Generating Greater Value from Herd Test Data


Livestock Improvement Corporation is a leading New Zealand Agritechnology company driving farm productivity and profitability improvements through its commitment to research and innovation. As a farmer owned co-operative, it's our vision to improve the prosperity and productivity of our farmers. LICs goal is to empower farmers with the genetics and information needed to produce superior livestock, information to improve livestock performance, and the hardware and systems that will improve on farm productivity and decision making.

LIC is the biggest provider of Herd Testing (HT) in New Zealand. We conduct the service for 72% of all New Zealand dairy farmers. Herd testing is the process by which dairy farmers estimate the productive output of individual cows in their herds. LIC’s staff visit each farm 2-4 times a year to collect milk samples and deliver them to LIC laboratories. Well over 20 million individual milk samples, collected from approximately 3.5 million cows, are analysed on LIC’s MilkoScanTM FTIR milk analysers every year. The subsequent data set includes the main production trait measurements like milk yield, fat, protein and lactose content, as well as somatic cell count (SCC) and some corresponding metadata.

The primary purpose for herd testing is to provide dairy farmers with information they require to make culling and replacement stock selection decisions. The HT data also contributes to the ‘industry good’ National Animal Evaluation system for dairy sire evaluation, developed and operated by LIC on behalf of New Zealand dairy farmers. LIC on the other hand uses the data for the analysis of the dairy cow population in order to assist farmers with on farm selection decisions, as well as the evaluation of LIC’s own sires in its sire proving scheme. LIC also use the HT data in wide range of R&D projects in the fields of genomics, farm management and farm automation.

LIC considers itself an heir of the very first herd test service provider in NZ and recently celebrated its 100–year anniversary of this core service. About a billion raw HT records collected over the last 50+ years are stored in LIC’s databases waiting for keen researchers.

In essence the challenge is developing a methodology for sensible multivariate analyses of the raw HT data. Specifically we see the following three areas for improvement:

  • Data Quality Control
    We need to learn how to differentiate “technical” outliers from “biological” ones. Delineation of these outliers will reduce data loss in our resulting analysis and enrich our data with biologically important variations, thus allowing more specific analytical methods. This is our primary reason for promoting this challenge
  • Analytical Process Control
    Current methods to account for outliers are limited to removal using statistically derived cut-offs. We require a method to detect and analyse these outliers in real time, which determines the cause of each outlier and accounts for it appropriately.
  • New ‘Big Data’ Analytics
    We want to uncover the potential of using these massive data flows to develop new analytical methods to uncover undetected variation and lift the power of our selection models.

We have data as described above and access to serious computational facilities. LIC can also provide experts in the field of herd testing and animal breeding who understand the underlying business need. These experienced applied scientists and business managers that believe in power of science are available to clarify the problem.