TL;DR
Custom Charts let Mitzu's Analytics Agent render ad hoc visualizations inside the conversation when the answer exceeds what any single Mitzu insight can show. Triggered automatically for derived ratios (DAU/MAU), period-over-period comparisons with percentage change, and overlaid funnel conversion trends. The agent does the cross-insight calculation itself, then renders a collapsible card with a violet "Custom" badge directly in the answer.
Custom Charts is a new capability inside Mitzu's Analytics Agent. When the answer to a question exceeds what any single Mitzu insight can show, the agent now runs the calculation across multiple insights and renders the result as an ad hoc visualization directly in the conversation.
The Analytics Agent could already calculate and explain complex answers — ratios, period comparisons, trends across multiple metrics. What it couldn't do was show you. Some insights just land differently as a chart. Custom Charts closes that gap.
Custom Charts are ad hoc — built for the answer in front of you, not saved anywhere. The underlying Mitzu insights the agent used to get there are surfaced alongside the chart, saveable to any dashboard as always.
1. What are Custom Charts?
A Custom Chart is an ad hoc visualization the Analytics Agent renders inside an answer when no single Mitzu insight can express the result on its own. The agent pulls data from one or more underlying insights, performs a calculation across them (a ratio, a percentage change, an overlaid trend), and embeds the resulting chart as a collapsible card with a violet "Custom" badge directly in the conversation.
Two things to keep in mind:
- Ad hoc by design. Custom Charts are built for the question in front of you. They are not saveable as insights, not addable to dashboards, and not openable in the explore view.
- The underlying insights are saveable. The individual Mitzu insights the agent assembled into the custom chart are surfaced alongside it. Each one can be saved, added to dashboards, or shared — the normal way.
2. Why visualization matters for an analytics agent
An analytics agent that only returns numbers in prose hits a ceiling fast. "DAU/MAU is 0.34, up from 0.31 last quarter" is accurate but harder to read than a line chart that shows the curve and the inflection. Most real product analytics questions are about relationships and movement, and relationships and movement are what charts encode well.
Before Custom Charts, the Analytics Agent could describe these answers and link to the underlying insights, but it couldn't compose them into a single visualization. That meant the user often had to flip between two saved insights and do the comparison by eye. Custom Charts removes that step for the cases where the chart only needs to exist for one answer.
3. When the agent renders a Custom Chart
The agent decides whether a question warrants a Custom Chart based on the shape of the answer. Three common triggers, documented in the Analytics Agent reference:
| Trigger | Example question | What the chart shows |
|---|---|---|
| Derived ratios | Show me DAU/MAU ratio over the last 90 days. | A single line: daily DAU divided by trailing MAU, plotted over time. |
| Period-over-period comparisons | Show me signups by plan tier in Q3 vs Q4 with the % change. | Two bar groups side by side, with the percentage delta annotated per tier. |
| Funnel trend overlays | Plot signup-to-activation conversion rate week over week alongside the trial-to-paid conversion rate. | Two conversion-rate curves on the same chart, sharing a weekly time axis. |
If the question can be answered by a single Mitzu insight, the agent uses that insight directly. Custom Charts appear only when the answer is genuinely a composition over more than one insight.
4. How a Custom Chart is built under the hood
The mechanic is the same one that powers the rest of the Analytics Agent. The agent assembles analysis specifications (funnel definitions, retention parameters, segmentation breakdowns). Mitzu's deterministic query engine turns each specification into SQL and executes it against the warehouse. The agent then composes the per-insight results into the final calculation — a ratio, a percentage change, an overlay — and renders the chart.
Two consequences of that design:
- Each underlying insight is reproducible. The same specification always produces the same SQL and the same numbers. The custom chart is a presentation layer over deterministic results, not a free-form output.
- You can audit any data point. Click into the underlying insight, see the SQL the engine generated, verify the methodology. The agent didn't author the query; the engine did.
5. Why Custom Charts are ad hoc by design
Custom Charts are deliberately not saved as insights. The reasoning is straightforward: a dashboard artifact should have a clear definition someone can review, edit, and version. A composition the agent ran for one question is not that — it's an answer, not a metric.
Treating the chart as an answer keeps two surfaces clean:
- The conversation surface — Custom Charts render inline, answer the question, and live with the conversation. They can be screenshotted, copied, linked, but they don't pollute the workspace's saved-insight catalog.
- The dashboard surface — saved insights and dashboards remain a curated set of named metrics. The underlying insights the agent used are saveable individually, so anything that should be promoted to a dashboard can be — explicitly, by the analyst.
If you find yourself asking the same custom-chart question twice, save the underlying insights and compose them into a dashboard view. That's the explicit path.
6. Concrete examples of Custom Charts in use
DAU/MAU stickiness over time
Asking "show me DAU/MAU ratio over the last 90 days" requires two distinct insights — a daily-active-users series and a trailing monthly-active-users series — and a per-day division. The agent assembles both, computes the ratio, and renders a single line chart inline. The DAU and MAU insights are surfaced beneath the chart, both saveable.
