kaizen in practice
D-KAI-11 | First published 2026-08-12 | Revised 2026-08-12 | Primary sources: 1 peer-reviewed study
Disclosure: Gemba Works Lab is supported by advertising. This page contains no affiliate links; if any are added in the future they will be clearly marked (per the FTC Endorsement Guides).
"We did a great kaizen event and six months later everything drifted back" is one of the most common complaints about continuous improvement — common enough that a research team from the University of Helsinki and Aalto University decided to actually measure it, rather than take it on faith. Their study, published in BMJ Open in 2023, followed 26 kaizen events across a large Finnish hospital system and tracked which ones held their gains and which ones quietly reverted. The setting is healthcare, not a factory floor, but the mechanics of why an improvement sticks or doesn't turn out to have very little to do with the industry — and everything to do with follow-up, workload, and culture.
KEY POINTS
Haapatalo, Reponen, and Torkki studied the Hospital District of Helsinki and Uusimaa (HUS) — 2.2 million people served, 27,000 staff — which had run 26 kaizen events across specialised care units between 2017 and 2020, each a five-day workshop following the standard preparation → event → follow-up structure (source: Haapatalo E, Reponen E, Torkki P. "Sustainability of performance improvements after 26 Kaizen events in a large academic hospital system," BMJ Open 2023;13:e071743). The researchers went back to the unit leaders — 21 of the 26 responded — with structured interviews plus whatever performance-metric reports the units still had on file, and rated each unit's long-term outcome on a four-level scale adapted from an earlier sustainability framework (Bateman & David):
| Level | What it means |
|---|---|
| A | Results sustained, objective met or exceeded; improvements spread to other units |
| B | Results and procedures sustained; objective met, but no further improvement |
| C | Qualitative improvements made, but not confirmed or sustained by the metrics |
| D | Struggled to make any improvement, and what little changed did not last |
Source: Haapatalo et al., BMJ Open 2023, sustainability levels adapted from Bateman & David.
Of the 21 interviewed units, 62% landed at sustainability level A or B — meaning they kept the improvement and, in most A-level cases, went on to spread the method to other units or apply it to new problems. Level C accounted for 14% (qualitative improvement noticed, but nothing the metrics could confirm), and level D accounted for 24% — five units where the kaizen event essentially left no lasting mark. Counting the five units that took part in an event but never responded to the follow-up interview as presumptive level D, the study's authors note that sustainability levels C and D together would then account for roughly half of all 26 events (source: Haapatalo et al., BMJ Open 2023).
Framed the other way: of the 21 units that could be assessed, 12 (57%) both reached their objective and held onto it long-term. That is a genuinely useful number for setting expectations before you run a kaizen event of your own — a well-run event roughly doubles your odds of a lasting result compared to a coin flip, but it does not make sustained improvement close to automatic.
The study's most useful contribution isn't the pass/fail rate — it's the "why." Researchers coded 15 explanatory factors from the interview transcripts and tied each quote to whether it explained persistence or decline. Three findings stand out for a small manufacturer reading this from outside healthcare:
Put together, the study's own framing is direct: level-A units mentioned reasons for persistence roughly nine times more often than level C/D units did, while level-D units mentioned reasons for decline more than twice as often as everyone else — mostly citing turnover, workload, and unclear objectives (source: Haapatalo et al., BMJ Open 2023).
None of the 26 events studied were in manufacturing, and the authors are careful to note the limits of generalizing across healthcare systems, let alone into industry. But the three factors above — follow-up discipline, protected time, and whether leadership visibly backs the effort — are structural, not medical. They map directly onto the shop-floor patterns this site covers elsewhere:
The honest summary of this research is not "kaizen doesn't last" — 57% of assessed units in this study did reach and hold their objective, a result the authors call worth the investment on its own. It's that sustaining a gain is a second project, distinct from and roughly as demanding as running the event that produced it, and skipping that second project is the single most reliable way to watch a good week of work quietly drift back to baseline.
Last updated: 2026-08-12 | Primary sources referenced: 1 peer-reviewed study | Figures in this article (62%, 24%, 57%, 4.9/2.4 self-assessment scores, 16-quote workload finding) are quoted directly from the cited paper's results and tables, not estimated.