Lineage you never had to write
Parsed from query logs on day one. Test it by renaming a column in staging and watching the blast-radius list appear before your pull request merges.
Why Northlane
The graph is table stakes. The difference shows up at seven on a Monday, when a column was renamed on Friday and sixty dashboards are quietly wrong. What happens in the next twenty minutes is the actual product.
We do not run a proof of concept as a sales stage. You create a read-only warehouse role, we parse thirty days of query history, and you are looking at your own lineage graph inside a day. If the graph comes back wrong, that is the evaluation result — and we would rather you find that out in week one than in month six.
Pricing is metered on rows processed and models managed, so seats are free and inviting an analyst never costs anything. Every incident review is published in full within five working days, including the ones that were our fault. Everything you build exports as plain SQL and YAML whenever you ask for it.
Measured, not claimed
Three numbers taken from customer workspaces and republished every quarter. The method behind each one is in the note under it.
11 days Decrease, an improvement.
Median time to a governed pipeline
First warehouse connection to a pipeline with lineage, tests and a named owner, across 62 deployments in 2025.
98.9 % Increase, an improvement.
Freshness SLA met last quarter
Across every pipeline on a Business or Enterprise plan. Misses are credited without anyone having to ask.
5 days Decrease, an improvement.
Median time to publish an incident review
Timeline, root cause and fix, on the public blog. Six years of them are in the archive, unedited.
Reported from customer workspaces, Q4 2025. Counting method and raw totals are published in the trust portal.
How we build
The lineage graph is parsed from SQL that actually executed against your warehouse. Nothing is declared in a YAML file, which means nothing can quietly fall out of step with what is running. The same graph feeds the monitors, the access policies and the cost report, so a permission change and a spend report can never disagree about which model is which.
The price of that decision is that we are close to useless on a warehouse with no query history — a greenfield migration gets little from us in the first fortnight. We would rather say that here than sell a two-week onboarding whose real content is writing metadata by hand.
How we sell
List pricing is on the pricing page and it is the same whether you arrive through a partner, a search result or an introduction. Seats are free, because charging for them would make you think twice about giving an analyst warehouse visibility, and rationed access is the exact problem this product exists to remove.
There is no renewal uplift written into the standard order form, no egress fee on an export, and no clause making an export conditional on being in good standing. The export endpoint sits in the same API reference as everything else, and it works on day two as well as it works on the way out.
In every order form
The short list
Written as claims rather than adjectives. Each one comes with a way to prove it wrong inside the first week of a trial.
Parsed from query logs on day one. Test it by renaming a column in staging and watching the blast-radius list appear before your pull request merges.
Every monitor resolves to the team that owns the model, with the failing SQL attached. Test it by breaking a load deliberately and reading the page you get.
Metering is on rows processed and models managed. Test it by adding twenty analysts to a trial workspace and watching the projected invoice not move.
Every incident, including the ones that were our fault, on the blog within five working days. Test it by reading the archive before you sign anything.
Open the incident archiveModels, tests and policies come out as plain SQL and YAML through the documented API. Test it on day two, not on the way out.
Roadmap items are labelled roadmap in the proposal, never capability. Test it by asking for anything on this page to be written into the contract.
Run all six inside a 14-day trial. We will help you try to break them.
Independently checked
Claims about our own product are ours to make. These are not. Each one is examined by a third party on a published cycle, and every report is available to read before you sign rather than after.
Certificates and reports are released under NDA through the trust portal, normally within one business day of the request.
SOC 2 Type II
Examined annually by an independent CPA firm across Security, Availability and Confidentiality. The current report covers the twelve months to September 2025.
CurrentISO/IEC 27001:2022
Certified information security management system covering the platform, corporate IT and both data-centre regions.
CurrentISO/IEC 27701
Privacy information management extension, audited on the same cycle as the 27001 certificate.
CurrentGDPR
EU representative appointed. Standard contractual clauses and the UK addendum ship inside the DPA, with no negotiation required.
In scopeHIPAA
Business associate agreement available on Enterprise. Protected health information is restricted to residency-pinned workspaces.
EnterprisePenetration testing
An annual third-party test plus continuous automated scanning. Critical findings are patched within 72 hours and the summary letter ships with the report set.
AnnualThe certificates and the current SOC 2 report are the authoritative documents. Nothing written on this page amends a signed agreement, and where the two disagree the agreement wins.
In their words
Each of these customers has agreed to take a reference call. Ask for the introduction during your evaluation and we will arrange it.
We ran Northlane and one other vendor against the same warehouse for three weeks. The graphs were comparable. What separated them was that one of the two told us which forty of our models nobody had queried since 2023.
34% lower warehouse spend in one quarter
Priya Raghavan
VP Data Platform · Halden Systems
The thing I did not expect was the incident review. They published their own outage in full five days later, with the commit that caused it. That told me more about who we were buying from than the demo did.
1,400 models migrated in nine days
Marcus Feld
Director of Analytics Engineering · Ardent Labs
Seat-free pricing reads like a marketing line until you stop rationing access. We went from fourteen people with warehouse visibility to ninety, and it never became a procurement conversation.
90 people with warehouse visibility, up from 14
Dana Okonkwo
Head of Revenue Operations · Corvane
Evidence
Three further deployments — none of them the teams quoted above — each one testing a different claim from the short list. Numbers signed off by the customer, no pilots, no anonymised logos.
Logistics 1,900 employees
Tests claim one, lineage nobody had to write. A five-year-old scheduler with no owner came out in nine weeks, on a graph parsed from query history rather than authored by hand.
68% fewer pipeline incidents
Financial services 5,200 employees
Tests claim five, an export that is not a favour. Old and new pipelines ran side by side until the column diff came back empty, and the whole regulatory review was answered from one export.
11-day regulated cutover
Freight and logistics 860 employees
Tests claim two, alerts that name an owner. A fifteen-hour handoff chain collapsed to one named engineer with the failing SQL already attached, across eleven reverse-ETL syncs.
19 hours to 4 hours MTTR
Start the conversation
No cost, no obligation — a real conversation about your organization, your goals, and your budget cycle. We'll help you figure out where you'd actually land, and tell you honestly if now isn't the right time.
No commitment to talk. No pressure to sign afterward.