The rotation idea
With three edge types (A, B, C) and three route groups of similar character (1, 2, 3), assign them in a Latin-square pattern so each type is used on each route group over three periods.
| Period | Route group 1 | Route group 2 | Route group 3 |
|---|---|---|---|
| First third of season | A | B | C |
| Second third | B | C | A |
| Last third | C | A | B |
Edges are changed at the period boundaries, or when worn out, whichever comes first, and each edge is logged by hours.
Data to record
The data set is what makes a three-way comparison possible, and every field has a job.
- Edge ID, type, supplier and price. Without the ID, the rest is anonymous.
- Truck, route group and period.
- Hours of plowing. The denominator for every comparison you will make.
- Wear readings and photos at fixed intervals.
- Failures, chips and strikes, with locations.
- Fitting time and any problems.
- Driver comments.
Decision rules to write in advance
Write the rules before the trial starts, because every rule is easier to agree while no result is on the table.
- The main measure, such as cost per hour of use including fitting.
- A minimum number of hours per edge before a reading counts.
- How to treat early failures: as data, not as exclusions.
- What difference is large enough to act on, for example a margin your board accepts.
Pitfalls
A three-way comparison fails quietly, and usually for one of these four reasons.
- Letting drivers pick edges. Preference is a variable you did not control, and it correlates with the result.
- Comparing calendar time rather than hours. Storms differ, and dates hide how much work each edge did.
- Changing equipment or routes mid-trial without recording it. The unrecorded change then explains the result instead of the edge.
- Stopping after one storm. One storm is weather; three are a pattern.
A note on fairness
Even with rotation, trucks differ: a newer truck, a better driver or a different plow can give one edge type an advantage. Log the truck and driver pool with each edge, and look at whether the differences between edge types are larger than the differences between trucks. If they are not, say so in the report.
Report
Summarise the design, data and result in a page, including limitations such as the number of edges and the length of the season, and keep the raw data.
Limits
A single season cannot settle every question. Treat the result as evidence for your own routes.
Boundary: No county result is reported; the design is a method, and one season cannot settle every question, so state its limits when you report.
Further reading on this site.
Studded, Segmented and Insert Edges for Ice Control: Comparing Three Approaches ↗A blade performs well on one route but poorly on another. How should we decide whether to standardize it? ↗Carbide snow plow blades ↗311 Complaint Logs as an Early Warning for Edge Problems ↗Snow Emergency Plans and Depot Readiness: Why Edge Stock Belongs in the Plan ↗Reading Public Bid Tabulations for Cutting Edges: What Unit-Price Spreads Tell You ↗
