Sanden Kotsu — Japan
At the Japanese bus operator Sanden Kotsu, treated service buses were compared against untreated control vehicles running in the same period — not against a historical baseline. Over three months the treated buses averaged 7.8% better fuel efficiency. It is the only dataset in our records with a genuine control group, which makes it methodologically the strongest Japanese trial we hold.
A control group is the difference between "the numbers got better" and "the numbers got better because of the treatment" — and this is the only trial in our records that has one.
| Operator | Sanden Kotsu, Shimonoseki, Japan — public bus operator |
|---|---|
| Fleet context | 240 vehicles, of which 183 are service buses |
| Vehicles measured | 2 service buses, July and August 2023 |
| Comparison | Against untreated control vehicles in the same period |
| Measured quantity | Fuel efficiency in km/L |
| Duration | Three months |
| Result | +9.84% and +5.75% individually; +7.8% average over three months |
Every measurement, not just the best one
| Bus | Month | Control group (km/L) | Treated (km/L) | Difference |
|---|---|---|---|---|
| Shimonoseki 230 A 5020 | July 2023 | 2.501 | 2.747 | +9.8% |
| Shimonoseki 230 A 5275 | August 2023 | 2.694 | 2.849 | +5.8% |
| Average over three months | +7.8% |
Method before result
Treated service buses were compared with untreated buses from the same operator running in the same period. That is the crucial design difference from every other trial in our records: season, weather, route conditions, fuel batch and traffic apply to both groups equally, so they cancel out of the comparison instead of contaminating it.
Fuel efficiency was recorded in km/L from the operator’s records over three months, with individual monthly figures for July and August 2023 and an averaged result across the full period.
The limits of this trial
We publish these next to the result rather than in a footnote. If a limitation would change how you read the number, you should read it at the same time as the number.
Known limitations
- Only two treated buses are documented individually, from a fleet of 240 vehicles.
- The composition of the control group is not documented in the records we hold — how many vehicles, which routes, how they were matched to the treated buses.
- We hold a derived summary of this trial, not the operator’s raw records. For publication with figures we would want the primary data.
- A three-month window on service buses is short relative to the year-round variation in urban bus operation.
Reading this evidence in context
This is the design we now recommend for every new trial, including the ongoing European ones: treat some vehicles, leave comparable ones untreated, run them in parallel, and compare the two groups rather than comparing one group against its own past.
The result — around 8% — sits below several of our uncontrolled trials. With one controlled trial we cannot claim that control groups systematically produce lower figures; what we can say is that this is the number we would defend most confidently, because the comparison group absorbed the season, the weather and the traffic along with it.
Questions operators ask about this trial
What is a control group and why does it matter here?
Why is this result lower than some of the others?
Can we run a pilot this way?
Provenance
- Operator
- Sanden Kotsu, Shimonoseki, Japan
- Comparison
- Against untreated control vehicles in the same period
- Measurement period
- July–September 2023
- Source
- Japanese field-study summary — derived record; operator raw data not held
- Last checked
- 15 August 2026
The only baseline that matters is yours
Every trial on this site was run on someone else’s vehicles, duty cycle and baseline. A NanoEFX pilot measures yours — and we will help you design it with a control group so the result stands up.