart |
| Feature Rankings | **Feature Rankings**: **Rank**, **Feature**, **Shown** (how many times the feature was shown), **Best**, **Worst**, **Net**, **Normalized Score** |
| Segment Impact | **Segment Impact**: for each feature and segment, **Segment Dimension**, **Category**, **Persona N**, **Share**, **Rank**, **Normalized Score**, **Best**, **Worst** |
| Population | The personas' demographic details |
| Raw | Responses per persona, with the number of completed and expected sets and the completion rate |

The **Feature Rankings** and **Segment Impact** sheets come with Excel filter buttons already in place. For download details, see [Exporting and downloading](/en/docs/results/exporting).

## Tips and common mistakes

- **Watch out for small segments.** In a segment of only a few personas, even a handful of choices change the scores noticeably. Check each segment's number of personas in the **Analysis by …** section.
- **Reuse the design.** You can use a Priority Ranking design you saved to the **Surveys** library with different populations ([Survey library and question design](/en/docs/research/survey-library)). To run a saved design on a ready-made population from a single screen, see [Quick Research](/en/docs/research/quick-research).
- **Get help from Argus.** Argus, SCL's AI research assistant, can fill in this form for you in a conversation; see [Working with Argus](/en/docs/argus/working-with-argus).

## Related pages

- [Feature Price Impact](/en/docs/research/feature-price-impact)
- [Starting a new study](/en/docs/research/new-study)
- [Segmentation](/en/docs/results/segmentation)
- [Exporting and downloading](/en/docs/results/exporting)
- [Credits](/en/docs/account/credits)