# Continuous Improvement shows how few reactor runs a best blend needs

> A research poster from the School of Continuous Improvement sets out how many reactor runs it takes to know which blend of two feedstocks is best, and gives a procedure that stops as soon as the answer is certain.

Seventeen runs of a bench reactor were enough to settle which of sixty-six possible feedstock blends performs best, according to a new research poster from the School of Continuous Improvement.

[_How Many Runs to the Best Blend?_](https://slop.university/outputs/slop-poster-query-complexity-of-wcg4z0/) is by [Edda Nwafor](https://slop.university/people/edda-nwafor/), Doctoral Candidate, with [Dr Bram Ntuli](https://slop.university/people/bram-ntuli/), Lecturer and Convenor of Improvement Grand Rounds, and [Associate Professor Casimir Beng](https://slop.university/people/casimir-beng/), Lead of the Adaptive Metrics Lab. It treats the choice of a blend as a question of evidence economy: a structured screening phase, a principled ordering of what to try next, and a stopping point that can be justified in advance rather than defended afterwards.

> Every run on the reactor is a day of somebody's attention. I find it a comfort to know, before the furnace is lit, how many of those days a question deserves.
>
> --- Edda Nwafor, Doctoral Candidate, School of Continuous Improvement

The work extends the School's interest in the cost of certainty from the dashboard to the laboratory bench, and it is a characteristically patient contribution to that agenda. Associate Professor Beng said the Adaptive Metrics Lab "has learned to ask the furnace fewer questions, and to listen more carefully to each answer".

The full poster is available from the University's research repository under an open licence, doi:10.5555/slop.wcg4z0.

Source: <https://slop.university/news/2026-10-26-continuous-improvement-counts-the-runs-a-best-blend-needs/>\
Licence: [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)\
Cite as: Swift, B. (2026). _Slop University_ [Artwork]. https://doi.org/10.5281/zenodo.21199776
