LLamasoft Supply Chain Blog

← Back to Blog

How to use benchmark data to calibrate your rates for design

By John Trestrail  August 12, 2015

Any modeler can tell you: a huge part of the supply chain modeling process is spent collecting and reconciling data. Hours spent to find data, adding it to the model, to only find that maybe the data was not quite as accurate as originally thought.

When building models to optimize your network to uncover savings in cost, service, risk and sustainability, there will invariably be a point in which a modeler runs into a blank space: the cost data for lanes that aren’t in the current history. Now what?

One approach is to run a regression analysis comparing historical rates to distance to fill the data gaps for new lanes.  This yields a cost per mile that can be used to estimate the transportation costs.  Distance is a major driver in most transportation modes.  But how well does it account for short-distance lanes?  Can it tell the difference between a headhaul and a backhaul rate?  Does it reflect the negotiating leverage of regular, large volume lanes?


Another approach is to pull benchmark or reference rates from a third-party data set.  This provides modelers with a good understanding of what shippers are paying on average in the marketplace.  But will your company perform above or below the benchmark?  Will using historical averages for existing lanes and benchmarks for new lanes bias the model towards the old or the new?

Our recommendation for the best practice is to use both regression analysis AND a benchmark rate data set.  Instead of regressing your historical rates simply against distance, look for the correlation between your rates and the benchmark rates.  The benchmark rates are composed of more pricing factors than merely distance.  You should find that generally your rates are above or below the benchmark rates.  You can than safely pull benchmark rates for new lanes and multiply them by a factor to calibrate them to be in-line with your historical rates.


Also, you can refine your regression analysis by isolating other factors such as:

  • Geographic regions
  • Lane volume levels
  • Equipment types
  • Service levels
  • Seasonal time periods
  • Hazardous vs. non-hazardous
  • and more



Click here for more info about the benefits of using benchmark data for supply chain modeling.