A lakehouse in production, and a next-generation BI tool, both built
Antvia Data is a bronze, silver and gold lakehouse with governance built into the platform rather than added on top, deployed in your own cloud and running for customers today. Antvia Intelligence is the BI half, and it reads the gold layer that Antvia Data produces.
Five kinds of source in, five surfaces out
Antvia Data takes five kinds of source and serves five kinds of consumer, with the governed middle doing the work.
- In
Five kinds of source
Databases, application logs, spreadsheets, event streams and APIs. Antvia supports 300 or more sources, and a new one is connected in under 30 minutes.
- Land
Kept exactly as it arrived
Nothing cleaned on the way in and nothing quietly dropped. When someone asks months later what the source system actually sent, the record is still there to read.
- Model
One set of agreed definitions
Joins are made once, centrally, and the definitions Finance and the business owner settled on live in one place instead of inside each dashboard that needs them.
- Govern
Checked on the way out
Column-level access control and PII masking apply to the query itself, so the same rules hold for a dashboard, a notebook, an API call and a model.
- Out
Five consumer surfaces
Dashboards, mobile, reports, AI and ML, and embedded. A new dashboard is live in under two hours. The AI surfaces train on the same gold layer the dashboards read, not on a copy of it.
Bronze, silver and gold, and what each layer is for
The three layers are the reason a figure in a board pack can be walked back to the record that produced it.
- Bronze
Raw, exactly as it arrived
The landing layer keeps the source payload unchanged. It is dull until an auditor or a sceptical CFO asks what the system of record actually sent, and then it is the whole answer.
- Silver
Cleaned, deduplicated, joinable
Duplicates resolved and keys aligned, so two systems that disagree are reconciled once here rather than differently inside every report that happens to touch them.
- Gold
Modelled and business-ready
Revenue by region, churn at thirty days, whatever your business argued about and settled. Every surface downstream reads this layer, which is why the surfaces agree with each other.
The figures Antvia publishes for the lakehouse
These are Antvia's own numbers for Antvia Data, published in full at antvia.io, alongside the comparisons against Snowflake, Databricks, Power BI and Microsoft Fabric.
What getting to first dashboards costs
The gap between these two is the reason the product exists, and it is a gap in months and headcount before it is one in money.
Where two to four weeks lands on a ten month build
One bar, drawn in days, carrying both of the published figures at once.
Governance is in the platform, not bolted to the side of it
These are the controls Antvia publishes, and they should be read as what the platform does.
Column-level access control
Access is decided at the column, not only at the table, so one surface can be correct and safe for two people with different entitlements.
PII masking enforced on every query
Enforced at query time rather than when a dataset is prepared, so it holds whether the query came from a dashboard, a notebook, an API or a model.
A governed upload workflow for business users
A spreadsheet can be brought in without becoming an ungoverned side channel. The file lands in the same pipeline, with the same checks, as everything else.
Lineage that can be audited end to end
Keeping bronze exactly as it arrived is what makes this possible. A figure in a board pack can be walked back to the record the source system sent.
You own it and you can leave with it
Antvia runs in your own VPC on AWS, Azure, GCP or on-prem, in open formats. The tables are readable by tools that are not ours, and you can export at any time.
One filter changes: what mainstream BI does next
Every tool answers the first question well, so the useful thing to trace is the second.
- Click
A filter changes
An analyst narrows to one region because the first chart looked odd. This is the cheapest thought in the process and it should be the cheapest action.
- Queue
The request leaves the browser
The tool cannot answer locally, so the question travels to the warehouse and joins whatever else is running. Your wait now depends on other people's jobs.
- Scan
The warehouse reads the data again
Often the same data it read for the same person a short while earlier. The meter runs on compute you have already paid for once.
- Wait
The analyst holds the thought
A few seconds is enough to break a train of thought. Long enough and the third question never gets asked, and that is usually the one that finds the problem.
- Bill
Curiosity turns up as spend
At quarter end someone asks why the warehouse bill grew. The honest answer is that people used the tool.
The same three questions, under each architecture
One analyst, one afternoon, three questions in a row, run once on mainstream BI and once on Antvia Intelligence.
- Q1 Revenue by region: scan the warehouse, waitQ1 Revenue by region: scan the warehouse, wait
- Q2 Narrow to the south: scan again, wait againQ2 Narrow to the south: re-cut in the browser
- Q3 Split by product: a third scan, a third chargeQ3 Split by product: re-cut again, no round trip
- The chat box sits beside the chart, so you retype the state it cannot seeIt sees your filter and the agreed definitions
What Antvia Intelligence does
Intelligence reads the gold layer that Antvia Data produces, which is why the model can hold the definitions the business agreed and show its working.
