Frequently asked questions
The questions we hear most often during evaluations and the first weeks of usage. Missing one? Email hello@responz.io and we'll add it.
Setup & onboarding
How long does the initial analysis take?
The free 20-conversation preview is ready within 60 seconds. A full 90-day analysis on a mid-sized workspace (~10,000 conversations) completes in 5–15 minutes. You can use the dashboard incrementally as conversations are processed.
Do I need to install anything?
No. ResponZ is fully managed. You connect Intercom via OAuth and use the dashboard in your browser. No CLI, no SDK, no webhook configuration on your side.
Can I connect more than one Intercom workspace?
Yes, on the Professional and Enterprise plans. Each workspace sits in its own logical tenant — the data is never blended in the dashboard.
Data & privacy
Do you train AI models on our conversations?
No. We have a zero-retention agreement with our LLM provider. Your data is not used to train any model, ours or theirs.
Where is the data hosted?
European Union — application on Vercel EU regions (Frankfurt, Paris), database on a managed Postgres in the EU. No personal data leaves the EU without Standard Contractual Clauses.
What happens to my data if I cancel?
Account data and analyses are deleted within 30 days of cancellation. Invoicing data is retained for 10 years to comply with French accounting law.
Can I get a DPA?
Yes. Email hello@responz.io and we'll send the up-to-date version.
Billing & credits
What counts as one credit?
One conversation analyzed end-to-end (scored, clustered, sentiment-tagged, checked for gaps). Re-analyzing the same conversation does not consume an additional credit.
What happens if I exceed my plan?
On Pro, additional conversations are billed at €0.054 each and added to your next invoice. You always see the projected overage in the dashboard before it lands.
Can I switch between Pro and pay-as-you-go?
Yes, anytime, from Settings → Subscription. Changes apply at the next billing cycle.
AI scoring
How is the quality score computed?
Each reply is scored on four criteria — relevance, accuracy, tone, completeness — using an LLM with your scoring policy as context (brand tone, custom rules, sensitive topics). The overall score is a weighted average you can tune.
Why does the score sometimes disagree with the customer's CSAT?
CSAT measures customer feeling at one moment, on a 5-point scale, with strong selection bias (only a fraction of customers rate). The AI score measures reply quality on objective criteria. Some disagreement is expected. The reliability tile on Insights tells you when disagreement is structural and the scoring policy needs tuning.
Can I override or correct a score?
Yes. Every score in the Quality module has a validate / correct toggle. Corrections re-tune the weights for your workspace nightly.
Limits & scope
Do you support helpdesks other than Intercom?
Not yet. ResponZ is currently built specifically for Intercom (Fin Agent scoring is a central feature). Zendesk and Front are evaluated for 2026.
Does ResponZ reply to customers automatically?
No, and it won't. ResponZ is read-only by design. We score and recommend; humans (or your existing Fin Agent) reply. This boundary is deliberate.
What languages are supported?
All major European languages and English. Asian languages work for clustering and sentiment but the quality scoring is currently more reliable on Latin-script languages.
Tip