Build a Portfolio
A portfolio is the foundational configuration in Research Setup. It tells the platform who you are, what market you operate in, and who you compete against. Everything downstream, segment auto-definitions, mindset metric defaults, and qualitative research framing, draws from this context.
Set it up once per client or project. Get it right and the platform’s recommendations will be meaningfully specific; leave it generic and the results will be just as generic.
What a portfolio contains
Section titled “What a portfolio contains”| Field | Purpose |
|---|---|
| Industry | The broad market category (e.g., US Auto Insurance) |
| Brand | The client or focal brand (e.g., GEICO) |
| Competitors | The brands you want to benchmark or track |
| Country / region | Scopes audience data and census extrapolation |
The platform stores this context and applies it automatically every time you create a segment, generate a segment definition, or launch a qualitative or quantitative study.
Step-by-step: create a portfolio
Section titled “Step-by-step: create a portfolio”- Go to Research Setup. Open the platform and navigate to Research Setup from the main navigation.
- Start a new portfolio. Select the option to create a new research project or portfolio.
- Set the industry. Choose or type your industry category. Be specific: “US Auto Insurance” will produce more targeted recommendations than “Financial Services.”
- Add the brand. Enter the client brand name. This becomes the point of reference for all AI twin conversations; the platform will frame interactions as if you represent that brand.
- Add competitors. List the brands you want to benchmark against. The platform uses these for brand-track surveys and to ensure segment definitions consider the competitive landscape.
- Save the portfolio.
How the portfolio feeds downstream features
Section titled “How the portfolio feeds downstream features”Segment Recommend: When you click Recommend while naming a new segment without a pre-written definition, the platform reads your portfolio and generates a definition grounded in the industry and brand context you provided. A portfolio that says “US Auto Insurance + GEICO + [competitors]” will produce a segment definition about auto insurance shoppers, not a generic demographic description.
Mindset metric defaults: When you build a twin, the system proposes initial mindset metric values (brand attachment, emotional affinity, category awareness, etc.). These defaults are informed by the industry context in your portfolio.
Research framing: In Converse, the platform knows which brand you represent. When you ask a question as yourself (the “as-me” role), the AI twin responds with that brand context in mind.
Rule: be specific, not broad
Section titled “Rule: be specific, not broad”The platform is not a general-purpose LLM. If your portfolio describes an audience of “18–25 males in the US,” you are describing roughly 40% of the country. Results at that level of generality are no more useful than asking a large language model directly. Be as specific as your client brief allows; the portfolio is where that specificity lives.