Your data program on Snowflake, sized, built and proven

A program has a life. There is a block for each stage

An idea, a number, a plan, a build, a run, and then growth that has to stay governed. Most teams are somewhere in the middle. Start at the stage you are in.

1. 1 · Know where you stand

   Assess

   Nobody can tell me what to fix first.

   Findings on architecture, security, workloads and cost, in order of priority.

   Private preview

2. 2 · Decide what it takes

   Estimate

   The committee wants a number by the end of the month. We have three quotes that don't compare.

   Scope, estimate, scenarios and a proposal, from your own documents.

   Early access

3. 3 · Agree on how and when

   Plan

   The budget is approved. Nobody agrees on what ships first, or who signs it off.

   The estimate you approved, turned into epics, sprints and tasks with hours and due dates, with burndown, velocity and a completion forecast to follow them.

   Early access

4. 4 · Build it and prove it

   Deliver

   The backlog is six months long, and nobody can prove that what we shipped is right.

   Governed agents, and accelerators for migration, dbt conversion, DataOps, account foundation, data quality and FinOps.

   Early access

5. 5 · Run it in the open

   Observe

   The dashboards were slow again on Monday. We found the query on Wednesday.

   What runs, waits and changes on each warehouse, down to the exact query.

   Early access

6. 6 · Grow under control

   Optimize

   Finance wants the bill split by team. We have one number for the whole account.

   Spend, owners and savings you can check.

   In development

Every stage has a tool. Nobody owns the hand-offs

• The finance warehouse still runs on Teradata. The committee meets in three weeks and wants a number. Three quotes, and none of them compare.

  A number you can defend

  An RFP, a migration inventory or a list of findings becomes lines of work, each with its effort, scenarios you can compare, and a proposal you can put in front of a committee.

  On a real bid: 3 hours of work apart from a two-week expert assessment.

• We signed Snowflake in June. Everyone has an opinion on what ships first. Nobody has a plan, and the foundation has already been built twice.

  An agreed plan, with dates

  Plan turns the estimate you approved into epics, sprints and stories, each with its hours and its due date. The Foundation accelerator in Deliver takes a blank account to a governed one, once.

  A plan generated from the number you approved, not from a blank page.

• The bill is up 38% since we moved the dbt jobs. Nobody can say where it went. And the AI use case needs a budget line next month.

  An owner for every euro, and a number for the next thing

  Observe shows what runs on each warehouse, down to the query. Optimize gives every team its share of the bill. The next use case goes through Estimate like any other program.

  Observe refreshes every 5 minutes by default. Optimize is in development.

Where this comes from

Every block was built on CRITHINK's own engagements before it was offered to anyone else.

enterprise migrations

100+

countries

7

engines in the migration accelerator

12

of work between our estimate and a two-week expert assessment, on one bid

3 h

Questions we get first

• Where does our data go?

  Observe runs inside your Snowflake account. In Deliver, the agents run on your engineer's machine, where their conversations stay, and the migration accelerator runs on your machines, on your network. Estimate and Plan are a web application that works from the documents you give it. The security page says what each block reads.

• Can the assistant touch production?

  No. It is read-only by default. In a delivery it writes in the dev target only, it never pushes to the integration or production branch, and nothing is published before a person approves the list of changed files.

• Are you certified?

  We do not claim SOC 2 or ISO 27001 today, and we will not show a badge before an audit is valid. The security page lists what each block reads and where it runs.

• Do we have to buy the whole platform?

  No. Each block works alone. A second one makes sense when one block's output is the next one's input: findings that need a price, a plan that needs delivering, a change that needs checking.

• Do the blocks talk to each other?

  Each block works alone today, and you carry the output of one into the next: findings go into Estimate as documents. Direct links between blocks are what we are building now.

• What if we need people, not software?

  The same engineers run migrations and platform work at CRITHINK consulting. You never have to buy it to use the software.