Data-driven decisions use facts and metrics to help you make smarter decisions. But what if the data you’re using is bad?
The line between good data and bad data is easily blurred. Simply, bad-quality data is inaccurate, out-of-date or hard to access.
Since the 1990s, Terry Griffin, agricultural economist and precision agriculture specialist at Kansas State University, has studied yield monitor data from thousands of fields. He offers stark observations.
“If it comes off a combine, it’s bad data,” he says. “It was measured incorrectly by a huge machine wobbling through a field. If we have 50,000 observations, 20,000 are misleading. That’s not the end of the world, as we can clean that data and detect erroneously measured observations, but normally farmers don’t do that.”
POST-HARVEST ANALYSIS
Typically, it takes about 45 minutes to process a field, Griffin says. This includes flagging data outliers for analysis or omission. To the machine data, you want to incorporate the meta data, which can include personal observations, random field notes and more.
Meet with your team once harvest is complete and have them bring maps and memories, suggests Farm Journal Field Agronomist Ken Ferrie. He and Griffin provide these tips for analyzing yield data:
1. Do not send yield maps to be analyzed by someone outside your farm. Only your team, including consultants involved with decision making, know what happened.
2. Use raw data, not contoured or “smoothed out” maps. Yield maps must show yield spatial variability. Set up seven to 10 ranges (yield increments) with equal points, whether the variation in a field is 15 bu. per acre or 50 bu.
3. Confirm consistent conditions. The combine should be operated in similar conditions to which the yield monitor was calibrated. For example, Griffin says, an abrupt change in speed results in a spike, up or down, in yield data. “These outliers can be seen by examining percent change in ground speed between yield data observations within a transect. Flag this data.”
4. Overlay soil type and topography on yield maps. This will help identify the cause of yield swings.
5. Compare yield data with pest records. Specifically analyze ear size and ear count.
6. Don’t rely on average yields. Dry areas will yield higher in wet years and vice versa. “Averages don’t necessarily reveal management zones,” Ferrie says.
