> ## Documentation Index
> Fetch the complete documentation index at: https://formbricks.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Creating your first charts
> Go from a survey template to a shareable NPS dashboard, and learn why Feedback Records - not survey responses - are what charts are built on.
Charts in Formbricks are not built on survey responses. They are built on **Feedback Records**, which is why a
brand-new workspace shows an empty Analysis page even when your surveys already have hundreds of responses.
This guide walks the whole path once: from the NPS template to a dashboard your team can read at a glance.
## Responses and Feedback Records
A **survey response** is one submission of one survey. It lives with that survey, keeps the exact shape of the
questions you asked, and is the right thing to read when you want to see what a single person said.
A **Feedback Record** is one answer to one question, rewritten into a shape that is the same for every source.
Each record carries when it was collected, which source it came from, what kind of question it was
(`nps`, `csat`, `rating`, `text`, ...) and the answer itself. One response becomes several records - one per
answered question.
That normalization is what makes records reportable:
* **One shape across sources.** NPS from a survey, a CSV upload, or the API all become `nps` records, so a single
chart can cover all of them.
* **Measures that know what a score means.** Because the record is typed, Formbricks can offer `NPS: Score`,
`CSAT: Score` or `NPS: Detractors` as ready-made measures instead of asking you to build the math yourself.
* **Comparisons across surveys.** Two surveys pointed at the same dataset can sit on the same chart and the same
time axis.
* **Enrichment.** Open text is scored for sentiment and emotion as it arrives, which gives you dimensions you never
asked a question about.
Creating records never changes or deletes your responses. The response stays exactly where it was; a record is an
additional, chart-friendly copy.
## What you need before your first chart
Three things have to exist, in this order. Skipping the first is the most common reason Analysis looks broken:
Datasets live at the **organization** level. Go to **Settings → Organization → Datasets** and click
**Create dataset**. Give it a name that describes the body of feedback rather than a single survey - most teams
start with one dataset such as "Voice of Customer".
Under **Workspace access**, pick the workspaces that should be able to read it. A workspace can only be linked
to one dataset, so this choice decides where that workspace's charts get their data.
The dataset then shows up in the list with the number of workspaces that can reach it.
Only Owners and Managers can create datasets. See [Feedback Datasets](/docs/unify-feedback/feedback-datasets) for
archiving and access details.
A source is what turns incoming data into records. Go to **Analyze → Feedback Data → Feedback Sources** and click
**Add feedback source**.
Pick **Formbricks surveys**, then **Select questions**, choose your survey and tick the questions you want to
report on. Formbricks detects each question's type for you, so an NPS question becomes an `nps` record without
any mapping work.
Leave **Import historical responses** on to backfill the responses you already collected. With it off, only
responses submitted from now on become records.
Charts live under **Analyze → Analysis**. Every chart is scoped to the dataset your workspace is linked to.
Once a source is live, the Feedback Sources page tells you which dataset it writes into. Several sources can feed the
same dataset - that is the point.
## Use case 1: An NPS program you can read at a glance
The most common starting point. Create a survey from the **NPS Survey** template (**Surveys → New Survey**, search
for "NPS"), publish it, collect some responses, and connect it as a source using the steps above.
Now open **Analyze → Analysis → Charts** and click **Create chart**. You can describe the chart in natural language
or build it by hand; the three charts below are built by hand so you can see exactly which controls do what.
The AI builder needs **Smart functionality (AI)** enabled for your organization and a configured AI provider. See
[AI Features](/docs/platform/features/ai-features).
### Chart 1: Overall NPS
Your headline number. One value, no dimensions.
* **Chart type:** Big Number
* **Measure:** `NPS: Score`
`NPS: Score` is the standard calculation - `((promoters − detractors) / NPS responses) × 100` - so the number is
comparable to whatever your team quotes today.
### Chart 2: NPS over time
A single score tells you where you are. The trend tells you whether what you did last quarter worked.
* **Chart type:** Line Chart
* **Measure:** `NPS: Score`
* **Add time-based grouping:** on
* **Field:** Collected At
* **Granularity:** Month
* **Date Range:** This year
The default date range is **Last 30 days**, which is why a brand-new trend chart often shows a single point. Widen
it before you conclude the chart is broken.
### Chart 3: Promoters, passives and detractors
The score hides the shape of your base. A ten-point NPS made of loud promoters and loud detractors is a different
business problem from one made of shrugs.
* **Chart type:** Bar Chart
* **Measures:** `NPS: Promoters`, `NPS: Passives`, `NPS: Detractors`
### Put them on a dashboard
Go to **Analyze → Analysis → Dashboards**, click **Create dashboard**, then **Add charts** and pick the three charts
you just built. Drag and resize the widgets into the layout you want; a dashboard-level **Date range** filter sits
above the grid.
## Use case 2: Why the score moved
The NPS template asks a follow-up question - "can you describe the reason(s) for your rating?" - and those answers
land as `text` records. Enrichment scores each one for sentiment as it arrives, so you can chart the *reasons*
without reading them all.
* **Chart type:** Pie Chart
* **Measure:** `Responses`
* **Filter data:** `Question` equals your follow-up question
* **Group data → Group By:** `Sentiment`
The filter matters. Without it the chart also counts the NPS number records, which have no sentiment, and roughly
half of your pie becomes a "Not enriched" slice.
From here, open **Analyze → Feedback Data → Feedback Records**, filter to the negative comments and read the
verbatims behind the slice. See [Enrichment](/docs/unify-feedback/enrichment) for what else is scored.
## Use case 3: One dashboard across surveys
This is where records earn their keep. Add a second survey - a CSAT survey after checkout or a support
interaction - and connect it as a **second source into the same dataset**. Both surveys now produce records on the
same time axis, so they can share a chart.
* **Chart type:** Line Chart
* **Measures:** `NPS: Score`, `CSAT: Score`
* **Add time-based grouping:** Collected At, Month, This year
Both measures are percentages of a kind, so they read sensibly on one axis: `NPS: Score` runs from −100 to 100 and
`CSAT: Score` is the share of responses rated 4 or 5 on the 1-5 scale. When relationship-level NPS and
transactional CSAT diverge, that gap is usually the interesting part.
## When a chart looks empty
The workspace has no dataset. Create one under **Settings → Organization → Datasets** and add this workspace under
**Workspace access**.
No source is writing into it yet, or the source was created with **Import historical responses** switched off.
Check **Analyze → Feedback Data → Feedback Sources**, and confirm the banner under the table names the dataset you
expect.
Widen the **Date Range**. It defaults to **Last 30 days**, and a trend chart at **Month** granularity over 30 days
has almost nothing to draw.
You are counting records that carry no value for that dimension - usually numeric records in a chart grouped by
sentiment. Add a **Filter data** condition on `Question` or field type to narrow the chart to the records the
dimension applies to.
## Next steps
Chart types, the AI builder, and dashboard permissions.
CSV uploads, API ingestion, and field mapping.
Sentiment, emotions, and language consolidation.
Cluster open text into themes.