Synthetic dashboard case study

Forgeline:
A Lead-Gen Funnel Dashboard, Built by Prompting an AI Agent

A fictional B2B SaaS dev-tool dashboard: hand-coded D3.js v7, seeded synthetic data, dark theme, and panel-by-panel iteration with GitHub Copilot CLI.

Author: Vitaliy MatiyashSingle-file HTML demo
Fictional data only. Forgeline, all leads, and all metrics are synthetic examples; no real personal or organization data is included.
The problem

Leads stall quietly.

Teams often know top-line lead volume, but not where buyers slow down — or why some convert fast while others disappear for months.

What teams need to see

  • Where leads drop between stages
  • Which intent signals predict purchase
  • How long each segment takes to convert
  • Which follow-up motions deserve priority
What we built

A dashboard that makes intent visible.

Forgeline shows the funnel, segment mix, time-to-convert, and weekly trend in one dark, interactive view.

5

funnel stages

4

intent segments

1

self-contained file

Forgeline · Lead-Gen FunnelSynthetic demo
4,200Leads
2,849Newsletter
1,797Demo
950Trial
507Buy
The funnel
Lead Submitted · 4,200 Newsletter · 2,849 Demo Watched · 1,797 Trial · 950 Purchased · 507
Overall conversion12.1%
Lead to newsletter67.8%
Newsletter to demo63.1%
Trial to purchase53.4%
Four intent segments

Conversion rate and speed by intent

SegmentVolumeConversionMedian speed
Researching / Lurking47%1.3%~85 days
Actively Scheduling Demos26%8.5%~38 days
Narrowing (3 platforms)18%27.7%~22 days
Final-stage (referral/demo)9%48.3%~7 days
The big insight

Customer intent drives both conversion rate and conversion speed.

Final-stage referral/demo

48.3%

converts in about a week. High intent, short path, direct sales motion.

Researching / Lurking

1.3%

takes roughly three months and rarely buys. Educate, nurture, and avoid over-spending.

Time-to-convert

The faster segments are also the better-converting segments.

Researching / Lurking
~85 days
Scheduling Demos
~38 days
Narrowing
~22 days
Final-stage
~7 days

Scale runs left to right from 0 to 100 days. Bars approximate segment spread; labels show median time-to-convert.

How it was built · 1
1

Frame the outcome: one dashboard, one fictional product, one clear insight.

2

Ask the coding agent to create a complete working artifact, not just snippets.

Build a single self-contained HTML dashboard for a fictional B2B SaaS dev tool called Forgeline. Use D3.js v7, seeded synthetic data, a dark theme, and show how lead intent changes funnel conversion and time-to-convert.
How it was built · 2
3

Pin the house style: dark panels, compact KPI row, gradient header, slate text.

4

Constrain delivery: no build step, no external slide framework, no real data.

Match the dark D3 house style: #0b0f19 background, #111827 panels, #1e293b borders, slate text, and accents #38bdf8, #a78bfa, #22c55e. Keep it as one HTML file and disclose that the data is synthetic.
How it was built · 3
5

Specify the seeded model and exact headline totals before asking for charts.

6

Tell the agent what must be true: segment rates, median speeds, and monotonic funnel stages.

Use seeded synthetic leads. The funnel totals must read: Lead Submitted 4,200 → Newsletter 2,849 → Demo Watched 1,797 → Trial 950 → Purchased 507. Segment medians should be about 85, 38, 22, and 7 days.
How it was built · 4
7

Iterate panel-by-panel: funnel first, segment chart, time spread, then trend.

8

Add a segment filter, tune numbers, verify safety, then publish the file.

Add clickable segment chips that filter every panel. Then verify the displayed totals, conversion rates, and synthetic-data disclosure. Do not use real people, real organizations, or environment-specific identifiers.
Tips for prompting coding agents
🎯

Be specific

Give exact metrics, labels, and success criteria.

🧱

Constrain

Name the stack, file format, palette, and data rules.

🪜

Iterate small

Ask for one panel or behavior at a time.

Verify

Request totals, syntax checks, and safety scans.

🧭

Give references

Point at files that express the house style.

Takeaways

The dashboard was the artifact. The prompt workflow was the leverage.

Intent segmentation changes the story from “how many leads?” to “which buyers are ready?”
Rate and speed belong together: high-intent leads convert more often and faster.
AI coding agents work best when goals, constraints, examples, and verification are explicit.
Synthetic data lets you publish compelling demos without exposing real records.
Explore more

Links

Use , Space, buttons, or click the slide to navigate.

Fictional demo · seeded synthetic data