Do your dashboards show reality?
StreamLiner analyzes your Looker, Metabase, GA4, or Power BI screenshots to spot biases, explain what you're really looking at, and prioritize fixes.
Acquisition · Septembre
Mis à jour il y a 4 min
Sessions
128,4 k
+12,4 %
Conversion
2,4 %
+0,3 pt
Revenu
84,2 k€
+8,1 %
Panier moyen
61 €
−2,0 %
Taux de conversion
Sessions par canal
Appareils
- Mobile
- Desktop
- Tablette
- Autre
Analysis
Done in 14 sAverage clarity
Readable, but the main chart scale visually amplifies the trend.
3 findings
Y-axis not zero-based
HighConversion curve 2.1% → 2.4%: the rise looks 3× larger.
Vanity metric in dominant position
Medium“Sessions” occupies the most visible area, unrelated to revenue.
Low-accessibility palette
Low2.1:1 contrast between Tablet and Other series.
Built for screenshots from your reporting tools
- Looker Studio
- Metabase
- Google Analytics 4
- Power BI
- Tableau
- Superset
A dashboard is for deciding, not for reassurance. StreamLiner reviews your dashboards like a senior analyst: it describes what the eye catches, highlights scale, window, and comparison biases, then tells you what to fix first.
Reading
Read before you interpret.
The agent describes visual hierarchy, identifies the dominant KPI, and reframes the dashboard's implicit question, so everyone looks at the same thing.
Biases
Explicit biases, evidence included.
Truncated scales, cherry-picked time windows, vanity metrics, misleading correlations: each finding points to the area and its severity.
Action
A prioritized action plan.
Not a flat list of notes: recommendations ranked by impact and effort, with the score you can reach once fixed.
- < 2 min
- from screenshot import to full report
- 4 axes
- analyzed: hierarchy, scales, accessibility, readability
- 0 – 100
- a clarity score and severity per finding
Framework
A knowledge base, not a throwaway prompt.
Each analysis draws on best practices: data visualization, cognitive biases, accessibility, and information architecture. It lives in your code; you extend it with your own rules.
- Data visualization
- Cognitive biases
- Accessibility
- Information architecture
{
"severity": "high",
"title": "Axe Y non zero-based",
"evidence": "Courbe conversion 2,1 % → 2,4 %",
"rule": "dataviz/truncated-axis",
"suggestion": "Revenir à zéro ou annoter l’échelle"
}Workflow
From screenshot to action plan in four steps.
- 01
Account
Username and email code, or Google in one click.
- 02
Context
Your role, BI tools, and the decision the dashboard should inform.
- 03
Import
Drop a PNG or JPG screenshot, add notes if needed.
- 04
Analysis
The agent returns a score, spotted biases, and prioritized recommendations.
See what your dashboards aren't telling you.
Import a screenshot and get your first report before the next meeting.