Monitoring a Dairy Farm’s Performance

Gerard Cramer, DVM DVSc

College of Veterinary Medicine, University of Minnesota

Summary

1. Monitoring and evaluation of dairy farm performance is a key component of a dairy production medicine program.
2. Evaluating records should focus on evaluating current performance and determining if further action is warranted
3. Monitoring should focus on asking questions related to process and not solely based on outcomes
4. There are various pitfalls of data that a veterinarian needs to be aware of to avoid drawing inaccurate conclusions
5. There is no perfect metric and a combination of metrics is required to get a true picture of dairy’s performance.

Introduction

Monitoring a dairy farm’s performance is a valuable service veterinarians should provide as part of a production medicine service. This production/performance medicine based service typically focuses on the optimization of the

• The health status of the herd
• The productivity of the herd.
• management processes/protocols to produce safe dairy and meat products
• The sustainability of the dairy industry with a focus on animal/people welfare and the environment.
• Overall dairy farm profitability.    

To properly monitor these activities and processes requires an understanding of the process and the data collection procedures.

Content

Reasons for Monitoring

The main reason for monitoring is to be able to make data based decisions on a dairy. Monitoring focuses on the observation of current procedures, events, or outcomes for the purpose of determining if actions are required. This monitoring process includes data collection, evaluation, and communication this process should occur in a frequent and systematic manner.  The objectives of monitoring are to:

• Recognize “normal”

• While recognizing that these normal/abnormal parameters are occurring in a biological system and cows are not homogeneous within dairies.

• Evaluate the impact of a planned change in a management or performance area
• Find unintended drifts or declines in procedures or performance
• Determine the potential causes of abnormal performance.  

Within these objectives, it is important to avoid inaccurate conclusions that lead to delays in action or alternatively, taking action when problems do not truly exist.

Goals vs Metrics

Goals are target levels of performance that farms are trying to achieve and should typically relate to a measure of profitability. Metrics are a type of measurement or a set of measurements that quantify results and are used to gauge some quantifiable component of dairy performance. Metrics that are monitored are typically numbers that represent some type of process and are always important in achieving a goal, but are not synonymous with the goal themselves. Goals should not the default thing to monitor as typically there are typically several processes that need to occur to reach this goal. To truly use monitoring to detect a change in a farm’s process it is necessary to monitor as close to the process a possible, not just the goal. Therefore the focus for analysis should be on process monitoring, not outcome monitoring.

Common Issues Related to Monitoring

Data Quality

While it is tempting to automatically start analyzing data when you receive it, the most important step to prevent erroneous conclusions is to check data quality. Data entry errors are common, yet typically do not impact the ability to draw conclusions unless they occur on a systematic basis or make data analysis more difficult. The most common problem is not having all the data and missing historical records (culled animals for example).  Tools to check data include examining minimums, maximums, and distributions to ensure they fit within expected values. Creating scattergraphs and histograms from the date are great tools to do this.

Lag

Lag is the difference between when an event occurred and when it is measured. Lag is inherent in many of our metrics but a good metric is one that minimizes the lag.

Momentum

Momentum occurs when recent changes (in either direction) are not reflected in the data because the metric contains a long time period or quantity of historical data.  

Bias

Bias is the systematic error in the data collection/analysis or interpretation of data and can lead to incorrect conclusions. Typically in dairy records bias results when cows are wrongly included or excluded from the calculation of a metric.

Variation

Most metrics used to monitor performance in the dairy industry are averages. Averages should not be interpreted without understanding the underlying distribution of the data to get a sense of the variation. Standard deviations are one tool used to quantify variation. Variation is normal in many processes and random variation inherent in any biological system. Special cause variation is referred to as noise and should be monitored to determine if a system is in control or out of control. Planned management changes can also affect averages and introduce variation and these should be accounted for in the analysis.  

Data Type

Typically data are classified into either qualitative or quantitative data.

Qualitative data

Qualitative data are typically categorical or descriptive and can be nominal (no ranking: male vs female) or ordinal (ranking, small vs large). This type of data is analyzed by counting occurrences within the data and displayed using frequency distributions.

Quantitative data

Quantitative data is data that can be counted or measured and can be continuous or discrete.  Discrete data are either counts or categories that contain intervals which have a specific meaningful/quantifiable gap (BCS). Continuous data is any data that contains no limit on values and typically are summarized using averages and measures to account for variation.

Ratios, proportions, risks, and rates

Discrete and qualitative data are typically analyses using ratios, proportions, risk or rates.

1. Ratios compare 2 numbers and the numerator and the denominator are mutually exclusive

1. male vs female

2. Proportions range from 0 to 1 or 0 to 100% and measure the frequency of something. With proportions, the numerator is a component of the denominator.  

1. Number of dry cows as a % of total cows in the herd

3. Risks are similar to proportion but a time period is typically implied or specified.

1. Risk of metritis in cows that calved between 30-60 days ago.
2. Risk does not mean negative.

4. Rate refers to the speed at which an event occurs

1. Includes a time component in the denominator

1. Number of new events of something divided by the population time at risk
2. True rates are not commonly used in dairy production medicine but rate is commonly confused with risk.

Sample size

For most herd sizes comparisons of metrics between time points or between individuals/treatment outcomes is difficult and not practical if one relies on statistically significant differences alone. Relying on numerical differences alone is also not appropriate as in most cases there are biases between the groups and creating a valid comparison between groups typically creates problems with sample size as comparison groups become too small. To counter this there are strategies that use less stringent confidence intervals (68%) or process control criteria. In most cases, the best approach is to weigh the cost of making a change when none exists (type 1 error) compared to the cost of not making a change when a problem exists (type 2 error).

Evaluation performance

The first step in evaluating a diary’s performance is to ask the right questions. This requires understanding the process and the outcome and then asking a question that evaluates the process and not just the outcome. Typically this involves asking a few key questions and only proceeding to ask additional questions if the key question indicates a problem. This is different than randomly evaluating graphs or reports in software programs. This question based approach is a more systematic approach and prevents getting stuck in rabbit holes. An additional consideration is that for most processes there are no perfect metrics to answer the questions of interest and a combination of metrics is required to draw appropriate conclusions.

Guided Learning Activities

Monitoring Milk production

• Develop 2-3 key questions that you would use to monitor a dairy’s milk production on a regular basis

• Focus on routine monitoring, not problem investigation.

• Answer the following questions

• Identify what data you need to answer your question
• If a parameter exists for your question, what pitfalls does it have?
• How frequently would you monitor this?

• What is the command or process to get this parameter in your dairy management software program?

Self Guided Learning Activity

Create List of Key Monitors

Using the approach outlined in the guided activity create a list of metrics you would routinely evaluate prior to your regularly scheduled visit to a dairy.  Include all areas that you think are important to monitor to meet the goals of a dairy farm.

Additional Resources

• Overton, M.W. 2011. Dairy Records Analysis and Evaluation of Performance. C.A. Risco and P.M. Retamal, ed. Blackwell Publishing Ltd., Oxford, UK. (Google drive link to chapter)
• Stewart, S., S. Eicker, and P. Rapnicki. Making More Effective Use of Your Data 2011 Western Dairy Management Conference. (website link to the paper) 

Acknowledgments

Adapted from Dr. Overtons’s chapter listed above