# Real-Time Focus Group

How to run a live group discussion with up to 20 personas, dig into their reasoning with follow-up questions and a reaction round, then read the themes, sentiment and opinion clusters.

Real-Time Focus Group is a live group discussion you run with up to 20 synthetic consumers (personas). You are the moderator: you write the questions, the participants you select answer in Turkish at the same time, and their answers appear in a chat window.

The method brings out the reasons behind a preference, the language consumers use and their objections. Example question: "If you saw this packaging on the shelf, what's the first thing that would come to mind?" Once at least 5 responses are in, you can have the AI prepare an analysis with a summary, sentiment, themes and opinion clusters.

## When to use it

- To see first reactions to a concept, packaging, ad or message, and the reasons behind them.
- To learn the words consumers use to talk about a topic before you write a survey.
- To understand in depth why a price or preference result came out the way it did.
- To ask about the same topic in two different wordings and see which framing gets better answers (A/B test).

A focus group is a qualitative method; other methods are better suited to measuring shares and prices. See [Choosing the right method](/en/docs/research/choosing-a-method). [Argus](/en/docs/argus/what-is-argus), SCL's AI research assistant, can also prepare the question guide and run the focus group for you, without a moderator. This route is covered in the Focus groups with Argus section below.

## Before you start

- **Credits:** You aren't charged a lump sum at the start of the session; 10 credits are deducted for each participant response.
- **Question guide:** A session can collect up to 200 responses. That's about 10 questions with 20 participants, or about 20 questions with 10 participants. Plan your guide accordingly.
- **Images:** If you plan to show packaging or an ad, prepare JPG, PNG or WebP files (up to 5 MB each). GIFs and other formats aren't accepted.
- **A/B decision:** If you plan to compare two question framings, turn on A/B mode before the first question.

## Setting up the study

1. Click **New Research**.
2. Select the **General Research** card.
3. On the **Choose Your Research Type** screen, select the **Real-Time Focus Group** card.
4. On the **Choose Population Mode** screen, click **Create New Population** or **Select Existing Population**. For a new population, **Population size** defaults to 20; you can enter a value from 1 to 20. If the existing population is larger, you select up to 20 participants from it in the next step.
5. While the personas are being prepared, the **Generating Participants** screen appears. The page refreshes on its own.
6. On the **Participant Selection** screen, select the participants (see the section below).
7. Click the continue button at the bottom (e.g. **Continue with 8 Participant(s)**). The **Focus Group Session** opens. This method has no **Research Preview** or **Start** step.

For the population form, see [Creating a population](/en/docs/populations/creating-a-population). In the Fashion vertical, a new population has 10–20 personas, and the **Odak Grubu Hazırlanıyor** (Preparing Focus Group) screen appears while it is being prepared. In Fashion, if the population has 20 personas or fewer, everyone is selected as a participant and the session opens directly.

### Selecting participants

In the **Participants** list on the right, each card shows the persona's name, [market segment](/en/docs/results/segmentation), socio-economic status (SES) and income group. Hover over a card to open a [profile card](/en/docs/populations/persona-profiles) showing age, gender, city, education, marital status, **Purchasing Behavior** and, if there is one, **Custom Feature**. Click a strategy in the **Smart Selection** section on the left, and up to 20 participants are selected automatically:

| Strategy | What it does |
| --- | --- |
| **By Segment** | Selects an equal number of participants from each market segment. |
| **By Age** | Selects a balanced mix from the 18-24, 25-34, 35-44, 45-54, 55-64 and 65+ groups. |
| **By Gender** | Builds a group balanced by gender. |
| **By Income** | Spreads participants evenly across income groups. |
| **Joint Composition** | Balances age, gender and income quotas together with personality diversity. |
| **Random** | Selects completely at random. |

When you click **Joint Composition**, the **Achieved composition** box opens. It shows a **Ready** or **Blocked** badge and three values: **Quota balance** (demographic balance), **Disposition spread** (diversity across the five core personality traits, i.e. OCEAN) and **Blocked share** (the share left out because demographic or personality data is missing). The **Realised state** table below it compares the population (**Pool**) with the selected group (**Group**). The other strategies select at random; if you click the same strategy again, you may get a different group.

For **Manual Selection**, click the cards one by one; once you reach 20 participants, the remaining cards can't be selected. **Select All** selects the first 20 personas in the list, and **Clear** resets the selection.

### Persistent Panel

> [!NOTE]
> The **Persistent Panel** box appears only in organisation accounts.

To reuse the participants you selected in later studies, type a name in the **Panel name** field and click the create button (e.g. **Create panel from 12 selected**). In your next focus group in the same vertical, select the saved panel from the list and click **Preview composition**. If the preview shows **Ready**, click **Sample panel and continue**: up to 20 participants are selected, with priority given to members used less in earlier studies, and the session opens.

In these sessions, participants speak in two waves: the extraverted personas answer first, and the second wave sees the first wave's answers. A very introverted persona may stay silent on some questions; no credits are deducted for a response it doesn't give.

