A fleet adopts a published cost-per-lane-kilometre figure for winter wear parts, applies it to its routes, and finds that two of its depots appear to be performing badly and one appears to be performing impossibly well. Writing the boundaries of the published figure before applying it takes an hour. The picture that emerges is a benchmark described in units and surfaces that do not match the boundaries the fleet drew.
This is a method guide for building a benchmark in-house. It covers why published averages behave the way they do, the fields a fleet has to capture to produce its own number, how to normalise across routes and surfaces, and how to use the result without creating a target that nobody can meet.
Why published averages rarely fit
A published average describes a mix no fleet works in.
Cost per unit of work for a plow edge varies with four things: the abrasiveness of the surface, the number of events, the labour rate the fleet pays, and the specification in use.
A published figure blends all four across the fleets that contributed, which means it describes a hypothetical fleet that resembles nobody. Applied to a specific route, it describes the distance between that route and the average rather than any inefficiency.
The same argument applies in the other direction: a fleet that appears to beat the average comfortably may simply work a less abrasive network, and the apparent efficiency will disappear when its conditions change.
What an average is actually describing
An average is a summary of a distribution, and the distribution is the useful part. Where a published figure comes with a range, the range is what tells a fleet whether it is inside the normal band for its conditions or genuinely outside it.
Two properties of the distribution matter in practice. Cost per unit of work is usually skewed rather than symmetrical, because a few routes are much more abrasive than the rest, so the mean sits above the median. And the range is wide: a fleet working clean pavement and one working gravel can differ by several multiples, which means a single figure is not a target at all.
Fields to capture per edge change
Six fields produce everything a benchmark needs, and all six come from records the fleet already holds or can start holding without new systems. The machine and the route it works, with the surface described. The edge specification and its batch reference. The labour hours for the change, separated into removal, fitting and inspection. The work done since the previous change, in events or lane kilometres. And the reason the edge was changed, which distinguishes planned replacement from a failure.
The sixth field is the one that makes a benchmark honest. A fleet that records only the changes will average planned and unplanned work together, and the resulting figure will move when the failure rate moves rather than when the fleet’s efficiency does.
Normalising for route and surface
Normalising means expressing the cost against the work done and the conditions it was done in, rather than against the calendar. Two normalisations are enough for most fleets: cost per event or per lane kilometre, and a surface classification applied to each route.
The surface classification does not need to be elaborate. Three or four categories — clean pavement, gritted pavement, sealed coarse surfaces, and unsealed — are sufficient to separate the routes whose costs are not comparable. The purpose is not precision but grouping: a fleet that compares a gravel route with a paved one is measuring the surfaces, not its own performance. The measurement that supplies the wear side of the figure is described on the blade wear measurement page.
Building a two-season baseline
Two seasons produce a benchmark; one produces a figure. The first season establishes the fields and the habit of recording them, and it also exposes which routes the fleet cannot yet describe accurately. The second season produces the first genuine comparison, because it is the first time two measurements taken on the same basis can be set against each other.
At the end of the second season the fleet has a rate per route group, the variation around it, and an explanation for the extremes. That combination is what a benchmark is; a single number is not. The rate also feeds the fleet’s longer planning, and the way it is used in a multi-year view is described on the procurement forecast page.
Using benchmarks in tenders
A benchmark changes what a fleet can ask for in a tender, and the table below sets out how each use works.
| Use | What it establishes | What to state with it |
|---|---|---|
| Setting an expectation | The interval the fleet currently achieves | The surface and event count it was measured on |
| Testing a bid’s credibility | Whether a quoted interval is plausible for the route | The benchmark rate and the variation around it |
| Comparing lifecycle offers | Whether a higher price buys a proportionately longer interval | The labour and downtime assumptions used |
| Reviewing performance after award | Whether the delivered interval matched the bid | The same fields that produced the benchmark |
The second row is where a benchmark earns its keep in procurement. A fleet that can say what interval its routes actually deliver can ask a supplier to explain a figure that differs from it by a wide margin, and the conversation moves from assertion to evidence.
Sharing data across departments
Benchmarks improve when more than one depot contributes, and sharing them raises three practical questions.
