Bief, the full simulation.
A green electricity supplier now buying all its energy on the wholesale market, with consumption forecasts still run on spreadsheets. The case followed on this page, the forecast for a Tuesday in November, on the eve of a cold snap forecast across the Alps.
A green electricity supplier now buying all its energy on the wholesale market, with consumption forecasts still run on spreadsheets.
Every morning, a consumption forecast that the procurement team can defend.
- Bief supplies renewable electricity to households and businesses across France from its headquarters in Grenoble. The company employs ninety people, including a five-person procurement team that buys power from generators and on the wholesale market.
- Until the end of 2025, part of the electricity sold by Bief came from ARENH, the mechanism granting access to EDF nuclear generation at a regulated price. It ended on 31 December 2025, and suppliers must now source entirely on wholesale markets or from their own generation assets.
- Consumption forecasting is handled in a spreadsheet using historical data and weather forecasts. When it is inaccurate, Bief incurs imbalance charges for the difference between what it procured and what its customers consumed, without being able to identify the source of the error the following day.
- Each morning, an agent prepares the consumption forecast for the coming days, customer by customer for large commercial accounts and by profile for households. It cross-references consumption history, weather forecasts, public holiday calendars, and customers joining or leaving.
- It also accounts for each variance from the previous day, such as colder weather than forecast or a client factory shutdown. The procurement team then determines the purchases and sales for the day.
What remains with the procurement team
The head of procurement decides on every purchase and sale, within the risk limits set by management. The agent places no market orders.
- Variance between forecast and actual consumption
- Variances explained the following day
- Time spent preparing the forecast
- Large accounts forecast individually
Billing queries
Complaints arrive by email and via the website form. The agent gathers the contract, meter readings, previous invoices and interaction history, then prepares the response.
The adviser reviews each response and decides on each goodwill gesture.
Contracts with small-scale generators
Bief also buys electricity from small renewable energy producers. The agent reads each proposed contract and compares its clauses with those of contracts already signed.
The procurement manager negotiates and signs each contract.
Metering data for energy audits
Since 11 October 2026, companies consuming more than 10 TJ of energy per year must have carried out an energy audit, unless they have an energy management system. The agent prepares a clear consumption statement for each relevant client, site by site.
The account manager approves each statement before sending.
Unpaid balances
The agent identifies overdue invoices and prepares reminders, including the solutions set out in the terms of sale.
Customer service decides on each payment plan.
the forecast for a Tuesday in November, on the eve of a cold snap forecast across the Alps
The forecast for a Tuesday in November, before and after.
- The forecast
- Every morning, the analyst fills in the spreadsheet with historical data and the previous day's weather.
- The agent delivers the forecast before eight o'clock, client by client for major business accounts.
- First workstream
- The forecast cold snap
- The cold snap is spotted late, when reading weather bulletins.
- The agent incorporates the weather forecast for the coming days and flags the expected increase in residential demand.
- First workstream
- The halted factory
- The cheese dairy halts a production line without warning, and the discrepancy only appears at the end of the month.
- The agent spots the drop the very next day and suggests adjusting the client's forecast.
- First workstream
- The previous day's discrepancy
- No one has the time to investigate why the forecast was off.
- The agent explains and categorises each discrepancy, from weather conditions to a client shutdown.
- First workstream
- The complaint
- A client disputes their bill, and the adviser searches for meter readings across several tools.
- The agent gathers the readings and history, then prepares the response.
- Workstream 2
- A producer's contract
- A small producer sends their contract, which is reviewed in full by the head of procurement.
- The agent compares its clauses with those of contracts already signed and flags any discrepancies.
- Workstream 3
- Collect
- Every morning, it retrieves the previous day's consumption readings and weather forecasts.
- Update the portfolio
- It takes into account new or departed clients and changing contracts.
- Forecast
- It prepares the forecast for the coming days, client by client for major business accounts and by profile for residential customers.
- Explain discrepancies
- It compares the previous day's forecast against actual consumption and explains each discrepancy.
- Flag
- It flags unusual situations, such as colder weather than forecast or a major client shutdown.
- The analyst reviews each alert.
- Propose
- It suggests an adjusted forecast and the volume variance to cover.
- The head of procurement decides on purchases and sales.
- Archive
- It stores each version of the forecast, along with the data used.
- Report
- Each week, it prepares a review of forecast quality and recurring discrepancies.
- Meter readings
- Client consumption data, transmitted by distribution system operators.
- Read-only
- Billing software
- Current contracts and client account changes.
- Read-only
- Weather forecasts
- Temperatures and sunshine levels for the coming days.
- Read-only
- Procurement spreadsheet
- Purchases and sales already settled.
- Read-only
- Forecast spreadsheet
- Daily forecasts and their discrepancies.
- Write
- Managing Director
- Sets the agent's rules with us and selects the next workstream.
- Head of Procurement
- Decides on every purchase and sale, within the risk limits set by management.
- The analyst
- Checks the forecast and flagged discrepancies each morning.
- Customer service
- Handles complaints and monitors major business accounts.
- The FDE
- Leads the engagement and writes the agent rules alongside the team. They are accountable for every agent.
- Forecasting agent
- Prepares the consumption forecast every morning.
- Variance analysis agent
- Explains and categorises the previous day's discrepancies.
- Client service agent
- Prepares responses to complaints and statements for business accounts.
- Supervisory agent
- Compares forecasts against actuals each week and alerts the FDE to any recurring drift.
- The forecast is inaccurate on a very cold day.
- The analyst reviews the forecast every morning, and the head of procurement maintains a buffer they determine independently.
- A major client changes their consumption without warning.
- The agent flags any sudden change the following day, and the account manager calls the client.
- The agent places a market order.
- It has no access to market tools. Only the procurement team buys and sells.
- Client data is leaked.
- Consumption data remains within the company's systems, and all access is logged. It remains hosted in Europe.
- A residential customer asks how their data is used.
- Customer service provides an answer, and their data is used solely to forecast their consumption and bill them.
- We measure the baseline over one month, including the variance between forecast and actual consumption, unexplained discrepancies, time spent each morning, and major clients forecast individually. The analyst drafts the forecasting rules with us.
- The agent prepares the forecast alongside the analyst for two weeks, and we compare both against actual consumption.
- Each morning, the procurement team works from the agent's forecast, verified by the analyst.
- We compare against the baseline and, with the Managing Director, select the next workstream.
- Variance between forecast and actual consumption
- Average daily variance, compared with the baseline month.
- Discrepancies explained the next day
- Share of discrepancies whose cause is identified the following morning.
- Time spent preparing the forecast
- Time reported each morning by the analyst, compared with the baseline month.
- Major clients forecast individually
- Business clients whose consumption is forecast separately.