Intuition Doesn’t Match Data

I’m a big believer in the practice of split-applying nitrogen — specifically sidedress-applying a portion of the nitrogen after planting. Being able to apply part of the nitrogen closer to the plant uptake always made sense to me because the nitrogen is less available for loss before that time. However, oftentimes it is difficult to find an advantage to sidedressing in the data.

It can drive me crazy having access to millions of acres of data and not have the data confirm something I believe! Or worse yet, show the exact opposite! For example, there have been plenty of years when fall-applied nitrogen appeared to do as well, if not better, than spring- or split-applied.

Here are some of the lessons I’ve learned over the years. First, there can be “sectional bias” in the data. For example, growers “select” their heavier soils for fall applications for agronomic reasons — ability to hold ammonium nitrogen. Likewise, they might choose their lighter soils for sidedressing. Simply comparing yield by nitrogen timing can lead to a wrong conclusion. Looking at nitrogen timing by soil type can be an obvious next step in correcting for selectional bias.

A second reason that sidedressing might not show up as being superior in data analysis is due to the fact that nitrogen applications might not be the yield-limiting factor! If applied nitrogen is not what’s yield-limiting, then timing of application won’t likely matter or show up in data analysis. In much of the high-organic-matter portion of the U.S. Corn Belt, university researchers estimate that 50% to 70% of the nitrogen that feeds the plant is soil-supplied and not fertilizer-supplied. It’s no wonder there are years when the relationship between applied nitrogen—regardless of timing — and yield is insignificant.

The last reason is the one I hate — it’s that I could just be wrong. When believing so strongly in a concept, it’s very difficult to consider that you have it wrong. But I’ve been there with sidedress nitrogen applications. From the data, the best explanation is that there are years when, in some areas, we don’t get enough rain after the sidedress nitrogen application to move that newly applied nitrogen into the root zone. In those drier years, the top of the soil profile is dried out and the plant is feeding deeper. You’ve all been to a field day where someone used a backhoe to do a “root dig” to illustrate this point. You may have walked away with a mental note that what is below ground is bigger than what is above.

What do you do when data analysis doesn’t support your own theory? For me, the first answer is to keep digging in the data. I’m still a believer in split-applying nitrogen — including sidedressing a portion — but living through those dry summers would lead me to get the work done early. My second answer is to focus on pounds of applied N per bushel. Can I produce the same or more with less applied nitrogen by altering timing?

Got Data?
  1. What are some of your theories or beliefs that are not showing up in your data? Can you dig any deeper?
  2. How might you weather-proof your nitrogen program? How might your data help in the process?


Originally published in Corn and Soybean Digest.

Can’t Save your Way into Prosperity

Sometimes life events leave a market for generations. The Great depression created generations of frugal farm family survivors. Feeding a family was a challenge — holding on to a farm was almost impossible for many. The “frugal stamp” wasn’t just left on the parents but their children and many times passed on to another generation.

Did your grandmother save everything? Save leftovers — not in Tupperware but in leftover butter or other plastic food containers? Save old jeans to have material for future patches. Did your grandfather save pieces of old lumber for the next project? One of the most popular radio talk shows, hosted by Dave Ramsey, encourages being frugal to climb out of debt.

In tight economic times, it’s easy to want to take a frugal approach on crop inputs. In the November issue of CSD, an article highlighted DuPont Pioneer’s findings that many fields sampled have below optimum soil test P and K levels, with their expert calculating over $4 billion in lost revenues for growers.

The elephant in the room is the negative impact that high cash rents can have on maintaining the productivity of many farms. In most markets, the competition for land is so fierce that growers resort to penny-pinching on P and K applications. Many don’t want to leave any nutrients for the next renter if they get out-bid.

The article also highlighted the new higher-yield reality of how much has to be applied to keep up with nutrient removals. The “old” shotgun of 400 lbs. of dry fertilizer every other year can mine soil test levels quickly.

I believe the yield loss associated with being frugal is far more significant than most agronomists and growers realize.

