Prompt recipes
Ask for the decision. Let VSLZ choose the visual.
The best prompts describe what the reader needs to understand, the evidence available, and the context that makes the answer trustworthy. You usually do not need to name a chart type or visualization library.
For several coordinated visuals, explore composition prompts and complete examples: evidence networks, shared-scale small multiples, comparisons, and charticles.
Recipes below include basic visuals and editorial previews for small multiples, distributions, and connected relationships. Editorial templates require an enabled deployment. Use basic charts for ordinary trends and target gaps. Relationships should not be described as causes without supporting evidence. Explore every chart in the rendering lab.
RANK
Which segment is carrying retention?
Leadership wants a fast read on the strongest segment and where retention risk begins.
Try this prompt
Use VSLZ to help a leadership team answer: Which customer segment is carrying retention, and where is the risk?
FY2025 net revenue retention:
- Enterprise: 118%
- Mid-market: 109%
- SMB: 96%
- Partners: 88%
Treat 100% NRR as the minimum healthy threshold.
Make the takeaway obvious, rank the segments, emphasize only what matters, and keep the period, unit, and source visible. Choose the visual form that best communicates the answer.
Source: Northstar operating case — illustrative synthetic data.Good result
Preparing the evidence view…
Why this works
- The user asks a decision question, not for ECharts or a bar chart.
- The threshold makes “risk” objectively interpretable.
- The result leads with the answer, then shows the evidence.
CHANGE DRIVERS
Why did margin improve?
Finance leadership needs to understand a movement, not merely see two endpoint values.
Try this prompt
Use VSLZ to explain to finance leadership why gross margin improved from 42% in Q1 to 44% in Q2.
Drivers, in percentage points:
- Pricing: +5
- Mix: +3
- Support cost: -4
- Infrastructure: -2
Show the bridge from the starting margin to the ending margin. Make the main positive and negative drivers easy to see, and keep the units and source visible. Choose the best visual for explaining the change.
Source: Northstar operating case — illustrative synthetic data.Good result
Preparing the evidence view…
Why this works
- It gives the starting point, ending point, and reconciled drivers.
- It asks for the communication goal (“explain the change”), not a library command.
- The result headline performs the arithmetic and states the conclusion.
PROCESS
Where does onboarding get stuck?
Operations needs to understand a workflow and the intervention point.
Try this prompt
Use VSLZ to explain this onboarding flow to an operations lead and make the main intervention point obvious.
Flow:
1. Invite workspace
2. Configure workspace
3. Check whether data is ready
4. If data is not ready → resolve setup friction
5. If data is ready → activate first workflow
6. The setup-friction path also continues to activate first workflow
7. Activation → first value
The main bottleneck is around configuration, data readiness, and resolving setup friction.
Show a simple process that can be understood in a few seconds. Keep the bottleneck visually prominent and avoid unnecessary decoration.
Source: illustrative operating process.Good result
Preparing the evidence view…
Why this works
- The user supplies the actual process semantics.
- The desired reader outcome is explicit.
- VSLZ is free to choose the diagram structure while preserving the intended emphasis.
NETWORK
Where will reliability work have the most leverage?
SRE wants to find shared dependency points and inspect ownership interactively.
Try this prompt
Use VSLZ to help an SRE team decide where reliability work has the highest leverage.
Dependencies:
- Web app → Identity
- Mobile app → Identity
- Pricing → Checkout
- Identity → Checkout
- Checkout → Orders
- Orders → Payments
- Orders → Events
- Payments → Events
Known ownership:
- Identity → Platform
- Checkout → Commerce
Show dependency direction, make shared convergence points stand out, and let node selection reveal ownership or role metadata when available. This is an exploratory dependency map, not a fixed process flow.Good result
Preparing the evidence view…
Why this works
- The user communicates relationships as facts rather than Mermaid or Cytoscape syntax.
- The prompt distinguishes an exploratory network from a process diagram.
- Interaction is requested because it helps the decision, not because it is decorative.
TREND
How do the three groups change across quarters?
An editorial preview that keeps the evidence, units, and interpretation together for an operating review.
Try this prompt
Use VSLZ to help an operations lead answer: How do the three groups change across quarters?
Use the small_multiples_line editorial template when available.
Evidence:
[
{
"period": "Q1",
"value": 2,
"panel": "A"
},
{
"period": "Q2",
"value": 4,
"panel": "A"
},
{
"period": "Q3",
"value": 6,
"panel": "A"
},
{
"period": "Q4",
"value": 8,
"panel": "A"
},
{
"period": "Q1",
"value": 3,
"panel": "B"
},
{
"period": "Q2",
"value": 6,
"panel": "B"
},
{
"period": "Q3",
"value": 9,
"panel": "B"
},
{
"period": "Q4",
"value": 12,
"panel": "B"
},
{
"period": "Q1",
"value": 4,
"panel": "C"
},
{
"period": "Q2",
"value": 8,
"panel": "C"
},
{
"period": "Q3",
"value": 12,
"panel": "C"
},
{
"period": "Q4",
"value": 16,
"panel": "C"
}
]
Period: 2025.
