MedtechLive
Any assembly, costed in seconds.
Together with Meissner, we built a planning platform on its own production data, so planners see labour, material and sterilisation pressure weeks ahead.
Meissner
- 17 weeksfrom concept to go-live, to the agreed timeline
- Any assemblycosted in seconds, including assemblies the site has never made
- 2 sitesCastlebar and Camarillo, where planners, supervisors and finance use it every day
The situation
Meissner makes the single-use assemblies and filtration systems that biopharma companies use to manufacture biologics, vaccines and advanced therapies. At its Castlebar site in the west of Ireland, production planning decides how labour, components and sterilisation capacity line up against a constantly shifting order book.
The weekly capacity plan ran on manual worksheets that took up to four hours to prepare. It covered final assembly only, not the departments making sub-assemblies, and it rested on originally estimated labour hours with no feedback from what production actually achieved. A filter left on or a figure from the wrong column could surface five or six steps later.
The knowledge to run it sat with a small group of people. A 20-page procedure could not capture how the work was really done, which made it hard to train anyone else or cover absence. And with little forward view of capacity, a labour gap spotted too late could delay critical components.
What we built
We started on the factory floor, sitting with the Castlebar planners to see how the work was really done before writing any code. Working software went in front of users at every stage, with operations, IT and finance involved throughout.
The platform has five capabilities, all driven by one prediction engine.
Forward Planner. Forecast labour against available labour, by department, for the weeks ahead, with past-due work carried forward.
Past Performance. Actual labour compared with predicted, by department, so the variances worth investigating stand out.
Material Availability. Which jobs are ready to start, which are waiting on a sub-assembly and which are missing components, weeks before the day of build.
Sterilisation Forecast. Pallet demand at the external steriliser, projected against contracted capacity.
Predictor. Labour and material cost for any assembly at any quantity, including assemblies the site has never made, with the breakdown behind every figure.
The engine learns from years of Meissner’s production records, borrows from similar parts when a part is new, strips out statistical outliers and fits a curve to the economics of production. One automated weekly upload refreshes it.
All data stays in Meissner’s own Azure environment. People sign in through Entra ID single sign-on, with role-based, site-scoped permissions. The platform was built from day one for two sites, Castlebar and Camarillo, and each site’s working calendar is configurable.
The decision that mattered
We tested large language models against a purpose-built statistical model on the same historical production records. The purpose-built model was more accurate, so it’s the one that runs. It drives every chart and every on-demand estimate, so the numbers agree everywhere and can be explained to leadership.
Where it stands
The platform went from concept to go-live in 17 weeks, to the agreed timeline, with every milestone met and no change orders. It has been live since May 2026.
One weekly upload replaces two to four hours of manual preparation a week, an estimated 100 to 200 hours a year for the planner role.
For the first time, planners see labour, material readiness and sterilisation pressure together, weeks ahead and by department. Material shortages are chased early, emerging labour gaps are caught while there’s still time to act, and the Predictor costs any order in seconds.
Planners, supervisors and finance teams in Castlebar and Camarillo use it day to day. Meissner’s own teams own and run it, and Meissner has built a global planning and efficiency programme around it.