- Whether all depots are recording the same fields in the same units, which is a prerequisite rather than a detail.
- Whether a depot’s rate is attributable to its conditions or to its practice, which requires the surface and route data to travel with the figure.
- Whether the comparison is being used to improve the fleet or to rank depots, since the second changes what gets reported.
The third question decides whether the data is usable. A benchmark used as a league table collects optimistic entries, while one used to find the routes that need attention collects accurate ones.
Governance and data quality
Three practices keep the benchmark trustworthy over several seasons.
- A single definition of each field, written down, so a labour hour means the same at every depot.
- A review of the raw records each season rather than only of the summary, since errors in a rate are invisible once it is averaged.
- A note of any change in specification, route or method, so a step in the benchmark can be explained rather than treated as a decline.
The second item is the one that catches a common problem: a depot that records labour from memory rather than from a job record will produce a rate that looks good and cannot be used for anything.
What to do with an outlier
An outlier is the most useful output a benchmark produces, and it should be investigated rather than excluded. Three explanations account for most of them: the route is genuinely more abrasive than its group, the machine has a condition that loads the edge unevenly, or the depot’s records are not being taken the same way.
Each of those has a different response, and the order to check them in is the cost of the check. Confirming the records takes minutes; confirming the machine condition takes an inspection; confirming the route takes a season of work data. Running the checks in that order is what stops a recording error being treated as a fleet problem.
When an external benchmark is still useful
External figures keep two roles. Before a fleet has its own data, an external range is a useful sanity check on its first measurements, provided the fleet reads it as a range rather than as a target. And once a fleet has its own rate, an external comparison against similar networks can show whether its performance reflects its conditions or its practice.
Both roles require the fleet to describe its own conditions precisely, which is the work this article is about. A fleet that cannot state its surfaces, its event counts and its labour rate cannot use an external figure for either purpose.
Publishing your own numbers
A fleet that has built a benchmark can contribute to the wider picture, and doing so improves the averages that other fleets use. The contribution is only useful with the boundaries attached: the surfaces, the event counts, the labour rate and the definitions used.
Publishing with the boundaries is also the best protection against being measured against a figure that does not apply. A rate that arrives with its conditions stated cannot be misapplied, and it gives the fleet a defensible position the following season. Cost and performance research for winter operations is published by the Transportation Research Board, procurement practice by the National Association of Counties and the American Association of State Highway and Transportation Officials, network guidance by the Federal Highway Administration, and fleet practice by the Snow and Ice Management Association.
SENTHAI manufactures plow edges, wear plates and carbide inserts in Rayong, Thailand, and produces edges to the pattern and profile each machine requires, which is what makes a cost rate comparable across a fleet. Fleets building a first benchmark can raise the edge specification details through the contact page, and the reporting format the figure feeds is described on the cost per lane kilometre page.
FAQ
Why are published blade cost averages unreliable?
Because they average across surfaces, event counts and labour rates that differ more between two routes in one fleet than between two fleets. A published figure describes the mix behind it rather than the conditions a specific fleet works in, and applying it to a route outside that mix produces a target nobody can meet.
What should be captured at every edge change?
The machine, the route and its surface, the edge specification and batch, the labour hours by task, the work done since the last change, and the reason for the change. Those six fields are enough to produce a rate, and they come from records the fleet already keeps.
How long does it take to build a usable benchmark?
Two seasons. The first establishes the fields and the recording habit, and the second produces a rate that can be compared against the first. A single season gives a figure; two give a rate and the variation around it.
Can in-house benchmarks be used in a tender?
They can be used to set expectations and to test whether a bid is credible, provided the benchmark is defined on the same basis as the bid. Using one without stating the surface, event count and labour assumptions behind it invites a dispute about the benchmark rather than about the bid.
Is there still a place for external benchmarks?
Yes, for two purposes: a first-year sanity check before the fleet has its own data, and a comparison against similar fleets working similar networks. Neither replaces the internal figure, and both are more useful once the fleet can describe its own conditions precisely.
Send the edge changes you have recorded and the routes behind them. The SENTHAI engineering desk will confirm the specification details the benchmark needs so the rate is comparable across the fleet.