Analytics explained
Everything you need to understand how your AI product is performing — and how to improve it.
Every product on Productised has its own analytics dashboard, accessible from the product canvas. It gives you a complete picture of how your product is performing: who's going through it, how they're completing it, what outcomes they're receiving, and what the AI can tell you about patterns across your responses.
The analytics dashboard has four tabs: Charts, All Results, AI Insights, and Talking Points.
Charts
The Charts tab shows your aggregate performance metrics and a trend chart over time.
Metrics
| Metric | What it means |
|---|---|
| All Sessions | Total response records created — including viewed, started, completed, and abandoned sessions |
| Completed | Sessions that reached the end of the experience and were marked as complete |
| In Progress | Active sessions that have started but not yet completed |
| Abandoned | Sessions that went inactive for over 24 hours before completing |
| Leads | Sessions where lead details (email or phone) were captured |
| Completion Rate | Completed sessions divided by started sessions |
| Avg Messages | Average number of user messages sent across started sessions |
| Avg Score | Average scoring percentage across responses where scoring is enabled |
| Documents | Completed sessions that generated a Document outcome |
| Reports | Completed sessions that generated a Report outcome |
| Offers | Completed sessions that generated an Offer outcome |
| Decisions | How the AI classified completed sessions based on the hidden decision model |
Trend chart
The area chart plots completed sessions over time — useful for seeing the impact of a campaign, a social post, or a change to your product.
Cost & Credits
At the top of the Charts tab, the Cost & Credits card shows exactly what this product has used — broken into the four numbers that matter:
| Tile | What it means |
|---|---|
| Credits this month | Plan credits this product has consumed in the current billing period. Pairs with the run count so you can see cost-per-run. |
| Credits all-time | Lifetime credit usage. Pairs with a credits-per-response ratio if any responses have been captured. |
| Real $ this month | Real Anthropic dollar cost — what this product would cost in the current month if you were paying Anthropic directly via BYOK. |
| Real $ all-time | Lifetime equivalent. |
Below the tiles:
- A 14-day sparkline of daily credit usage — useful for spotting spikes after a campaign or a config change.
- Top cost drivers — which nodes in your workflow consumed the most credits this month (e.g. AI Advisor 77%, Standard Output 21%). If one node dominates and feels too expensive, tighten its prompt or move it to a smaller model.
- Powered by — the models that ran this product (e.g.
claude-sonnet-4-6,claude-haiku-4-5). Productised routes calls automatically based on phase.
If you have BYOK active (your own Anthropic key plugged in at Settings → Secrets), the card switches to "BYOK is on" and the real-$ tiles show your actual Anthropic spend — not an equivalent. Plan credits are not consumed.
If you don't have BYOK, the footer suggests it: "this product would cost you ~$0.84 this month direct to Anthropic" — a real number from your own usage, not an estimate. Click through to Settings → Secrets to add a key.
All Results
The All Results tab shows every individual response — a filterable table with one row per session.
Each row shows the session status, when it was created, whether lead data was captured, the decision result label (if a decision model is active), and which outcome was generated.
Click any row to open the full response — including the person's collected answers, the AI's decision result breakdown, the conversation history, and the generated outcome.
See Viewing results for a detailed walkthrough.
AI Insights — AI Product Coach
AI Insights analyses your responses and gives you an objective view of how well your product is working — where it's performing strongly, where it's creating friction, and exactly what to change to improve it.
This is not a dashboard of numbers. It is an AI-powered review of your product, run against your actual responses.
What AI Insights produces
Click Generate Insights to run the analysis. The AI reviews up to 20 recent responses and produces:
Output quality score A 1–10 score with a label and observation — how well the AI's generated outcomes matched what the conversation warranted. A strong match means the outcome felt earned and relevant. A weak match suggests the system prompt or collection goals need refinement.
Response themes Recurring patterns across your responses, categorised as positive signals, opportunities, or warnings. Positive themes confirm what's working. Opportunities suggest directions to develop. Warnings flag consistent problems worth addressing.
Confusion points Questions where people's responses were consistently vague, off-topic, or incomplete — with a pattern description and a concrete suggestion for how to improve that question.
Prompt improvement suggestions Specific, actionable suggestions for improving your AI Product's system prompt or individual node configurations — based on the gap between what the AI collected and what it produced.
Per-response insights For each sampled response: the inferred intent behind their answers, an output match rating (strong / partial / weak), and a specific observation about that session.
Talking Points
The Talking Points tab converts your response data into live audience intelligence — structured content you can use in webinars, keynotes, workshops, and podcasts.
If you speak to audiences that overlap with your product's users, Talking Points gives you something no generic research can: data from your own audience, analysed and formatted for presentation use.
Modes
Select the format before generating. Each mode shapes the output for a different presentation context:
| Mode | Best for |
|---|---|
| Webinar | Educational, framework-led presentations with Q&A |
| Sales Keynote | Pain → tension → solution arc, designed to convert |
| Workshop | Discussion prompts and group exercises |
| Podcast | Conversational hooks and soundbites |
What Talking Points produces
Audience summary — a plain-English description of who your audience is, based on the patterns in their responses.
Surprise finding — the most counterintuitive data point from your responses, formatted as a stat, an explanation, and why it matters.
Openers — three ways to open your presentation: a question, a statement, and a data point — each with a note on why it works for your specific audience.
Talking points — your key themes, each with: the percentage of respondents it applies to, a core point to make, an audience question to pose, and a transition tip.
Session arc — a suggested open, middle, and close for your presentation, built around your actual data.
Call to action — a suggested CTA for the end of your talk, informed by what your audience said they need next.