TML read as a specification
Exported TML tells us exactly which columns, joins and formulas a worksheet carries. That turns the rebuild into defined, estimable work instead of screenshot-matching against a live tenant.
Replacing search-first analytics
A ThoughtSpot to Power BI migration: TML rebuilt by hand, and a straight answer on search
We read your TML to specify the rebuild, turn worksheets and models into one governed semantic model, recreate the ten to twenty Liveboards people actually open, and give your business users a real replacement for typing questions — Copilot on a model prepared for it.
Representative dashboard — sample data
The question nobody scopes
Rebuilding the boards is easy. Replacing the search box is the decision.
Two conversations decide whether this move goes well. The first is technical and tractable: TML gives you a readable specification of every table, join, formula and Liveboard, so the rebuild is well-defined work rather than archaeology. The second is the one that gets skipped — a group of business users has spent two years typing questions into a search bar, and somebody has to tell them truthfully what they will do instead.
We answer that in the scope estimate, not at go-live. Most typed questions turn out to be a handful of repeating shapes that belong on a report page with slicers; the genuinely ad-hoc remainder is what Copilot is for, and Copilot only behaves if the semantic model underneath it was built for it.
What the rebuild covers
- TML as the spec
- Worksheets → semantic model
- Liveboards → report pages
- Saved Answers triaged
- RLS rules rebuilt
- Search → Copilot
Try us risk-free for 4 hours
Buy a block of prepaid Flex hours and, if you're unhappy within the first 4 hours, we refund you in full — no questions asked. A TML export review plus a Liveboard usage pull is the ideal first task: it costs very little and it tells you how much of your estate is worth moving at all.
How we approach it
What a deliberate rebuild gets you that a lift-and-shift does not
The point of leaving a search-first platform is not to own a smaller copy of it. It is to end up with definitions your finance team will sign.
The search question answered first
Before any page is built we categorise what users actually type, then show which questions become slicers on a page and which need Copilot on a prepared model. No one finds out at launch.
Ruthless triage of saved Answers
Search-first tools generate saved Answers by the hundred, most viewed once by one person. We pull usage and recommend retire, merge or rebuild on each, so you pay to move what matters.
One model, agreed definitions
Worksheet formulas become DAX measures on a single governed model. Where two worksheets defined margin differently, you pick one — that argument is cheaper now than after go-live.
Your warehouse is untouched
Snowflake, BigQuery, Redshift, Databricks or Synapse stays exactly as it is. This is a modeling and reporting change, which is usually what makes the internal approval straightforward.
Reconciled, not approximated
Each rebuilt page is run against its Liveboard original on identical filters and dates. Variances are explained individually, and occasionally the new number is the correct one.
Why does TML make this easier than most migrations?
ThoughtSpot Modeling Language is a scriptable YAML format used to export, edit and move metadata objects — models and worksheets, views, tables, saved Answers and Liveboards. Power BI cannot import it, and anyone claiming a converter is overselling. Its real value is that it removes guesswork: instead of reverse-engineering behaviour from a running tenant, we read the joins, the formula expressions and the Liveboard layout as text, and price the rebuild from that.
Worksheets — called models in more recent releases — are the closest thing ThoughtSpot has to a semantic layer, and they are where the interesting differences live. One worksheet usually flattens several tables into a single searchable surface. Power BI wants the opposite: a fact table surrounded by dimensions, with relationships declared. So the rebuild is a reshape, not a copy, and it is the step that decides whether the finished model is fast.
How do ThoughtSpot objects map to Power BI?
This is the map we estimate against. Where something has no honest equivalent, the table says so rather than inventing one.
| ThoughtSpot | Power BI equivalent | What changes in practice |
|---|---|---|
| Worksheet / model (TML) | Semantic model | A flattened searchable surface is reshaped into a star schema with declared relationships |
| Table & join definitions in TML | Tables plus model relationships | Join type and cardinality become explicit, and ambiguous paths have to be resolved |
| Worksheet formula | DAX measure | Aggregation and filter logic is rewritten once and reused by every page |
| Liveboard | Report page | Tiles become visuals on a laid-out page; Liveboard filters become slicers with explicit scope |
| Saved Answer | A visual, or nothing at all | Most are retired on usage evidence; the rest are folded into an existing page |
| Search bar (natural-language query) | Copilot for Power BI, on a prepared model | Answer quality depends on model naming, descriptions and agreed measures — it is preparation work, not a switch |
| SpotIQ automated insights | No like-for-like equivalent | Replaced in practice by designed comparison and variance pages, plus Copilot follow-up questions |
| Monitor / KPI alerts | Power BI alerts and report subscriptions | Thresholds are re-declared against the new measures; delivery moves to email subscriptions |
| Row-level security rule | Row-level security role | Group-based rule expressions usually consolidate into one dynamic, testable role |
| Embrace live connection to the warehouse | DirectQuery, Import, or a composite model | Storage mode becomes a per-table decision instead of a platform-wide one |
Microsoft's documentation on DirectQuery in Power BI (opens in new tab) covers the trade-offs we weigh per table, and our guide to DirectQuery vs Import mode is the plain-English version. The measure and relationship work sits on our Power BI data modeling and DAX page, and security rebuilds follow Power BI row level security setup.
What actually happens to the people who type questions?
This deserves a straight answer, because it is the one thing a search-first platform does that a report tool does not do the same way. Power BI has offered natural-language questions for years through Q&A, but Microsoft has now published that Q&A experiences are going away in February 2027 (opens in new tab) and recommends Copilot instead. Designing a 2026 migration around Q&A would be building on a deprecation notice, so we do not.
