New Follow a KPI back to the file it actually comes from
Everscope reads the metadata of your Power BI and Fabric tenant and answers four questions you cannot answer today: what do I have, what is wrong with it, what changed since last week, and what can I show an auditor.
Metadata only. Everscope never reads a data value and never writes to your tenant. No Premium capacity, no XMLA endpoint, no sales call.
The problem
Mid-size BI teams do not lose control through one bad decision. They lose it through four years of reasonable ones. Nobody set out to build two hundred measures nobody references.
Question one
Somebody built a report for a meeting that stopped happening two years ago. It still refreshes every night and still turns up when people search. Nobody dares delete it because nobody knows who asked for it.
Question two
Two dashboards disagree about last month’s margin. Both are right, given their own definition. The person who wrote one of them left in March, and the workspace has no owner recorded.
Question three
A measure moved two points overnight. Power BI will tell you it changed. It will not tell you who changed it, what it looked like before, or whether anyone approved it.
The product
Not a dashboard about your dashboards. Each screen answers one question and ends in a decision you can point at later.
Semantic models, reports, lakehouses, pipelines and notebooks in a single filterable list, with owner, last activity and open issues. Beside it, which of the five analysis layers the scan was permitted to read, because a layer nobody could read is a gap and not a clean result.
How it works
Everscope has no model that guesses what might be wrong. Every finding comes from a named rule with a version number and the exact metadata it evaluated. Run it twice on the same estate and you get the same answer twice.
object is unreachable from the fully scanned dependency graph AND report usage was collected
object, its expression, and every reference searched
If the scan could not read report definitions, the rules that depend on them are suppressed rather than reported as clean. An empty result and an unread layer mean opposite things, and the product will not let them look alike.
A published rule is never edited. A changed condition becomes a new version, so a finding from last quarter still says exactly what it was tested against.
No language model reads your DAX and forms an opinion. That is what makes a finding something you can take to a change board.
Refresh load
Unread reports still refresh. Full reloads still set the nightly peak, and the peak is what your capacity is sized on. Everscope shows where the time goes, per model, from the refresh history it already collects.
Total refresh time per semantic model, longest first, with how many runs and how many of them failed.
Which hour carries the most refresh time. The peak is the number that sizes a capacity, not the average.
Capacity units and euro amounts are not in the APIs Everscope reads: they live in the Capacity Metrics app. Rather than estimate them and call it a finding, this screen reports the thing it can prove.
Security
The question your IT security team asks first is what leaves the tenant. The answer is metadata, and nothing else.
Object names, structure, DAX and M text, usage counts and refresh timings. Not one row of your data is read, transferred or stored.
A Full Scan requests Report.ReadWrite.All because Microsoft requires it for a read-only call. Everscope has no code that writes to your tenant. You will see that scope on the consent screen, so here it is first.
All processing and storage in EU regions. A signed processing agreement is available before you enter a card.
Everscope works through the standard REST and Fabric APIs. Nothing in the product requires Premium or Fabric capacity.
A scan without tenant admin consent can list what exists but cannot read structure, expressions or report definitions. It will find very little, and Everscope says so on screen rather than letting an empty list look like a clean estate. The full scan reads five layers and needs one admin approval, once.
Pricing
No demo to sit through, no implementation fee, no minimum term.
To find out whether you have a problem at all.
File scans, no card needed
Scan a fileFor the BI team that owns the estate and gets asked about the bill.
Billed monthly, cancel any time
Start with GovernedFor a single team that wants the backlog visible and owned.
Billed monthly, cancel any time
Start with TeamNeed customer-run collection, longer history or invoicing on purchase order? Write to hello@everscope.app and you get an answer, not a discovery call.
Why this exists
Five years of building BI teaches you that the hard part is not the first dashboard. It is year three, when the estate has grown past anyone’s memory of it and every question about it starts with “I think”.
I once spent the better part of a week working out which reports were still being read, by hand, across a few hundred objects. It was tedious, it was not difficult, and a computer should have done it. That week is why this exists.
The rule-based engine is not a technical preference, it is the whole point. A finding is only worth anything if you can defend it in a room full of people who would rather not clean anything up. “The model thinks this is unused” loses that argument. “Rule ES-USAGE-001 version 3.0 found no reference in the dependency graph, here is the trace” wins it.
Questions
A Lite Scan uses three delegated read permissions and works on plain Pro licences. A Full Scan adds Tenant.Read.All and, honestly, one write scope: Report.ReadWrite.All. Microsoft requires it for the read-only Get Report Definition call, which is the only reason it is there. Everscope has no code path that writes to your tenant, and you will see that scope on the consent screen rather than discovering it later.
Less than you would like, and the product says so on screen. Lite reaches the catalog layer only, so rules that need model structure, DAX or Power Query cannot run. On a tenant where a Full Scan returns a hundred and twenty-five findings, Lite returns a handful. That gap is the permission, not your estate, and the coverage panel names which layers were missed rather than leaving you to assume everything is fine.
No. Everscope reads through the standard Power BI and Fabric REST APIs and uses no XMLA anywhere. Nothing in the product requires Premium or Fabric capacity.
No data values. Everscope reads and stores metadata: object names, table and column names, relationships, DAX and M expression text, usage counts and refresh timings. Expression text can contain a value if someone typed one into a measure, and it is treated as confidential accordingly. Everything is processed and stored in EU regions.
Because governance findings get argued with. A deterministic rule produces the same finding on the same estate every time, states its version, and shows the metadata it evaluated. That is defensible in a change board meeting. A probabilistic answer is not, however good it sounds. There is no model anywhere in the analysis path.
The admin portal tells you what exists right now. It does not tell you what changed since last week, it does not group problems into rules with evidence, and it has no concept of an owner, a decision or a deadline. Everscope keeps every scan, diffs them, and turns a finding into something with a name on it and a date next to it.
Scanning stops at the end of the paid period. You can export the full evidence trail before or after that, and the stored metadata is deleted within thirty days. No exit interview.
One scan, every layer it is allowed to read, every finding visible. No card, no call, no obligation to like the answer.