Analyse all your feedback in one place, with every insight checked. Quantify what each problem costs, on your own data.
Priced on your own data.
The business case builder turns a fix into a three-year return in euros. You enter the financial inputs; it adds the expected improvement, a low and a high scenario, and the risk.
Churn drivers
Pick a driver to price its fix.
Fix the top-up confirmation delayFix the eSIM transfer at activationExplain roaming charges before the trip
Customers affected
Revenue per customer, a year
Cost of the fix
Fewer leave, from the analysis8 to 15%5 to 10%3 to 6%
Present value, three years
€0
Low and high scenario, after the cost
Return on the fix
0
Risk-adjusted
Year 1
Year 2
Year 3
“Users rated the platform strongly for use cases such as text and sentiment analytics, where it performs above the category average.”
InsightNarrator analyses every answer, in the language it was written in.
Scale into sandsiv+ when ready
Same company, same Swiss rules.
InsightNarrator
Our agentic analysis module turns raw customer feedback into findings, dashboards, stories and business cases, and cites the customer’s words behind every claim. It is the Analyse & Quantify module of sandsiv+, and it is also sold on its own with a free trial.
FilesCSV and Excel, up to 100 MB a file, 50 files at onceAvailable
Survey tools26 connectors, synced by hand, hourly, daily or weeklyAvailable
Reviews25 public platforms, plus an AI reader for any web addressAvailable
The assistant on every page of InsightNarrator. It does the work instead of describing it: it creates datasets, runs analyses and agent runs, builds dashboards and surveys, and writes stories.
Why is churn high among prepaid customers in Ticino?
Prepaid detractors complain about billing far more than postpaid customers, and the gap grows every week.
Source: your data · Confidence: high · 1 query, 412 rows, 0.8 s
One answer, with its source
Every answer shows where it came from
The source, a confidence level, the query it ran with its row count and time, and how fresh the data is. An optional deep review has a second reviewer challenge the answer before you see it.
ToolsWell over a hundredAvailable
ExportsCSV, Excel, Word or PowerPointAvailable
SkillsData analyst, file exporter, survey designer, and your ownAvailable
Outside toolsThrough MCP, when a workspace admin switches them onAvailable
Sentiment and risk
Two analyses read every comment.
Aspect-based sentiment
Several aspects per comment, each placed in your categories, with its sentiment (positive, negative, neutral or mixed), a confidence level, the words it rests on and an explanation.
Complaint risk
A score from 0 to 100 and a level (low, medium, high, critical), across 28 risk factors in six groups, from service quality to escalation signs, with the evidence.
RowsUp to 100,000 a jobAvailable
LanguagesEach comment analysed in the language it was written inAvailable
AutomaticSet to run on a dataset as new feedback arrivesAvailable
The analysis catalogue
More than 30 analysis types: drivers, root causes, jobs to be done, moments of truth, churn risk, share of voice and more.
Groups
Customer feedback, voice of the customer, customer journey, experience measures such as NPS, CSAT and CES, behaviour, competitive intelligence, operations, and synthesis across analyses.
How they read the data
Catalogue analyses work from a random sample of up to 500 comments, sized to the model. Aspect-based sentiment and complaint risk read every comment.
Your ownCustom analyses and instructions for each runAvailable
ScoresScores, categories and cross-tabs computed in the database, with a written narrativeAvailable
BriefScoty drafts the objective and the key questions and proposes up to five analysesAvailable
Your definitionsA Knowledge Pack of your metrics, glossary and rules, approved by a personAvailable
The agent team
A team of ten agents. NarratorGPT plans the run and writes the report.
A run
Five phases: set-up, analysis, audit, synthesis, report. The agents query your data with read-only SQL. The Quality Auditor scores confidence from 0 to 100 and flags contradictions. You can steer them while they work. A run takes minutes, not seconds.
AlwaysNarratorGPT, Data Scientist, Domain Expert, Synthesis Analyst, Quality AuditorAvailable
As neededSentimentAI, ThemeMapper, TrendRadar, SegmentLens, RiskAlertAvailable
ReportsMarkdown, HTML or JSON, with an evidence appendix of queries and row countsAvailable
What the analysis finds, in the form each audience reads.
Dashboards
Sixteen widget types, filters, templates and dashboards built by AI, with version history and auto-refresh. Exported as PDF, PNG, JPEG, SVG, HTML or JSON, or shared through a read-only link you can revoke.
Stories
Narrative documents built from blocks: text, tables, images and widgets, with comments and version history. A story can regenerate daily, weekly or monthly. Revenue stories cover acquisition, retention and expansion.
LibrarySaved reports with folders, tags and favouritesAvailable
MonitoringRoutines that alert on sentiment drops, new topics and rising risk, in the appAvailable
Business caseThree-year risk-adjusted return and present value, from your inputsAvailable
Languages and AI
Feedback in any language is analysed as written.
You choose the models
A workspace can keep to a pure EU stack, including models hosted in Switzerland, or use the latest frontier models. The model can be set per workspace, per task and per run.
Report languages57, chosen per runAvailable
TrainingYour data never trains a shared modelAvailable