Value AG is a leading company for real estate valuation in Germany. Gemma Analytics built demand forecasting with machine learning on a Snowflake data platform for Value AG, so operational teams plan staffing against expected demand instead of static rules.
The starting point
As an independent partner to banks, insurers and investors, Value AG values properties of all asset classes. Its work covers valuation reports, on-site inspections and market data, with teams on site all over Germany. How many people are needed, and when, depends on how many valuation orders come in, and that demand changes over the course of the year.
Staffing was planned with manual estimates and static rules. Seasonality and current trends slipped through that kind of planning. The result was over-staffed quiet weeks and bottlenecks at the peaks, a mismatch that costs money in one direction and service quality in the other.
What we built
Gemma Analytics has worked with Value AG since 2023. It began as data analytics consulting: interviews across the business, an infrastructure recommendation and a roadmap, followed by building a central data platform. Data from the internal systems for orders, appointments and inspections, spreadsheets on SharePoint and a tour planning interface come together in one Snowflake warehouse and are modelled with dbt. On that basis Value AG reports on order volumes, utilisation and service levels.
The machine learning forecast models run on this data. They take trends, seasonality, product mix and external factors into account and are recalculated automatically. The forecasts feed directly into operational planning, so team sizes follow the demand that is actually coming instead of a fixed rule.
The result
The forecast reaches 90% accuracy, and over- and understaffing fell by up to 50% across all regions. Because the forecast is built into planning, Value AG works with lower costs, better service levels and teams that fit the workload, in quiet weeks as well as at the peaks.
Tools and setup
Snowflake is the central data warehouse. Apache Airflow orchestrates the data pipelines, which run in Docker containers, and dbt handles the data modelling. The code is versioned in Azure DevOps, the forecast models run as scheduled tasks inside Snowflake, and Power BI dashboards show forecast, plan and actuals to the business teams.