### Focus groups with Argus

Instead of running the session yourself, you can have Argus run it. Based on the topic you discuss in the conversation, Argus prepares a question guide of 3–10 questions and sets up an unmoderated session with 4–20 participants. It selects the participants from the population at random; each participant answers the questions in the guide in order.

As in a session you run yourself, 10 credits are deducted per response. Before the session starts, however, the cost of the whole guide is checked against your balance: number of questions × number of participants × 10. For example, 8 questions × 10 participants × 10 = 800 credits. If your balance isn't enough, the session doesn't start and Argus tells you so.

If your balance is enough, Argus shows the **Launch the focus group** card above the message box. Read the card and click **Approve and launch**; once approved, the session can't be cancelled. If autopilot is on, the card doesn't appear. For the card and autopilot, see [Approvals and Autopilot](/en/docs/argus/approvals-and-autopilot).

The session moves forward question by question without you, and the analysis is generated automatically after the last question. You'll find the result as a card under the **Focus Group** filter on the **Reports** page; like other sessions, the card opens the **Odak Grubu Analizi** (Focus Group Analysis) page.

## Running the session

The participant list is on the left and the chat on the right. At the top are the **A/B Test** button, the **New Chat** button (once the chat has started) and a counter of responses used (e.g. 40/200); hover over the counter to see how many more questions you can ask. If SCL has brand or market information on file that matches your topic and vertical, the session draws on it automatically from the first question. In that case, participants know the brand to different degrees, as people do in real life.

### Asking a question

1. Type your question in the **Write your question...** box. Press Shift + Enter for a new line.
2. If you like, add up to 4 images with the **Image** button. The images are shown to participants along with the question.
3. Check the credit estimate next to the box (e.g. **~80 credits**).
4. Click the send button or press Enter. Under the question, you first see "… is preparing", then "… is writing".
5. Read the answers. Each answer appears with the participant's name and the time it was sent; below it there may be a short list of key points. If a participant can't produce an answer, the warning "… could not answer" appears; if no answer arrives within 180 seconds, "… took too long" appears.

You can leave the page; answers keep being generated in the background. While answers are being generated, the send button turns into a red **Stop** button; clicking it cancels the pending answers.

### Follow-up questions and the reaction round

To ask a follow-up question, type your next question. From the second question on, each participant answers having seen the key points the others raised on the previous question. To send a question to only some participants, click them in the list on the left: instead of **Send to everyone**, the strip shows the number you selected (e.g. **Send to 3 people**). To send to everyone again, click **Everyone**.

A **Tepkileri al** (Get reactions) button appears next to any question with at least 3 responses. When you click it, the system picks the sharpest quote and asks the 3–5 participants furthest from that view whether they agree with what was said, and why. Their answers appear in the **Tepki turu** (Reaction round) thread under the question. You can run only one reaction round per question.

The **AI Suggestions** button, with the sparkle icon, suggests three questions based on how the conversation is going; each suggestion carries a **Probing**, **Clarifying**, **Follow-up** or **Exploration** tag. Click a suggestion to put it in the message box; you can edit it before you send it.

### New chat

**New Chat** › **Start New Chat** archives the current chat and opens a new session with the same participants and a fresh allowance of 200 responses. Earlier sessions appear in the **Oturumlar** (Sessions) list in the header with an **Archive** badge. After the first question, the session's participants are locked. To change participants, first start a new chat, then go back to the selection screen with the **Participant Selection** button at the bottom.

### A/B test

An A/B test asks the same participants about the same topic in two different wordings.

1. Before the first question, click **A/B Test**. The chat splits into **Panel A** and **Panel B**; both panels have the same participants.
2. In each panel's own box (**Write a question to Panel A...**, **Write a question to Panel B...**), type one version of the question. To add an image, use **Add Image**.
3. If you like, get up to three question pairs with **A/B Suggestions**; the B version, tagged **Alternative**, asks the question with a different framing.
4. Send each panel's question separately.

> [!WARNING]
> Once the first question is sent, the session's structure is locked: A/B mode can't be turned on and **Close A/B** doesn't work. If you want to run an A/B test later, open a **New Chat** and click **A/B Test** before the first question.

## Settings and limits

| Setting | Limit |
| --- | --- |
| Number of participants | 1–20 |
| Responses per session | Up to 200 (including reaction rounds and both A/B panels) |
| Question length | Up to 5,000 characters |
| Images per question | Up to 4; JPG, PNG or WebP; up to 5 MB each |
| Reaction round | One per question (one per panel in A/B); needs at least 3 responses; 3–5 participants |
| Question suggestions | 3 questions; up to 3 question pairs in A/B |
| Analysis | At least 5 responses |

## Calculating credits

**Credits = number of participants you send the question to × 10.** Each response in a reaction round also costs 10 credits. Before a question is sent, the cost of the whole round is checked against your balance. If your balance isn't enough, the question isn't sent and a warning shows the credits required and the credits available.

