THE SEMANTIC CANVAS FIELD GUIDE / 7 MIN READ
The dashboard
is the beginning.
How governed metrics become a clearer story—and a better next question.
Follow the transformation ↓Revenue is growing.
Is the growth healthy?
Follow the same business through a rough dashboard, an AI-assisted redesign, and an executive review. Scroll to see the view change.
Ending MRR
What drove the $4k increase?
- New business+$10k
- Expansion+$8k
- Contraction−$4k
- Churn−$10k
$100k in July → $104k in August
A collection of metrics is only a beginning.
Ending MRR
What drove the $4k increase?
- New business+$10k
- Expansion+$8k
- Contraction−$4k
- Churn−$10k
$100k in July → $104k in August
Imagine the monthly leadership review. Revenue is up. Someone asks whether the business is getting healthier. The dashboard contains a revenue total, a retention number, and several charts. Yet the room is still assembling the answer.
That is a design problem with analytical consequences. When related evidence is separated, readers compare the wrong things or stop at the most reassuring number. In our example, new business and expansion add $18,000 while churn and contraction remove $14,000. The movement chart explains the $4,000 net gain, but the scattered layout leaves the reader to connect those losses to retention.
The goal is to make the important relationship easier to notice. The same illustrative data will remain underneath every version you see here.
Design test: can a reader identify the central question before opening a chart menu?
Tell the dashboard what job it has.
Reimagine with AI
✦ Powered by AIThis guided example uses a prepared proposal. In the builder, your configured AI provider inspects the canvas and catalog.“Make it look better” leaves the important choices unstated. A useful brief names the audience and the decision: help leadership understand whether growth is healthy and where to investigate next.
Reimagine with AI starts from the current canvas and available governed metrics. Its proposal can change queries, chart types, comparisons, grouping, sizes, and supporting content. The definition of revenue is still supplied by the semantic model. The design brief determines how that evidence is presented.
The live builder can inspect catalog information and query data before proposing changes. This prepared walkthrough compresses that process so you can inspect the outcome. It makes no claim about generation speed, and it does not contact your AI provider.
A useful brief combines an audience, a question, and the decision the evidence should support.
Make the relationship visible.
How recurring revenue is changing
What is happening to retention?
94% means the existing revenue base shrank 6%, including expansion.
New business offsets losses in the existing base
- New business+$10k
- Expansion+$8k
- Contraction−$4k
- Churn−$10k
$100k in July → $104k in August
The proposed dashboard places ending revenue, retention, and new business together. Those cards establish the result and the tension: total revenue grew, but the existing revenue base contracted.
An 18-month revenue trend then gives the latest value a trajectory. A retention trend adds the history of the existing revenue base, while the movement breakdown remains visible to explain this month’s gains and losses. These are changes to what the dashboard can explain, as well as to how it looks. A short reading guide connects the headline numbers to the evidence below.
Use Before and Proposed to compare the two versions. The most important test is whether the new arrangement makes a useful question easier to ask. A denser dashboard or a longer list of charts is not necessarily an improvement.
Read in order: result → pattern → supporting evidence → next question.
Preview is where your judgment enters.
AI proposes the changes. You choose what to apply. Each change here has its own undo.
An AI-generated design is a proposal. In the builder, the preview renders the proposed charts against connected data before applying them. Validation checks the metric selections and proposed composition. The current dashboard remains intact until you apply the change.
After applying a redesign, the next adjustment may be straightforward. A compact executive view suits a regular review; Tell a story gives the evidence more room for a presentation. Try both stages here, then undo them one at a time.
The design should remain editable. Pinned sections preserve intentional placement, existing filters remain attached to their tiles, and changed data or filters require a fresh preview. A team should be able to keep what works and refine what does not, without treating the first proposal as final.
A useful AI workflow makes the proposed change inspectable, editable, and reversible.
A better story still needs a defined calculation.
$104k revenue.
94% retention.
New business can grow total revenue while the existing customer base contracts.
- Starting recurring revenue
- $100,000
- Expansion
- +$8,000
- Contraction
- −$4,000
- Churn
- −$10,000
- Retained + expansion
- $94,000
- New business
- +$10,000
- Ending recurring revenue
- $104,000
In August, the sample begins with $100,000 of monthly recurring revenue. Expansion adds $8,000. Contraction removes $4,000, and churn removes $10,000. The existing revenue base therefore contributes $94,000: net revenue retention is 94%.
Another $10,000 of new-business revenue brings ending MRR to $104,000. Total MRR is up 4% even though the existing base shrank. These statements are compatible because retention excludes new business. Putting the metrics together reveals the distinction.
The definition is only one part of trust. Reporting dates, filters, source coverage, and access scope determine what a reader can compare. These charts use illustrative closed monthly observations through August 2026. In a real dashboard, a partial current month needs a coverage check. The figures alone do not identify affected accounts or establish why they left.
Know the metric, the reporting period, and the boundary of what the evidence supports.
Keep a path from the signal to the next question.
Net revenue retention
Below this example’s expected range of 98.8%–109.2%.
17 prior monthly observations · Heuristic rangeThe dashboard has done its first job when the team can name the next investigation: which accounts explain churn and contraction, and what should change before the next forecast? That work needs more evidence than a trend line.
A personal chart watch can help bring attention back to the question. In the builder, save the dashboard, enable chart notifications, preview an unusual-change check, and choose its frequency. The sample’s August retention is below a heuristic range calculated from the prior 17 months. That is a signal to investigate, not a diagnosis.
Checks require the running server and valid data access. Updates appear inside Canvas. Chart conversations keep discussion beside the saved evidence. The complete workflow is a repeatable review: inspect the signal, ask the next question, improve the view, and return when something changes.
The outcome is a clearer decision process, with a dashboard that can evolve alongside it.
Let the question lead.
How recurring revenue is changing
What is happening to retention?
94% means the existing revenue base shrank 6%, including expansion.
New business offsets losses in the existing base
- New business+$10k
- Expansion+$8k
- Contraction−$4k
- Churn−$10k
$100k in July → $104k in August
Try the controls yourself, or watch the complete transformation in 90 seconds.
Method, sample data, and product scope
The SaaS values are illustrative and are shared by the watch, hands-on, and reading formats. Dollar amounts in the reconciliation add up to the headline cards. The 18 monthly observations end in August 2026. The unusual-change example uses the builder’s balanced, trend-aware scaled-MAD rule with 17 prior observations.
The walkthrough demonstrates the updated local builder’s AI redesign, layout refinement, and undo workflow. AI requires a configured provider; proposals vary by data and goal. Sample buttons use prepared changes, do not make live AI requests, and do not enable real watches. Chart notifications are personal, in-app updates. Ingest provides staged source review, and workspace access still governs the real app links.