Units: "units".
Source: VSLZ synthetic editorial examples.
Compare panels on a shared scale; preserve missing observations as gaps.
Lead with a supported takeaway, keep the exact data available, and preserve all observations.Good result
Preparing the evidence view…
Why this works
- The question identifies what the reader should learn.
- The complete evidence, period, source, and units travel with the prompt.
- The preview uses the same records and preserves the limits of the evidence.
DISTRIBUTION
How are the observed values distributed?
An editorial preview that keeps the evidence, units, and interpretation together for an operating review.
Try this prompt
Use VSLZ to help an operations lead answer: How are the observed values distributed?
Use the histogram_distribution editorial template when available.
Evidence:
[
{
"value": 0
},
{
"value": 7
},
{
"value": 14
},
{
"value": 2
},
{
"value": 9
},
{
"value": 16
},
{
"value": 4
},
{
"value": 11
},
{
"value": 18
},
{
"value": 6
},
{
"value": 13
},
{
"value": 1
},
{
"value": 8
},
{
"value": 15
},
{
"value": 3
},
{
"value": 10
},
{
"value": 17
},
{
"value": 5
},
{
"value": 12
},
{
"value": 0
},
{
"value": 7
},
{
"value": 14
},
{
"value": 2
},
{
"value": 9
}
]
Period: 2025.
Units: "units".
Source: VSLZ synthetic editorial examples.
Use six equal-width bins; count every observation once and label the count axis.
Lead with a supported takeaway, keep the exact data available, and preserve all observations.Good result
Preparing the evidence view…
Why this works
- The question identifies what the reader should learn.
- The complete evidence, period, source, and units travel with the prompt.
- The preview uses the same records and preserves the limits of the evidence.
RELATIONSHIP
How does each group move through the two measured dimensions?
An editorial preview that keeps the evidence, units, and interpretation together for an operating review.
Try this prompt
Use VSLZ to help an operations lead answer: How does each group move through the two measured dimensions?
Use the connected_scatter editorial template when available.
Evidence:
[
{
"x": 1,
"y": 2,
"entity": "A"
},
{
"x": 2,
"y": 4,
"entity": "A"
},
{
"x": 3,
"y": 6,
"entity": "A"
},
{
"x": 1,
"y": 3,
"entity": "B"
},
{
"x": 2,
"y": 6,
"entity": "B"
},
{
"x": 3,
"y": 9,
"entity": "B"
}
]
Period: 2025.
Units: "units".
Source: VSLZ synthetic editorial examples.
Connect each group in supplied observation order; do not imply causation.
Lead with a supported takeaway, keep the exact data available, and preserve all observations.Good result
Preparing the evidence view…
Why this works
- The question identifies what the reader should learn.
- The complete evidence, period, source, and units travel with the prompt.
- The preview uses the same records and preserves the limits of the evidence.
More editorial chart questions
These recipes complete the editorial catalog: radar profiles alongside the small-multiple line, histogram and connected-scatter recipes above. Each includes the same evidence used by its live preview.
compare
Multi-criteria comparison
Use radar compare to answer a focused question with complete, synthetic evidence.
Try this prompt
Use VSLZ to compare. Use the radar_compare template when enabled. The following input contains the complete illustrative evidence and encodings. Preserve every value and identify the source as synthetic. Keep exact values available in a data table. Do not infer causation or invent missing observations.
{
"insight": "Entity A leads on reliability while Entity B leads on speed",
"title": "Multi-criteria comparison",
"dataContext": {
"unit": "units",
"period": "2025",
"source": "VSLZ editorial fixture",
"synthetic": true
},
"type": "line",
"data": [
{
"criterion": "Speed",
"score": 8,
"entity": "A"
},
{
"criterion": "Quality",
"score": 6,
"entity": "A"
},
{
"criterion": "Reliability",
"score": 9,
"entity": "A"
},
{
"criterion": "Speed",
"score": 6,
"entity": "B"
},
{
"criterion": "Quality",
"score": 9,
"entity": "B"
},
{
"criterion": "Reliability",
"score": 7,
"entity": "B"
}
],
"encoding": {
"category": "criterion",
"series": "entity",
"value": "score"
},
"template": "radar_compare"
}Good result
Preparing the evidence view…
Why this works
- The template matches the reader's question.
- The prompt and preview share the same complete evidence.
