Is there “fake” data on farms? Do we know how to convey the importance of good record keeping?

There are a wide variety of data management programs on the market today and almost all farms use one of these programs in their day-to-day operations. Despite the continuous and widespread use of these programs, not all those involved are aware of the importance of the reliability of the data collection that we subsequently enter into the program.

The most common inaccuracies in data collection

All farms correctly record the number of piglets weaned, the most important production data. Apart from that, not all farms have the same reliability in data collection. Some of the most common errors are:

• Not recording all stillborns.
• Recording piglets born very weak as stillborns.
• Failure to record piglet deaths in lactation and/or their causes.
• Incorrect recording of the causes of piglet deaths.
• Failure to record piglet fostering between sows.
• Equalizing litter numbers and subsequently recording births.
• Collecting data two or three days after farrowing and recording the piglets present then.
• Failure to record or incorrectly recording the reasons for sow losses (e.g., sows culled for lameness are recorded as dead).
• Failure to remove euthanized or culled sows from the database (“phantom sows”).

The most common reasons for incorrect data collection

In the vast majority of cases, the failures that I detect in data entry are not intentional, the managers are simply not aware of the importance of the reliability of the data. These would be the most common reasons:

• Data collection protocols inherited from previous farm managers.
• Time savings in management tasks.
• Productivity bonuses that are not optimally designed.
• Employees' self-imposed demands to deliver the best possible results.
• Failure to record data immediately (automatically or manually).
• Poor training on the importance of correct data collection.
• Lack of clear guidelines when recording data.

The importance of data collection reliability

To improve the farm's production results we must identify the points with the greatest potential for improvement, and to do so it is critical that the reliability of the data entered allows us to identify these critical points.

Let's look at just a few examples:

If we want to improve piglets weaned per litter we will have to assess whether it is more feasible to reduce pre-weaning mortality or increase piglets born alive (Figure 1).

• To increase piglets born alive we will have to assess whether the critical point of improvement is to increase the total born or reduce stillborns.
• If not all the stillborns are recorded, we could think that the problem is the lack of total born (and we would review the management protocols during insemination) when perhaps the problem lies in excess stillborns and we should review the management protocols during farrowing.
• If piglets born weak are recorded as stillborns, we would believe that the problem is the lack of born alive and we would be lowering the real percentage of pre-weaning mortality.
• If piglet fostering is not recorded, it will be more difficult to identify the sows that have many piglets die in lactation repeatedly and should be culled. Since these sows end up weaning piglets received from other sows, their poor performance may go unnoticed.
• If we cull sows with many returns to estrus but do not record the reason for culling, it will be more difficult to identify and solve cases of low farrowing rates.
• If there is an excess of piglets over 15 days old that die by crushing, we almost certainly have a problem with the temperature control of the farrowing rooms (rooms that are too cold or nests that are too hot). But to detect this, we must record the piglet deaths, with their real reason and date.
• If I do not correctly record the sows removed from the farm and do not delete them from the program, we will have "phantom sows”, females that are no longer present on the farm but that the program still considers in its calculations. These animals generate an increase in NPD (non-productive days) and will cause us to waste a lot of time looking for what could be the problem.

How to detect cases of low reliability in data collection

It is not always easy to say with absolute certainty that data collection on a farm is not as reliable as it should be, but these are some of the signs that can help us detect it:

• Farms that are among the best in their class in pre-weaning mortality and, at the same time, among the worst in stillborns. Are they recording piglets born alive with low viability as stillborns?
• Farms with low total born and very low stillborns. Are they recording all stillborns?
• Farms with a very flat curve on total born by parity (Figure 2) Are they equalizing litters and then recording data?

• Prolificacy distribution: Is there a litter size with a much higher percentage (more than 25% frequency) than the other litter sizes?

• If we analyze the statistical deviation of total born, born alive, and weaned, the deviation of total born should be greater than that of born alive and born alive greater than that of weaned (Figure 4). This does not always occur on farms with low reliability in their data collection.

Conclusion

The use of management programs provides numerous advantages and uses such as creating task lists and production reports. Based on the analysis of production data, we can detect critical points with potential for improvement, with the aim of increasing the farm's productivity, which would result in reduced production costs.

In order to improve production results through effective technical audits, it is essential that the data entered into the management program accurately reflect what is actually happening on the farm.