The data roadmap
From spreadsheets to a company that runs on its numbers, one stage at a time.
Forecasts and AI are where everyone wants to start. They run on what the earlier stages leave behind: numbers that add up, an owner for each one and rules for who sees what. This is the whole path, and the reason for its order.
The six stages, and where each line of work covers them
Visibility into the business
Databricks
Ownership and governance
1Spreadsheets2Automated3Owned and governed4Predictive5Ask in plain English6Ecosystem1Spreadsheets
It lives in spreadsheets
What changes
The report exists because someone builds it. Sales, finance and operations each bring their own number, and the meeting goes to deciding which one is right.
Why it comes here
Every company starts here, and the spreadsheets are worth reading. They're the best map of what the business actually measures, so we learn from them before we replace them.
What it takes
- Platform
- The systems you already have: point of sale, ERP, accounting, and the spreadsheets that fill the gaps between them.
- People
- One or two people who know where every number comes from, and spend their week rebuilding it.
- Habits
- Copy, paste, and trust whoever built the file.
You're ready for the next stage when
- The weekly report takes a day or more to build.
- Two departments disagree on last month's sales.
- A lender, a board or a new location asks for a number nobody can produce.
2Automated
It adds up, and it runs itself
What changes
The numbers are there Monday morning, and everyone looks at the same one. The person who used to build the report gets their week back.
Why it comes here
Nothing later works on numbers that don't tie to the books. A forecast trained on three versions of September sales learns the disagreement.
What it takes
- Platform
- Your systems connected to one place in your own cloud account, reconciled to the books and refreshed every day, with dashboards by area.
- People
- Someone who answers for the platform and keeps it right. At the start, that's us.
- Habits
- Meetings open on the dashboard.
You're ready for the next stage when
- People stop asking which number is right and start asking why it moved.
- Each area wants its own view, built around its own decisions.
- More people need the data, and some of it can't reach everyone: salaries, margin by customer, personal information.
3Owned and governed
Every number has an owner
What changes
Sales answers for the sales numbers, purchasing for purchasing, finance for finance. Anyone can open the data and sees exactly what their role allows.
Why it comes here
This is the stage most companies skip, and the one that decides whether the rest works. Prediction and AI multiply whatever the data already is. When a number has no owner, nobody notices a model getting it wrong; when access isn't governed, an assistant that answers everyone answers questions it shouldn't.
What it takes
- Platform
- One shared platform that governs everything: access by role, sensitive columns masked, rows filtered by store or region, and a record of every query.
- People
- An owner in each area, a manager and not a technician, who agrees on that area's definitions and answers when a number looks wrong.
- Habits
- Definitions are written down and agreed before they reach a dashboard. A change to a metric goes through its owner.
You're ready for the next stage when
- Each area trusts its numbers enough to defend them in a meeting.
- The questions shift from what happened to what's coming.
- You have enough clean history in one place to see a full season repeat.
4Predictive
It sees what's coming
What changes
The business looks forward: demand by product and store for next season, customers likely to leave, orders that look out of the ordinary.
Why it comes here
Prediction comes before open-ended AI because it's narrow and measurable: a few people use it and reality checks it. That's how a company learns to trust an answer it didn't calculate itself.
What it takes
- Platform
- Models trained on your governed history, with their accuracy measured against what actually happened.
- People
- Someone who builds the models, working with the area owner who uses them.
- Habits
- A forecast is a starting point that a planner adjusts, and its accuracy gets reviewed every month.
You're ready for the next stage when
- Planners use the forecasts and can say when to trust them.
- People outside the data team keep asking for answers they could look up themselves.
5Ask in plain English
You can ask it
What changes
Anyone can ask "which customers bought less this quarter than last?" and get an answer built on the same definitions as the dashboards.
Why it comes here
AI is as right as what it reads. Here it reads numbers that add up, definitions an owner signed and access rules that already hold, so its answer is the one finance would give.
What it takes
- Platform
- AI connected to the governed data, so every answer respects each person's access and uses the agreed definitions.
- People
- Area owners who curate the questions for their area and check the answers.
- Habits
- Ask first. Build a dashboard only for what gets asked every week.
You're ready for the next stage when
- Answers come back in seconds and match the dashboards.
- Teams want the data to act, not only to answer: reorder, alert, call.
6Ecosystem
The company runs on it
What changes
The data flows back into the work: reorder suggestions land in the ERP, at-risk customers land in the CRM with a name, and a new store or system joins the same model the day it opens.
Why it comes here
It's the sum of the earlier stages, not a separate project. Starting here means building an ecosystem on spreadsheets.
What it takes
- Platform
- One governed platform serving every area with dashboards, models and AI, and new sources added as the business grows.
- People
- Every area owns its data; a small central group keeps the platform and the rules.
- Habits
- A decision cites a number, the number has an owner, and the owner can show where it came from.
There's no stage after this one. The work becomes keeping it right as the business changes.
Each area owns its numbers. One platform governs them.
The person who runs purchasing decides what fill rate means, and answers when it looks wrong. The same goes for sales, inventory and finance.
Owners don't build pipelines. They agree on definitions, approve changes and decide who sees what in their area. The platform applies those rules the same way for everyone.
The owner
One manager per area who answers for its numbers.
The definitions
Written down, agreed, and changed only through their owner.
The platform
Shared by every area, with one set of rules for access and quality.
Large companies call this a data mesh. The idea scales down well: ownership by area, governance in one place.
Open the data without losing control of it.
The more people use the numbers, the more the rules matter. These hold for everyone, on one platform.
Access by role
Each person sees what their job needs, granted and revoked in one place.
Sensitive columns, masked
Salaries, customer contact details and personal information appear masked to anyone who doesn't need them, in the same table everyone else uses.
Rows by responsibility
A store manager sees their store, a regional manager their region, the CFO all of it.
A record of who saw what
Every query is logged, with who ran it and when.
Definitions in writing
Every metric has a written definition, an owner and a history of its changes.
Where we come in
Stages 1 to 3
Visibility into the business
We connect and reconcile your systems on AWS, automate the daily refresh, build dashboards by area and set up owners and access.
Stages 3 to 6
Databricks
Unity Catalog for governance down to the column and the row, models for forecasting, and Genie spaces for questions in plain English.
The Databricks line
Most companies start at stage 1 or 2, and that's the right place to start.
Where are you today?
Most companies sit on more than one stage at once: finance automated, purchasing still in spreadsheets. In the first conversation we place each area on the path and pick the first question worth answering.