I believe the yield loss associated with being frugal is far more significant than most agronomists and growers realize. I believe that for many growers the “optimum” soil test level is even higher than university definitions of “optimum”. This chart shows one way to analyze data across a grower’s 1,700 acres of soybeans — examining the relationship between low to high yield acres and the corresponding soil test P and K levels.

If these were your yield results, what would consider your optimum soil test levels? That’s the power of your data — it can lead you to customize your approach to what is best on your farm versus a statewide average.

There are many strategies to address the fertility needs of the crop. Strip tillage and deep banding of nutrients is a great way to compensate for low fertility fields and maximizing return for nutrient dollars invested. Irrigated sand requires that we spoon-feed nutrients.

For most nutrients, feeding the crop is a combination what we apply, including manure, and what the soil supplies. We shouldn’t measure our success by whether we raised soil test levels but by our yields and cost/bushel. I believe your data will lead you to find the balance in your approach to input decisions. In corn and soybean production, you can spend yourself poor but you can’t save yourself into prosperity.


Originally published in Corn and Soybean Digest.

Dig Deep into Data

One of the services Premier Crop provides growers with is community or group data analysis from a regional basis. Growers can confidentially and anonymously share their field data with other growers in their region through written data-sharing agreements. Group data analysis complements field- and grower-level analysis; growers receive actionable field intelligence they can use on their own fields.

At Premier Crop, we are constantly challenging growers and their advisers to dig deeper into their data. For example, a few years ago one of our customers, a grower we will call Bob, wanted to use data to make decisions on trying to capture the 40-cent-per-bushel early harvest pricing premium his local ethanol plant was offering. Bob’s data analysis considerations started out with the yield penalty for planting earlier-maturity hybrids in his area.

Current and historic data analysis of his and other growers’ data in Bob’s adviser’s 100,000-plus acres of multiyear group-data database showed a consistent 15-bushel-per-acre penalty (185 vs. 170) for planting early maturities over the past five years. The adviser and the grower calculated that he needed a 35-cent-per-bushel premium to break even.

Bob’s second question was whether the yield penalty was universal across all soil types. His adviser added soil type to the relative maturity query and the results showed a 10-bushel disadvantage on the lighter soil types, which was about 25% of the grower’s operation. The overall yields weren’t as high on the lighter soils, which might be part of the reason why the penalty was lower.

Planting earlier maturities on lighter soil-type fields dropped the breakeven premium to 23 cents per bushel — offering a potential $30-per-acre advantage to planting an early-maturity corn and capturing the premium.

The adviser suggested they dig even deeper and look at planting date, as well. The adviser’s instincts led to an even more powerful discovery. The reward to earlier planting was (18 bushels-per-acre) more significant on the lighter soil types than the normal and heavier soil types (10 bushels per acre).

This last discovery didn’t make sense at first. Like many growers, Bob liked to plant his “best” fields first — best being those with the highest yield potential. Because the lighter soils had less yield potential, they tended to be treated as the secondary ground. But as Bob and his adviser considered what the data analysis was showing, they realized that later planting dates meant pushing pollination back into the hotter and drier weeks of summer. The lighter soils didn’t have as much water-holding capacity as the normal and heavy soils, which were better able to handle stress without losing yield.

So, did Bob plant earlier-maturity corn on his lighter soils first to capture the early harvest fall premium? More than likely, but I’m not positive. However, I do know the entire process led to the purchase of an additional planter. The planting date data analysis was so overwhelming it was easy to justify the addition of equipment needed to plant all his acres earlier. As convincing as the analysis was on the lighter soils, Bob was willing to potentially lose 10 bushels an acre on his best soils by switching planting order.

And imagine his delight in being able to use data to justify to his wife why buying more equipment was the right answer!


Originally published in Corn and Soybean Digest.

Share Your Farm Data?

It’s hard to avoid hearing about the promise of “big data.” Thanks to Edward Snowden’s proposition, we need your non- partnership to use your data to revelations, it is also easy to spin conspiracy theories. There are many big-data analytics examples cited, such as Amazon and Netflix. They suggest books and movies we may enjoy based on what we have “liked” in the past or what other people who seem similar to us like. Google, the National Security Agency and others evidently collect data about what we search, what and whom we email, and much more.