- Sources, units and exact values remain inspectable.
Add icons, glyphs and supplied images
Use identity only when it helps the reader. Supply image bytes inline and keep text labels; there is no photo upload or server storage service.
IDENTIFY
Use icons without distorting the comparison
Add meaningful team identity to a quantitative chart with a separate label key.
Try this prompt
Use VSLZ to compare the supplied synthetic team output. Use the curated icons in this specification to identify the teams; retain text labels and do not replace the numeric bars with images. Preserve the source and exact values.
{
"type": "bar",
"title": "Output by team",
"insight": "Product delivers twice the platform output",
"dataContext": {
"source": "VSLZ visual identity fixtures",
"synthetic": true
},
"data": [
{
"team": "Product",
"value": 40
},
{
"team": "Platform",
"value": 20
}
],
"encoding": {
"category": "team",
"value": "value"
},
"visualAssets": {
"version": 1,
"definitions": [
{
"id": "product",
"kind": "icon",
"key": "team"
},
{
"id": "platform",
"kind": "icon",
"key": "server"
}
],
"bindings": [
{
"target": {
"type": "category",
"field": "team",
"value": "Product"
},
"assetRef": "product",
"placement": "label"
},
{
"target": {
"type": "category",
"field": "team",
"value": "Platform"
},
"assetRef": "platform",
"placement": "label"
}
]
}
}Good result
Preparing the evidence view…
Why this works
- Icons have a clear identity role.
- Numeric geometry remains unchanged.
- The complete evidence and bindings are supplied.
CONNECT VIEWS
Combine node badges and chart evidence
Show a mechanism and independent chart evidence with inline raster and icon identities.
Try this prompt
Use VSLZ to render this synthetic evidence_network infographic. Preserve the supplied inline image and icon definitions, visible labels and source context. No server photo storage or remote image fetch is needed. Treat the network as structure, not proof of causation.
{
"template": "evidence_network",
"title": "Delivery and output",
"insight": "Product leads output while the API connects delivery services",
"dataContext": {
"source": "VSLZ visual identity fixtures",
"synthetic": true
},
"blocks": [
{
"id": "network",
"kind": "network",
"role": "mechanism",
"spec": {
"title": "Delivery dependencies",
"insight": "The API connects the client to the database",
"dataContext": {
"source": "VSLZ visual identity fixtures",
"synthetic": true
},
"layout": "dagre",
"nodes": [
{
"id": "client",
"label": "Client"
},
{
"id": "api",
"label": "API service"
},
{
"id": "db",
"label": "Database"
}
],
"edges": [
{
"source": "client",
"target": "api"
},
{
"source": "api",
"target": "db"
}
],
"visualAssets": {
"version": 1,
"definitions": [
{
"id": "client",
"kind": "image",
"src": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAIAAAD8GO2jAAAAXUlEQVR4nO3VsQkAIAxEUbdwSAdxWCsLe8E6OWKCBOTg6v+6pIy5nq4QIHBWW9cWBUDayCDAWMeGClzVgSEDjrpmZADuumgQIEDg01PhNsRO3j+4MkAh+yfHRyAf2Hl+rBGQyfWTAAAAAElFTkSuQmCC",
"alt": "Synthetic blue circle badge",
"fit": "contain"
},
{
"id": "server",
"kind": "icon",
"key": "server"
},
{
"id": "store",
"kind": "icon",
"key": "database"
}
],
"bindings": [
{
"target": {
"type": "entity",
"field": "id",
"value": "client"
},
"assetRef": "client",
"placement": "node"
},
{
"target": {
"type": "entity",
"field": "id",
"value": "api"
},
"assetRef": "server",
"placement": "node"
},
{
"target": {
"type": "entity",
"field": "id",
"value": "db"
},
"assetRef": "store",
"placement": "node"
}
]
}
}
},
{
"id": "output",
"kind": "chart",
"role": "evidence",
"spec": {
"type": "bar",
"title": "Output by team",
"insight": "Product delivers twice the platform output",
"dataContext": {
"source": "VSLZ visual identity fixtures",
"synthetic": true
},
"data": [
{
"team": "Product",
"value": 40
},
{
"team": "Platform",
"value": 20
}
],
"encoding": {
"category": "team",
"value": "value"
},
"visualAssets": {
"version": 1,
"definitions": [
{
"id": "product",
"kind": "icon",
"key": "team"
},
{
"id": "platform",
"kind": "icon",
"key": "server"
}
],
"bindings": [
{
"target": {
"type": "category",
"field": "team",
"value": "Product"
},
"assetRef": "product",
"placement": "label"
},
{
"target": {
"type": "category",
"field": "team",
"value": "Platform"
},
"assetRef": "platform",
"placement": "label"
}
]
}
}
}
]
}Good result
Preparing the evidence view…
Why this works
- Both panels share one readable composition.
