KIASA

Case study

Varda Energy reads two million meters without a mainframe

Varda Energy moved two million smart meters off a 1980s mainframe in groups of 50,000. Readings now reach customers in 15 minutes, and not one bill went out wrong.

  • January 20, 2026
  • 6 minute read
  • Energy and Utilities
A road curving past wind turbines.
Client
Varda Energy
Industry
Energy and Utilities
Region
Europe
Meters
2 million
Duration
26 months, 2023 to 2025
Team
12 at peak

The challenge

A mainframe built for one reading a month

Varda Energy supplies power to two million homes and businesses. Its meter data ran on a mainframe designed in the 1980s for one reading per meter per month. Smart meters send one every half hour: 96 million readings a day, arriving faster than the overnight batch could process them.

Customers saw their usage the next day at best. Late or missing readings caused billing exceptions that a team of 30 worked through by hand.

Only four people still knew the mainframe's code well, and two of them planned to retire within three years. The regulator's deadline for half-hourly settlement was 2026.

The challenge

The mainframe had served Varda well for forty years. But it was built to read each meter once a month and bill from that reading. When smart meters arrived, the volume of data grew more than a thousandfold. The overnight batch began to overrun into the morning, and on the worst days it did not finish at all.

Every late batch meant late readings, and every late reading risked an estimated bill. In 2022, Varda's billing team handled 41,000 exceptions by hand. Running the mainframe cost more each year, and the people who understood it were leaving.

The approach

The first step was not code. We spent three months with Varda's four mainframe specialists, writing down what the old system actually did. They found 300 rules for validating, estimating and correcting readings, many of them added after some past incident and documented nowhere else.

We then built a meter data platform in the cloud that treats readings as a stream rather than a batch. Each reading is validated, corrected or estimated by the same 300 rules within minutes of arriving, and passed on to billing, settlement and the customer app.

“The mainframe team were not in the way of the project. They were the project. Nobody else knew what the old code really did.”

Sofia Lindgren, Chief Information Officer, Varda Energy

Then came what Varda's team named the quiet migration. Meters moved to the new platform in 40 groups of 50,000. For six months, both systems processed every reading, and every bill was compared before it was sent. Any difference held the bill until a person had checked it. Customers were never told, because nothing about their bills was meant to change, and nothing did.

Railway tracks under a web of power cables.
Half-hourly readings from two million meters, validated and passed on within minutes.

What the new platform does

  • Takes in 96 million readings a day as they arrive, not in a nightly batch.
  • Applies the same 300 validation and estimation rules the mainframe used, now written down and tested.
  • Shows customers their usage in the app within 15 minutes.
  • Supplies half-hourly data for settlement, as the regulator now requires.

The outcome

The last group of meters moved in August 2025, and the mainframe was switched off in November. Billing exceptions have fallen by 60%, and the team that handled them by hand now works on improving the estimation rules instead. The platform costs 34% less to run than the mainframe did in its final year.

96 millionMeter readings processed every day, each within minutes of arriving.

Two of the four mainframe specialists now lead Varda's platform team. Varda runs the platform itself, ships changes every week, and has started using the same data to plan where the grid needs strengthening.

Results

Faster data, fewer exceptions, lower cost

2 million
Meters read every half hour, without a mainframe.
15 min
From a meter reading to the customer's app, instead of the next day.
60%
Fewer billing exceptions to resolve by hand.
34%
Lower running cost than the mainframe in its final year.

Varda Energy metering and billing data, 2022 to 2026.

A red suspension bridge rising out of fog.

“We moved two million meters and not one customer called to ask why. For a migration, silence is the best review you can get.”

Kenji ArakawaHead of Metering Systems, Varda Energy

How the work ran

The quiet migration

  1. Write down the rules

    Three months with the mainframe specialists, documenting 300 rules for validating and estimating readings.

  2. Build the stream

    A cloud platform that processes each reading within minutes, using the same rules, tested against old results.

  3. Move 50,000 at a time

    40 groups of meters, moved one after another from early 2024 to August 2025.

  4. Bill in parallel

    Six months of comparing every bill from both systems before it went out.

  5. Switch off and hand over

    The mainframe went dark in November 2025. Varda's own team runs the platform.

Still running something on a mainframe?

We move core workloads off mainframes in small, reversible steps, with the people who know the old system at the center of the work.