Copilot for Power BI is the destination, and it is genuinely good on a model that was prepared for it — friendly table and column names, written descriptions, agreed measures, hidden key columns. It is unconvincing on a model carrying raw source-system names and no descriptions, which is exactly the model a hurried migration produces. Microsoft's Copilot in Power BI overview (opens in new tab) sets out what it does; our Power BI Copilot consultant page covers the capacity requirement and the model preparation honestly, including when the answer is that you are not ready yet.
The other half of the answer is less glamorous and more effective. When we categorise real search logs, the majority of questions are a dozen recurring shapes — this metric, by that dimension, for this period, versus last year. Those belong on a page with slicers and a properly built date table, where the answer is instant, consistent and does not depend on how the question was phrased.
How big is this really, and what does it cost?
Smaller than the Liveboard count suggests. Search-first adoption produces volume — saved Answers accumulate, Liveboards get cloned per team — and usage data almost always shows a core of ten to twenty boards carrying the decisions. We scope to that core deliberately. A migration that promises to move everything is a migration that spends your budget recreating content nobody opened last quarter.
Work runs on prepaid Flex hours that never expire: Starter $2,500, Growth $4,600, or Power $8,000, with hourly rates from $100 on the largest block. The TML review and usage triage normally fits inside a Starter block and produces an item-by-item estimate for everything after it. Individual boards typically ship within 3–5 business days of approval, and running both platforms through one reporting cycle is worth the overlap — users compare on live warehouse data, and the ThoughtSpot contract ends with nothing stranded.
Who this is for
Analytics and finance teams running ThoughtSpot on a cloud warehouse, with a Microsoft 365 estate already paid for and a renewal that no longer clears the value bar. It suits organisations willing to make definition decisions in transit; it suits nobody who wants an identical copy, because an identical copy of a search-first estate is not a thing Power BI is trying to be. If more than one platform is in play, the broader Power BI migration service covers the Excel and SSRS routes, Looker to Power BI migration is the closest sibling for teams also leaving a code-modelled cloud platform, and Domo to Power BI migration handles Beast Modes, Magic ETL and PDP policies.
What you get
- TML-derived object and formula inventory
- Usage-based retire / merge / rebuild calls
- A star-schema semantic model with tested DAX
- A written plan for replacing search
- Reconciliation against the Liveboard originals
10–20
boards that usually matter
3–5
business days per board
You
own everything
Start with the TML review and usage triage, then decide the rest with real numbers in front of you. See Flex pricing.
How it works
From TML export to a governed reporting estate
- 1
Export, read and triage
You export TML for your worksheets, Liveboards and Answers; we turn it into an object and formula inventory, pull usage for every board, and recommend retire, merge or rebuild on each. You approve the list before a page is built.
- 2
Model first, pages second
Surviving content is rebuilt on one star-schema semantic model over your existing warehouse, with worksheet formulas rewritten as DAX and the model named, described and shaped so Copilot can answer against it later.
- 3
Reconcile, enable, hand over
We compare every page to its Liveboard original, explain the variances, rebuild the security rules as tested roles, turn on the agreed search replacement, publish to your workspace and take your users through it.
Representative perspective
We had over three hundred saved Answers and assumed that was the scope. Usage said nine boards carried every decision we make, and the budget conversation changed that afternoon.Director of analytics, healthcare services group
Representative perspective
Being told upfront what our users would lose, and what they would get instead, was worth more than another vendor promising it would feel the same.Finance director, mid-market SaaS
Get a real number before your renewal date
Tell us how many Liveboards and worksheets you have, which warehouse sits underneath, and when the contract comes up. We'll scope the TML review and usage triage in hours so the decision is made on evidence.
FAQ
What teams ask us before they commit
Not directly. TML is ThoughtSpot's own YAML format and Power BI cannot read it. What TML is genuinely good for is reading: we export it to get an exact, readable specification of your tables, joins, formulas and Liveboard layouts, then rebuild each object deliberately in Power BI and reconcile the output.
Two things, and we are honest that neither is identical. Copilot for Power BI answers questions in natural language on a well-prepared semantic model, and well-designed report pages with slicers cover most repeat questions outright. Microsoft has also announced that the older Power BI Q&A experiences retire in February 2027, so we design for Copilot.
No. Whatever warehouse ThoughtSpot queries through Embrace stays exactly where it is. Power BI connects to it natively, and we choose DirectQuery or Import per table rather than for the whole model, so the migration changes the modeling and reporting layer only.
Far fewer than you have. We pull usage for every Liveboard and saved Answer, and the list that survives is typically ten to twenty that genuinely drive decisions. Rebuilding the rest costs real money to recreate content nobody opens, so we recommend retire or merge and let you approve each call.
They are rebuilt as Power BI row-level security roles and tested by viewing the report as each role. ThoughtSpot filters rows with rule expressions evaluated against group membership; Power BI filters with a DAX expression on a role. The logic usually simplifies, because a single dynamic role replaces many per-group rules.
Related work
This page is specific to the ThoughtSpot path; mixed-estate links for Excel, Looker and Domo sit above. The model rebuild leans on our Power BI data modeling and DAX service, the natural-language replacement is covered in depth by our Power BI Copilot consultant page, and once the new estate is live a Power BI health check keeps refresh, performance and security honest. Every task is scoped from your prepaid block of senior-led hours — no project minimums, and hours are deducted only for work you approve.
Book a free demo
Get senior Power BI help, on demand
Tell us the task. We scope it, you approve, and a senior consultant delivers — typically in 3–5 business days. Prepaid hours that never expire.
- Prepaid hours from $100/hour — and they never expire
- Senior-led delivery — 20+ years on the Microsoft platform
- Typical 3–5 business day turnaround
- Full refund if you’re unhappy in your first 4 hours
The fastest way to reach us is the form — tell us the task and we'll reply within one business day.