Transforming FP&A with AI Modernization

Transforming FP&A with AI Modernization

Skip to content

Blog / Product and Technology / How We Use CoCo and Streamlit to Make Long-Range Planning Conversational

JUL 08, 2026 / 11 min read Product and Technology How We Use CoCo and Streamlit to Make Long-Range Planning Conversational

Brad Floering

Long-range planning is one of the most important exercises a finance team runs, but it is also one of the hardest to scale.

At Snowflake, we needed to forecast out 10 years across a business that had become much more complex: more than 40 entities, each with more than 100 cost centers and hundreds of spend categories underneath that. We needed that level of detail because the long-range plan was not just a finance artifact. It had become a data set that other teams depended on.

Tax, for example, does not just need a high-level expense forecast. They may need to understand the split between goods and services, legal entities, jurisdictions, and other operating details that matter for planning. Treasury needs a view into cash. Workforce planning needs headcount assumptions. Executives need to understand the strategic tradeoffs across growth, margin, investment and free cash flow.

The planning model had to support all of that.

So, like many companies, we did what finance teams often do when the business outgrows the tooling: We built a gigantic Excel model.

And over time, it became what most large finance models eventually become. It worked, but it looked like Frankenstein’s monster. New tabs were added. New formulas were bolted on. New logic was layered onto old logic to accommodate a changing business. The model became incredibly valuable, but also increasingly difficult to maintain, govern and scale.

That was the first problem we set out to solve.

From spreadsheet model to planning platform

More than a year ago, we rebuilt the long-range planning model in Snowflake and used Streamlit as the UI layer for analysts and executives to interact with the forecast.

That became Snowplan, our internal long-range planning application.

The goal was not to create a dashboard. It was to create a planning platform. We wanted an experience that still felt intuitive to finance users, but with the scale, governance and compute power of Snowflake underneath it.

In Snowplan, analysts can update assumptions through an editable Streamlit interface. Those changes are written directly back to Snowflake, where the model runs, and the updated outputs are immediately resurfaced in the app. No broken formulas. No file saves. No wondering which version of the model is the source of truth.

That architecture changed the planning process.

Instead of maintaining a massive offline workbook, we now had a governed application connected to the raw data sources that already power the business. Actuals could flow into the model without someone spending hours updating files. Assumptions could be versioned. Scenarios could be compared. Different users could interact with the same planning platform at the right level of detail based on their role.

For individual contributors and associates, Snowplan provides granular input pages, assumption management, scenario creation and version control. For directors and managers, it provides visibility into logic and assumption changes for review and approval. For executives, it provides a consolidated view of the P&L, free cash flow and key scenarios.

That matters because long-range planning is not just a modeling exercise. It is an alignment exercise. The more time finance spends maintaining the model, the less time we spend pressure-testing strategy with the business.

Data used in the image above is fully synthetic data.

Why building in Snowflake changed the model

The most important decision we made was to build the model where the data already lived.

Because Snowplan runs on Snowflake, it is connected to our raw data sources and governed data model. That means we are not wasting time manually updating the model with actuals or reconciling offline data pulls. The model is tied into the same environment where finance data, permissions, logic and history already exist.

That creates several advantages.

First, the model can scale. A 10-year forecast across entities, cost centers, spend categories, headcount, revenue, balance sheet and free cash flow creates a significant amount of data. That is exactly the type of workload Snowflake is built to handle.

Second, the model becomes easier to govern. Access can be managed through Snowflake role-based permissions and row-level controls, so users see the right data and functionality based on their role. Executives do not need the same interface as the analyst building the assumptions, and analysts do not need to create separate exports for every stakeholder.

Third, the model becomes reusable. Once the planning logic is built in Snowflake, it does not have to be limited to the annual long-range plan. The same platform can support workforce planning, stock-based compensation modeling, treasury cash forecasting, hedging, legal entity forecasting, COGS planning and M&A scenarios.

That is the bigger story. Snowplan is not a one-off planning app. It is becoming a platform for finance planning.

Snowflake CoCo made scenario planning conversational

Streamlit made Snowplan scalable and usable. Snowflake CoCo made it conversational.

Before CoCo, Snowplan already gave us a better way to manage long-range planning. Analysts could update assumptions in the app, run scenarios and compare outputs. But the user still had to know where to go, which assumption to adjust and how to interpret the downstream impact.

CoCo changes that interaction model.

Now, instead of navigating through pages of assumptions, I can ask a question in plain English. I can ask CoCo to compare two versions of the forecast and summarize the major drivers. I can ask what changed between the plan we showed the board last year and the latest version we are preparing now. I can ask for the net result of the changes, the key drivers of margin expansion or dilution, and the assumptions that matter most.

That is incredibly powerful in executive planning.

When preparing for a board discussion, the question is rarely “can you give me the latest forecast?” The question is usually, “what changed, why did it change and what does that imply for the story we are telling?” CoCo helps compress that analysis from a manual comparison exercise into a conversation.

The...