The balance in the sidebar updates as soon as the question is sent. If a response is retried, credits aren't deducted twice.

| Case | Calculation | Credits |
| --- | --- | --- |
| One question to 8 participants | 8 × 10 | 80 |
| 6 questions to the same group | 6 × 8 × 10 | 480 |
| Follow-up question to 3 participants | 3 × 10 | 30 |
| Reaction round | (3–5) × 10 | 30–50 |
| A/B: 10 participants, one question per panel | 2 × 10 × 10 | 200 |
| A full 200-response session | 200 × 10 | 2,000 |

Question suggestions, the analysis, **Recalculate** and downloads don't use credits. For the credit rules for creating a population, see [Credits](/en/docs/account/credits).

## Reading the results

To open the **Odak Grubu Analizi** page, click the **Analytics** button in the session, or the card under the **Focus Group** filter on the **Reports** page. In **Reports**, every session with at least one message is listed as a separate card; the card says **Analytics Ready** or **Analytics not ready**. On the analysis page:

1. If there are at least 5 responses, click **Generate Analysis**. The analysis can take a few minutes; you can leave the page and come back in the meantime.
2. If you asked new questions after the analysis, click **Recalculate**. The analysis opens as it was saved and doesn't pick up new responses automatically.
3. To continue the chat, click **Back to Session**.

The boxes at the top show **Katılımcı** (Participant), **Soru** (Question), **Toplam Yanıt** (Total Responses), **Genel Duygu** (Overall Sentiment) and **Uzlaşı** (Consensus). Apart from the counts and the average word count, every metric is an assessment the AI makes by reading the transcript.

| Metric | Meaning |
| --- | --- |
| **Genel Duygu** | The overall tone of the discussion, from −1 to +1, shown as a percentage. Above 20% counts as **Pozitif** (Positive), below −20% as **Negatif** (Negative), and anything in between as **Nötr** (Neutral). |
| **Uzlaşı** | How much participants agree with one another; 0% means no consensus at all and 100% means full consensus. |
| **Duygu Dağılımı** (Sentiment Distribution) | The share of positive, neutral and negative responses. |
| **Anahtar Kelimeler** (Keywords) | Up to 12 phrases that appear in the responses; the text size shows their importance. Click a phrase to open the related responses in a side panel. |
| **Ana Temalar** (Key Themes) | The topics that stand out; **Ağırlık** (Weight) shows how much a theme was discussed. |

Tabs:

- **Genel Bakış** (Overview): a short **Özet** (Summary), the sentiment distribution, keywords and themes.
- **Kümeler** (Clusters): 3–5 groups of participants with similar views; each cluster's name, sentiment, members and standout demographic traits.
- **Custom Feature:** only if the population has a [custom feature](/en/docs/populations/custom-features); for each value, the number of participants and responses, and **Ort. kelime** (Avg. words), the average word count.
- **A/B Karşılaştırma** (A/B Comparison): only in A/B sessions; the **Genel Kazanan:** (Overall Winner:) badge (**Panel A Kazandı** (Panel A Won), **Panel B Kazandı** (Panel B Won), **Berabere** (Tie)), **Genel Duygu**, **Katılım Skoru** (Engagement Score) and **İçgörü Kalitesi** (Insight Quality) for each panel, and a **Çerçeveleme Farkı** (Framing Difference) and **Kazanma Sebebi** (Winning Reason) for each topic. In A/B sessions, **Genel Bakış** and **Kümeler** cover **Panel A** only.

**Delete** deletes only the computed analysis; the chat and the report card aren't deleted. For the general layout of reports, see [Reading a report](/en/docs/results/reading-a-report).

## Exporting

- **Download Excel** is available after the first question and contains the full transcript of the chat. On the Raw sheet, each response is one row: the participant's name, age, gender, city, education, income group and SES, the A/B panel, the question, the answer, whether the response came from a reaction round (yes/no) and the time of the response. The other sheets contain the summary, clusters and themes, segment impact, the A/B comparison and the population. Sheet names and headings are in English; the content (the answers and the analysis text) is in Turkish.
- **Download PDF** appears once the analysis has been generated and downloads the **Focus Group Report**: **Session Summary**, **Perception Aspects**, **Opinion Clusters**, **A / B Panel Analysis** if there is an A/B test, **Highlighted Questions** and **Summary**.

Files are prepared in the background. For details, see [Exporting and downloading](/en/docs/results/exporting).

## Tips and common mistakes

- **Ask clear, single-topic questions.** Participants usually answer in two to four sentences.
- **Send follow-up questions to the relevant participants.** Sending a follow-up question only to the relevant subgroup saves both credits and your 200-response allowance.
- **Seek out differing views on purpose.** Because participants see the key points from the previous question, their views can converge over time; **Tepkileri al** brings opposing views to the fore.
- **Use Smart Selection for a balanced group.** **Select All** doesn't build a balanced group. If **Joint Composition** shows **Blocked**, try another strategy or a larger population.
- **Treat the results as directional.** The answers come from personas played by language models; validate high-stakes decisions with real consumers. See [Accuracy and limitations](/en/docs/getting-started/accuracy-and-limitations).

## Related pages

- [Starting a new study](/en/docs/research/new-study)
- [Creating a population](/en/docs/populations/creating-a-population)
- [Credits](/en/docs/account/credits)
- [Persona profiles](/en/docs/populations/persona-profiles)
- [Working with Argus](/en/docs/argus/working-with-argus)