> ## Documentation Index
> Fetch the complete documentation index at: https://docs.coloop.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Top Tips and Prompting Advice
> This guide will walk you through when to use Grids versus the AI Chat in CoLoop, helping you choose the right tool for different stages of your analysis. You’ll also find top tips for prompting the AI effectively to get richer, more targeted insights.
* [Top Tips Overview ](/docs/analysis/top-tips-and-prompting-advice#top-tips-overview)
* [When to Use the Grids versus the Chat ](/docs/analysis/top-tips-and-prompting-advice#when-to-use-the-grids-versus-the-chat)
* [Prompting Dos and Don'ts ](/docs/analysis/top-tips-and-prompting-advice#prompting-dos-and-don%E2%80%99ts)
* [Prompting Practical Tips](/docs/analysis/top-tips-and-prompting-advice#prompting-practical-tips)
### Top Tips Overview
### When to Use the Grids versus the Chat
| Analysis Grid | AI Chat |
| :---------------------------------------------------------- | :---------------------------------------------------------------------------- |
| Open analysis: “I’m not sure what I’m looking for just yet” | Closed analysis: “I know what sorts of things I’m trying to find or validate” |
| Good for exploring themes | Good for finding specific quotes / generating summaries |
| Based on one piece of evidence for every participant | Based on top N pieces of evidence |
| Can’t ask follow up questions | Can ask follow up questions |
### Prompting Dos and Don'ts
✅ **Rephrase your question:** Just like with Google — if you don’t find what you’re looking for the first time, try asking differently.
✅ **Ask specific questions:** CoLoop uses your prompt and project description to find answers. If it’s unclear what to look up, the answer won’t be helpful.
✅ **Ask simpler questions:** CoLoop considers one thing at a time. If you ask three questions at once, the response will be less detailed than if you asked them one by one.
❌ **Don’t ask quantitative questions:** The AI chat isn’t reliable for “how many” or “how much” questions — use the analysis grid instead.
✅ **Always check the full evidence panel:** The quote you’re looking for might not be at the top — scroll to review everything.
✅ **Use filters when possible:** Segment filters reduce the amount of data CoLoop processes and improve accuracy.
✅ **Reset the chat:** When changing topics, reset the chat to avoid carryover from earlier questions.
### Prompting Practical Tips
| Guidance | Example |
| :--------------------------- | :------------------------------------------------------------------------------------------------ |
| Be Specific | What are good quotes to support the idea that participants found the packaging difficult to open? |
| Break up complicated prompts | What are the main advantages of concept A? |
| Provide Examples | Find me some similar quotes to this one “ “ |
| Use Quotes | Did anyone mention “Product X” in their responses |
| Use Segments | What are some quotes from @non-users to support the idea that … |
| Tell CoLoop how to answer | What are some of the main concerns participants expressed? Answer in a bulleted list. |
| Give CoLoop feedback | Improve your last response by adding more detail about X |
### Detailed Example Prompts
For more involved analyses, here are detailed, ready-to-adapt prompts grouped by goal. Swap the placeholders — like `[topic]` or `Segment A/B` — for your own study.
**Segment comparison**
* Compare Segment A and B's top concerns. List each segment's top 3–5 with quotes, then highlight where they diverge most.
* Compare first reactions to Concept A vs B — emotional response, language used, and objections raised.
**Contradictions and conflict**
* Find where participants contradict themselves. Quote both statements, note the context shift, and what it reveals.
* Find topics where participants disagree. Group opposing camps with quotes and the underlying assumption driving each.
**Surprises and outliers**
* Surface the 5 most surprising or counterintuitive findings. Explain why each is surprising with supporting evidence.
* Identify outlier participants. Summarize their view and assess whether it's fringe or an early signal.
**Language and emotion**
* Identify recurring metaphors and phrases used to describe \[topic]. Group by theme, note which recur vs one-offs.
* Map emotional intensity across topics. Which spark strong emotion vs neutral mentions? Quote the strongest moments.
* Run a rigorous qualitative analysis: top themes with frequency, key tensions, segment differences, and gaps. Quote and cite for each.
* Run a JTBD analysis. For each job: functional, emotional, social; trigger; current alternatives; success criteria. Rank by frequency and intensity.
* Run a thematic analysis. Produce themes → sub-themes with counts, quotes, and business meaning. Flag thin themes.
* Build personas from this data. Each with behaviors, goals, frustrations, drivers, language, anchor quotes, and source participant IDs.
* Run a decision-driver analysis for \[key decision]. Separate stated reasons from revealed reasons where they diverge.
* Run a journey analysis for \[process]. Map stages, emotional highs/lows, friction points, and where journeys diverge by segment.
* Identify the strongest new market opportunity. Specify segment, unmet need, why now, evidence strength, and key risks.
* Suggest 3–5 product or feature opportunities. For each: unmet need, affected segments, evidence strength, impact vs effort.
* Suggest 2–3 positioning directions based on participants' language and emotional drivers. Include target segment and supporting quotes.
* Recommend which segments to prioritize vs deprioritize, based on signal strength, intensity of need, and willingness to act.
* What follow-up research is needed? Identify questions raised but unanswered, plus sample or methodological gaps.
* Pressure-test \[hypothesis]: evidence for, evidence against, what's missing, and a confidence verdict (strong/mixed/weak/contradicted).
* Find the strongest counter-arguments to \[finding]. Quote participants whose views complicate it and explain a skeptic's read.
* Does the data support \[new idea]? Identify evidence for/against, segments most/least likely to respond, and untestable assumptions.
* Stress-test for sample bias. What users, contexts, or viewpoints are under-represented, and how might that skew conclusions?
* What evidence would disconfirm my hypothesis that \[X]? Specify what future research would need to show.
* Write a compelling narrative for \[key finding]: hook, finding, why it matters, 4–6 supporting quotes, nuance, implications.
* Write a one-page executive summary: top 3 findings, implications, next steps. Assume the reader has 2 minutes.
* Reframe \[key finding] for \[audience: product/marketing/leadership]. Adjust emphasis, language, and the "so what."
* Pull 10–15 of the most powerful verbatim quotes. One line of context each, tagged by theme.
* Show the full evidence base for \[claim]: every supporting quote, every contradicting quote, with participant IDs, then a verdict.
* How many participants raised \[topic], in what context, with what intensity? Separate unprompted mentions from prompted ones.
* Audit \[previous summary or finding] against the raw data. Flag overstated, under-evidenced, or unsupported claims.