Export to CSV: taking your LHO data into the tools you already use
Your Databricks cost analysis rarely stops in one place. The same numbers feed the annual budget, the chargeback model that already holds cloud and SaaS spend, the team dashboard, the board pack.
Nearly every table and every cost chart in Lakehouse Optimizer has an export control on it, and the file it produces carries the filters you had applied when you clicked — so LHO's analysis lands in those systems without an integration project, API credentials, or a wait on a platform team.
This is not a complete catalogue of every export button. It is something more useful: the three shapes the export control takes, so you can recognise it on any page you land on — followed by eight worked examples from the pages teams use most. Once you know the shapes, you will find exports on pages this article never mentions.
- 1 Spotting the export control on any page
- 2 Reading the dialog before you click
- 3 Eight places teams export from
- 3.1 1 · Cost Over Time, by tag — the chargeback export
- 3.2 2 · Cost Over Time, by feature — where the money actually goes
- 3.3 3 · Workspace Cost — the input to a per-team model
- 3.4 4 · Forecasting — the export for budget conversations
- 3.5 5 · Workflows — job-level cost for the teams that own the jobs
- 3.6 6 · Recommendations — the savings backlog, as a worklist
- 3.7 7 · Incidents — a record of what was caught, and when
- 3.8 8 · Vendor Consolidation — evidence for a migration business case
- 4 Connecting exports to other systems
- 5 Good to know
Spotting the export control on any page
The control takes one of three shapes, depending on the kind of page.
Shape | Where to look | Tooltip | Screenshot |
|---|---|---|---|
Next to the page title | A small upload arrow right of the heading, on table-based pages. The most common shape — and it follows you as you drill down into a workflow, a run, then the clusters underneath. | Export to CSV | 
|
In a chart card's toolbar | On chart-based pages the control sits in the top-right of the chart panel, not near the heading. Look only at the page title here and you will wrongly conclude there is no export. | Export CSV |
|
Behind the ⋮ menu | Some pages tuck it into the three-dot menu beside the title. Check there before deciding a page has none. | Export to CSV |
|
A fourth, rarer shape exists for administrators: on some settings screens the icon sits in the table's own header row and downloads immediately.
A greyed-out control means the view has no data for your current filters; widen the date range and it becomes available.
Reading the dialog before you click
Most exports open a short dialog first. Treat it as a receipt — it states the date range, the workspace, the tab you are on and, where relevant, the number of rows.
Scope is the one thing to check. Table exports write only the rows on the page you are viewing, so a table showing 25 of 300 rows gives you 25. Chart and summary exports write the whole selected date range.
One habit works everywhere. Before exporting a table, raise the rows-per-page control to cover all your rows, then check the count in the dialog matches what you expect.
Everything else — subscriptions, workspaces, date range, the tab, the Filters drawer — comes along automatically. Filenames record it too:
workloads_workflows_jobs_2026-07-01_to_2026-07-29_analytics-prod_page_size_25_page_1.csvThe workspace, range and page size are all in the name, so months later you can still tell which filter produced which file.
Eight places teams export from
Each of the following is a page where export earns its keep. They are examples, not the full list.
1 · Cost Over Time, by tag — the chargeback export
Spend Insights → Cost Over Time → Tag · chart toolbar shape
If you only ever use one export, this is the one. It breaks spend down by any tag you govern — business unit, cost centre, project, environment — and turns that view into the allocation table your finance or FinOps process actually runs on.
This is the one export that asks you a question, and the answer matters:
Top N values (remaining grouped into "Other") mirrors what the chart shows — the largest values individually, everything else collapsed into one row. Good for a summary slide.
All values without grouping into "Other" writes every tag value as its own row. This is the one you want for chargeback, because "Other" is not a cost centre and nobody can be billed for it.
Two controls on the page shape the file as well. Top N values to display sets the number offered in that first option. And Roll up case-variant values carries into the export: with it on, prod, Prod and PROD merge into a single row rather than fragmenting one cost centre across three. On a real estate with imperfect tag hygiene, that toggle is the difference between a usable allocation file and one somebody cleans by hand every month.
Breadth here is controlled by the Top N / All values choice in the dialog — not by the row checkboxes in the table, which govern the chart and the table on screen.
Worth exporting alongside the tagged values: the Untagged and Shared rows. Untagged spend is the honest measure of how far your tagging discipline has to go, and showing it as a line in the allocation model is usually what finally gets it fixed.
2 · Cost Over Time, by feature — where the money actually goes
Spend Insights → Cost Over Time → Feature · chart toolbar shape
The same page, on its Feature tab, breaks the same spend down by what it was spent on — jobs, all-purpose compute, pipelines, SQL warehouses, AI, platform services — with cloud and Databricks cost separated and month-over-month change alongside.
Here the row checkboxes do carry into the file. Untick the categories you are not discussing and the export contains only the ones you left ticked — useful when you want a clean two-line comparison rather than a fourteen-row table somebody has to filter. The cost-type selection carries too: choose Databricks-only or cloud-only and the file is written to match.
This is usually the export behind a monthly cost review: same filters, same tabs, exported on the same day each month, appended to a running sheet that shows the trend.