- The image is supplied inline with alternative text.
- No causal claim is inferred from the network.
Build a story from several views
These recipes include complete illustrative evidence and a live composition. Infographic features must be enabled in your deployment. Choose a single chart when it answers the question on its own.
VISUAL STORY
Evidence network
A relationship structure needs quantitative context. Use a network alone when no supporting quantities are needed.
Try this prompt
Explain this delivery network and quantify its output in one visual.
Audience: an operations team reviewing the year. Use the following illustrative data for 2025, measured in units. Source: VSLZ illustrative operating case. Preserve all values and label the data as synthetic.
[
{
"panel": "network",
"nodes": [
{
"id": "intake",
"label": "Intake"
},
{
"id": "north",
"label": "North",
"metadata": {
"team": "North"
}
},
{
"id": "delivery",
"label": "Delivery"
}
],
"edges": [
{
"source": "intake",
"target": "north"
},
{
"source": "north",
"target": "delivery"
}
]
},
{
"panel": "trend",
"data": [
{
"period": "Q1",
"value": 12,
"team": "North"
},
{
"period": "Q2",
"value": 18,
"team": "North"
},
{
"period": "Q3",
"value": 24,
"team": "North"
},
{
"period": "Q4",
"value": 32,
"team": "North"
}
]
},
{
"panel": "bars",
"data": [
{
"period": "Q1",
"value": 12,
"team": "North"
},
{
"period": "Q2",
"value": 18,
"team": "North"
},
{
"period": "Q3",
"value": 24,
"team": "North"
},
{
"period": "Q4",
"value": 32,
"team": "North"
}
]
}
]
Keep the panels readable on a phone and make exact chart values available. Explain the relationship between the views without claiming that the process caused the observed growth.Good result
Preparing the evidence view…
Why this works
- The question gives every panel a shared purpose.
- The prompt includes the evidence, units, period, and source used in the preview.
- The explanation connects the views while keeping causal uncertainty visible.
VISUAL STORY
Small multiples story
Repeat a chart shape with shared units and domains. Use a comparison chart for only a few values.
Try this prompt
Show North and South output as small multiples with a shared scale and linked quarter selection.
Audience: an operations team reviewing the year. Use the following illustrative data for 2025, measured in units. Source: VSLZ illustrative operating case. Preserve all values and label the data as synthetic.
[
{
"panel": "trend",
"data": [
{
"period": "Q1",
"value": 12,
"team": "North"
},
{
"period": "Q2",
"value": 18,
"team": "North"
},
{
"period": "Q3",
"value": 24,
"team": "North"
},
{
"period": "Q4",
"value": 32,
"team": "North"
}
]
},
{
"panel": "south",
"data": [
{
"period": "Q1",
"value": 8,
"team": "South"
},
{
"period": "Q2",
"value": 10,
"team": "South"
},
{
"period": "Q3",
"value": 12,
"team": "South"
},
{
"period": "Q4",
"value": 15,
"team": "South"
}
]
}
]
Keep the panels readable on a phone and make exact chart values available. Use the same numeric scale for both teams and link quarter selection.Good result
Preparing the evidence view…
Why this works
- The question gives every panel a shared purpose.
- The prompt includes the evidence, units, period, and source used in the preview.
- A shared scale makes the team comparison fair; linked quarters help inspect the same period.
VISUAL STORY
Cause and effect story
A proposed mechanism needs outcome evidence and a caveat. Do not label correlation as proven causation.
Try this prompt
Show the proposed delivery mechanism with its outcome trend and explain the causal uncertainty.
Audience: an operations team reviewing the year. Use the following illustrative data for 2025, measured in units. Source: VSLZ illustrative operating case. Preserve all values and label the data as synthetic.
[
{
"panel": "mechanism",
"steps": [
"Intake",
"Review",
"Delivery"
]
},
{
"panel": "trend",
"data": [
{
"period": "Q1",
"value": 12,
"team": "North"
},
{
"period": "Q2",
"value": 18,
"team": "North"
},
{
"period": "Q3",
"value": 24,
"team": "North"
},
{
"period": "Q4",
"value": 32,
"team": "North"
}
]
}
]
Keep the panels readable on a phone and make exact chart values available. Explain the relationship between the views without claiming that the process caused the observed growth.Good result
Preparing the evidence view…
Why this works
- The question gives every panel a shared purpose.
- The prompt includes the evidence, units, period, and source used in the preview.
- The explanation connects the views while keeping causal uncertainty visible.
Advanced users can still request a specific visual when the communication job calls for it. Start with what the reader needs to decide.