Quarter-over-quarter signups by plan tier
"Signups by plan tier in Q3 vs Q4 with the % change" needs two segmentations plus a per-tier delta. The custom chart renders both quarters side by side with the percentage change annotated for each tier. The underlying Q3 and Q4 segmentations are surfaced and saveable as standalone insights.
Two conversion funnels on one trend chart
For "plot signup-to-activation conversion week over week alongside trial-to-paid," the agent builds two funnels, extracts the weekly conversion rate from each, and overlays them on a shared time axis. Useful for spotting whether two stages of the lifecycle move together or diverge. Both funnels are saveable insights on their own.
7. How Custom Charts interact with Planning Mode
Custom Charts compose with the rest of the Analytics Agent surface. In Planning Mode, a multi-step investigation can include a step that produces a custom chart — for example, "compare the conversion curves for paid and organic users on one trend chart" as the final step of a retention deep dive. The plan is approved up front; the chart appears inline when the plan executes.
For one-shot questions, no plan is needed — the agent decides whether a custom chart is the right shape for the answer and renders it directly.
8. Who Custom Charts are for
- Product managers — DAU/MAU stickiness, feature-adoption trends across cohorts, conversion overlays at lifecycle stages. Questions where the answer is a relationship between metrics, not a single number.
- Growth and marketing leads — campaign period-over-period deltas, channel-mix comparisons, signup-quality ratios. The percentage change is usually what matters; the custom chart renders it inline instead of requiring a manual spreadsheet pivot.
- Data analysts — quick exploratory comparisons before deciding whether something deserves a saved insight. Custom Charts are scratchpads with real warehouse data behind them.
- Founders and operators — quarterly board-deck inputs, executive-summary ratios. Ask once, screenshot, move on.
9. Warehouse readiness — the prerequisite
Custom Charts run inside the Analytics Agent, which means the same prerequisites apply: event data already lives in a modern cloud warehouse — Snowflake, BigQuery, Databricks, Redshift, or ClickHouse — and Mitzu's Configuration Agent has indexed it into the semantic layer. If event data still sits in a third-party vendor silo without warehouse export, Mitzu is not the right fit yet.
Frequently asked questions
What are Custom Charts in Mitzu?
Custom Charts are ad hoc visualizations Mitzu's Analytics Agent renders inline when an answer requires composing more than one Mitzu insight. Typical cases are derived ratios like DAU/MAU, period-over-period percentage comparisons, and overlaid funnel conversion trends. They appear as collapsible cards with a violet "Custom" badge inside the agent's response.
Can I save a Custom Chart as an insight or add it to a dashboard?
No. Custom Charts are ad hoc — they cannot be saved as insights, added to dashboards, or opened in the explore view. The underlying Mitzu insights the agent used to build the custom chart are surfaced alongside it and are saveable to dashboards the normal way. If a recurring question is calling for the same custom chart twice, save the underlying insights and compose them into a dashboard explicitly.
When does the Analytics Agent decide to render a Custom Chart?
When the answer can't be expressed by a single Mitzu insight. The most common triggers are derived ratios (one metric divided by another), period-over-period comparisons that require a calculated delta, and multi-series overlays on a shared axis. If a single insight already answers the question, the agent uses that insight directly.
Does the Analytics Agent write SQL to produce a Custom Chart?
No. The agent never writes SQL — not for regular insights and not for Custom Charts. Each underlying insight is built from an analysis specification (funnel, retention, segmentation, journey) that Mitzu's deterministic query engine turns into SQL. The custom chart is a composition over those deterministic results — a ratio, a delta, or an overlay — calculated by the agent and rendered inline.
Can I see the SQL behind a Custom Chart?
Yes — indirectly. Each underlying insight is reviewable on its own, with the SQL the deterministic engine generated. Open the insight from the card surfaced alongside the chart, inspect the SQL, verify the methodology. Custom Chart calculations on top of those insights (ratios, deltas, overlays) are documented in the chart's own metadata.
Do Custom Charts work in Slack or only in the in-app agent?
Custom Charts render in the in-app Analytics Agent. For surface-by-surface availability, see the Custom Charts documentation.
How are Custom Charts different from saved chart insights?
A saved insight is a named, reviewable, version-able artifact tied to a single analysis specification. A Custom Chart is an answer — a one-shot composition the agent produced for a specific question. Saved insights belong on dashboards and in the data catalog; Custom Charts belong in the conversation that produced them.
Is Custom Charts part of the standard Mitzu plan?
Custom Charts are included in the Analytics Agent and available in workspaces with the agent enabled. See Mitzu pricing for plan details.
Related reading
- Planning Mode in Mitzu — review the agent's plan before it runs
- Why agentic analytics needs a product-analytics-shaped semantic layer
- How Mitzu generates verified, trustworthy SQL
- What is agentic analytics?
- Warehouse-native analytics: benefits and how it works