3 · Workspace Cost — the input to a per-team model
Spend Insights → Workspace Cost · ⋮ menu shape
This page answers "which workspace spent what", broken out by feature, resource or provider.
Where many organizations map workspaces to teams or environments, this export is the second half of the chargeback picture: tags give you the logical allocation, workspaces give you the structural one, and cross-checking the two is how you find spend that belongs to somebody who does not know they own it.
4 · Forecasting — the export for budget conversations
Spend Insights → Forecasting · chart toolbar shape
This is the only export that contains projected spend next to budget, commitment and actuals. That combination is what makes it the natural attachment to a renewal discussion, a quarterly business review, or an internal "are we going to blow the budget" conversation.
The dialog is short, and its Export Option follows the granularity you were viewing — monthly, weekly or daily.
Because the forecast usually needs to sit next to the rest of the annual plan, this export tends to be the most frequently repeated one in the product — often monthly, into the same workbook.
5 · Workflows — job-level cost for the teams that own the jobs
Workloads → Workflows · title icon shape
This page lists every job with its cost, run statuses, duration, data usage and analysed runs.
This is the clearest illustration of the page-scope rule, because the dialog tells you exactly how many rows you are getting.
Teams use this to put the cost of the jobs a group owns straight into that group's own planning and reporting. The same export control follows you further down — into a single workflow's run history, then a single run's tasks and clusters — so when a specific job needs investigating, the underlying rows come out at whatever grain the conversation needs.
6 · Recommendations — the savings backlog, as a worklist
Optimize → Recommendations · ⋮ menu shape
This page ranks the optimization opportunities LHO has found, with the suboptimal cost attached to each. Exporting it turns an in-product list into something a team can actually work.
The pattern here is simple and common: export the list, add an owner column and a target date, paste it into your tracker or share it as a sheet, then re-export a month later to see what moved. LHO decides what is worth doing; the CSV is what lets a team divide it up and be held to it. The export follows the tab you are on, so you can pull cost-savings and performance opportunities as separate worklists.
7 · Incidents — a record of what was caught, and when
Optimize → Incidents · ⋮ menu shape
This page has two tabs, and each exports separately. Cost Control covers cost spikes measured against your policy thresholds.
Efficiency covers the other kind — idle warehouses, over-provisioned drivers, all-purpose compute running job workloads.
One naming detail to expect: the Efficiency tab exports with the subtype Performance — you will see it in the dialog and in the filename.
Exported incidents make a good monthly operations record — what was detected, against which threshold, and how quickly it was dealt with. Teams that report on platform health to a steering group tend to keep this one on a schedule.
8 · Vendor Consolidation — evidence for a migration business case
Optimize → Vendor Consolidation → Migration Opportunities · ⋮ menu shape
This page shows which workloads depend on third-party systems, with the data transferred, the vendor execution time and the wait time each one costs you.
A consolidation proposal lives or dies on evidence, and this is the export that supplies it: the per-workload numbers behind "we are paying twice, and waiting". Note that the export is on the Migration Opportunities tab; the Transferred Data & Cost tab beside it is a visual summary and has no export of its own.
Connecting exports to other systems
Column headers stay stable between exports, so the same destination keeps working month after month. That predictability is what turns a manual download into a routine.
Destination | What to send | What it gives you |
|---|---|---|
Spreadsheets and finance models | Cost Over Time by tag + Workspace Cost | Export on the same day each month, paste into the next rows, and existing formulas and pivots pick it up. These two cover a full allocation model — the logical split and the structural one. |
BI tools | Any export, into a watched folder | Power BI, Tableau and Looker read CSV directly — point a dataset at the folder and the existing refresh does the rest. |
A warehouse, or back into Databricks | Cost Over Time, on a schedule | Loaded into a Delta table, exports become a cost time series you can join to headcount, product lines or revenue. Cost per team is a spreadsheet exercise; cost per customer usually needs that join. |
Chargeback and showback | Cost Over Time by tag, All values | Internal billing needs a periodic table of "this dimension, this period, this amount" — exactly this export's shape. |
Ticketing and planning tools | Recommendations, plus an owner column | Jira, Azure DevOps, ServiceNow and Asana import CSV in bulk, so the savings backlog becomes tracked sprint work. |
Audit and board packs | Whichever export backs the figure | Cost data is restated as providers settle billing, so a dated export is a point-in-time record — what the number was the day you looked. |
Three habits make this smoother. Keep the same cadence and the same filters, because comparability beats freshness. LHO aggregates on UTC, so align reporting periods to it or month boundaries drift. And rename the tag and Workspace Cost exports if you archive them — their filenames do not record the workspace.
Good to know
Export is deliberately simple, and it is worth being clear about the edges.
Table exports are page-scoped — set the rows-per-page control first. Exports are manual: there is no scheduling and no "export everything" button, so a monthly routine is a calendar reminder rather than an automation. The format is CSV only. Not every page has an export — visual summary pages and trend tabs generally do not, and the ⋮ menu is worth checking before you conclude one is missing. On AWS deployments the Subscription field does not appear, because it is an Azure concept.
This article describes Blueprint Lakehouse Optimizer as of version 3.4.0. Later releases may change features, terminology, or screenshots.