It does the interactive work in the browser
Filtering, slicing and re-cutting a result you are already holding happens on the client, so the second and third questions are not charged like the first.
It sends to the backend only what needs it
Fresh data, large scans and anything that must be governed centrally still run on the server. The backend is not removed, it is asked less.
It puts the model inside the analysis
The model sees the state you are in, the definitions your business agreed and the columns in front of you, then works on the analysis with you.
It reads one set of binding definitions
A person and a model asking about margin, churn or active customer read the same definition, from the gold layer, rather than two teams arguing faster.
It shows its working
Any number traces back to the query and the source behind it. An answer that cannot be explained does not survive its first meeting with Finance.
Antvia Intelligence on screen
Reserved for a real capture from the product, not a stand-in.
How the engagement runs
An Antvia engagement runs the way every Woodfrog engagement runs, starting with a paid audit you keep either way.
- Week 1
The audit, two calls and a one-pager
A fixed fee, two calls and a written one-pager: what your estate costs to run, which surfaces carry decisions, and whether Antvia belongs in your case at all. You keep the document whichever way you decide.
- Weeks 2 to 6
Build, in your own cloud account
A pod of three works inside your VPC on AWS, Azure, GCP or on-prem. Across more than fifty projects our median to a first production-grade artefact is six weeks.
- Week 7 onward
Operate, with the keys in your hands
We run it while your team takes it over. The tables sit in open formats readable by tools that are not ours, so leaving is an export rather than a migration project.
Where Antvia is the wrong answer
There are cases where the honest recommendation is that you do not need this, and it is cheaper for everyone if that comes out in week one.
Your current stack answers the questions people ask
If the warehouse you have serves the business at a cost and speed you are content with, replacing it is a project with no return attached. We would rather say so than sell you a migration.
One system holds everything and it agrees with itself
A lakehouse earns its place when data arrives from several systems and those systems disagree. If a single database answers your questions today, this is heavier than the question needs.
The problem is what got built, not what it runs on
Most of our analytics delivery has happened inside the platform the client already runs, including Apache Superset. If the platform is sound and the estate on top of it is the problem, a new platform is not the fix.
You need published proof for the BI half before you commit
Intelligence is built and in use, but there is no published benchmark, price or reference customer for it. If your decision depends on those, better you know now than in week four.
Questions buyers put to us about Antvia
The six that come up in almost every first call.
What can I actually buy today?
Antvia Data is in production with customers, deployed in their own cloud, and its figures are published at https://antvia.io. Antvia Intelligence is built and comes into the engagements we run. What does not exist yet for Intelligence is a published benchmark, a price list or a named reference customer.
How does this compare with Snowflake, Databricks or Microsoft Fabric?
Antvia makes that comparison itself, so read it as Antvia's argument rather than as a neutral one. Those platforms are capable and well marketed, and the difference Antvia points at is the cost model rather than a number: compute billed per query or per processing unit, seats charged separately, governance and real-time features sitting in premium tiers, and egress charged per GB. Antvia charges for the sources it connects and the metrics it governs, with no per-query charge, no per-user fee and no feature gating, and the infrastructure stays yours. The second difference is tenancy: on Antvia the platform runs in your own cloud account in open formats, so you can leave with it. The full comparisons are at antvia.io/compare.
What is in the benchmark report, and what does it cost?
80+ benchmarking metrics across every layer of the stack, measured on 480 GB across 3.2 billion rows and twelve source systems, head to head against AWS Athena, Snowflake, Google BigQuery and Databricks. Methodology, sample SQL and architecture diagrams are included, with an optional 30-minute walkthrough. Ask us here, or write to hello@antvia.io and it comes straight back.
Is Antvia certified for SOC 2, ISO 27001, HIPAA or GDPR?
We make no certification claim and you should not read one into this page. What Antvia publishes are capabilities: column-level access control, PII masking enforced on every query, a governed upload workflow, and auditable lineage from bronze through to gold. If a certification is a gate in your procurement, ask us before you shortlist.
If the interactive work moves into the browser, what happens to governance?
Definitions, access control and anything requiring central enforcement stay on the server, and PII masking stays on the query. What moves to the client is the re-cutting of a result the user is already permitted to see.
We have just paid for a BI licence. Why take on another tool?
You probably should not. If your current tool answers the questions your team asks, at a speed and price you are content with, keep it and let us fix what got built on top of it. Antvia Intelligence is for the case where exploration itself has become the expensive part.
Start with the audit and find out which half of Antvia you need
Two calls, a fixed fee, and a one-pager you keep either way: what your current estate costs to run, which surfaces actually carry decisions, and which half of Antvia belongs in your case. If your existing tool is fine, we will say so.