There seem to be several “values” from big-data analytics. Many companies’ goals are to monetize data through better value propositions to their customers, like selling advertising, how they position products, differential pricing, etc. Another goal is to reliably predict behavior. For example, Bob shares a common background and behavior as Tim, and Tim likes this brand; therefore, Bob must like it, as well.

Reliably predicting behavior, better product value propositions and differential pricing are all examples of how companies could use your agronomic data. For example, your field’s soils are X, Y, Z, and hybrids A, B, C outperformed hybrids Q, R, S on 85% of the X, Y, Z soils; therefore, A, B, C is the best value proposition for you. The fact that your soils are X, Y, Z is “public knowledge” — because the soils data- base is public.

But a company might say, “to really perfect our value proposition, we need your non-public agronomic data. Why don’t you send us your historic yield data, your fertility data, your management information, etc.”

What can it hurt? After all, they only want to help you. The reality is, we all share “our” data with other companies, either intentionally in exchange for a benefit or inadvertently because we wanted a cool “app,” and sometimes the trade-offs are worth it.

For me, the difference between consumer data sharing and sharing your georeferenced agronomic data is profound. Your reading choices might influence what advertising you see; your agronomic data is your “business” data. There is nothing anonymous about GPS data.

Anyone with a tractor guidance system has heard about different levels of GPS accuracy. Any of those accuracies are more than sufficient to provide site-specific data about your fields when you use a documentation system.

For Premier Crop, building a partnership to use your data to make better agronomic decisions has always begun with this foundation: The grower owns the data, data is only pooled with permission, and even pooled data belongs to and is for the benefit of the growers who shared data. But Premier Crop’s history and operating principles don’t mean that’s the right way or the only way to use data to benefit growers.

Sharing your data with seed, crop protection, nutrient or machinery suppliers can make business sense. These companies sell you products that are important to your business and profitability. Sharing your data to help them provide better recommendations may well be worth any trade-off. Most important is to think through those trade-offs and each “partnership proposal.”

Got data?

  1. What are the partners going to do with the data? Perhaps more importantly — what will they assure you (in writing) they will not do?

  2. Some of your partners may be fearful of missing out on the “next big thing” if they provide the wrong answers to your questions. What questions are you asking?

Originally published in Corn and Soybean Digest.

Success in VR Seeding

I try very hard not to fall in the trap of “geek” talk. It’s a special challenge, as we have a company full of geeks! Even though we hire a lot of “normal” people, within a year we typically convert them into farming geeks/ nerds. They still look normal, but they aren’t. I describe what we do as the collision of agronomy and technology. It’s the technology side that leads us to become geeks.

One word I’ve tried to avoid overusing is “spatial” — rhymes with racial — defined as relating to space and the relation- ship of objects within it. Precision ag tools allow us to map spatial differences within a field; yields and fertility are common examples. For many, maps have been a powerful way to visualize the spatial variability that exists in most fields. For some, these maps confirm what we’ve always known to exist — significant differences within fields.

For almost a decade, Premier Crop and our customers have experienced terrific success in creating value for growers with variable-rate seeding prescriptions. Our process is simple. Create management zones by identifying the best part of the field, where you want to be more aggressive (A zones). C zones or other lower letter zones are usually used for the lower-yielding area, where being more conservative with seeding rates might be warranted. B zones end up with the “average” part of the field — neither great nor limited.


Now, with millions of acres successfully and profitably variable-rate seeded, one observation worth sharing is that it is very rare for a management zone to run in a straight line from one end of a field to the other. This management zone map is the rule, not the exception.
In our company’s early years, we did plenty of strip trial approaches where treatments and checks ran across the
entire field. But we came to understand that most strip trials are running treatments across tremendous spatial variability, which is frequently ignored in the analysis. Averaging the results for treatment strips and ignoring the effect of spatial differences tends to mask what is really happening. Gains in the A zones can be “averaged” down by the results from C and D zones.

For many, analyzing variable-rate planting with strip trials
has resulted in conclusions that aren’t positive for the practice. Our results and experiences have been the opposite. Using higher and lower Learning Blocks within management zones and analyzing yield results have led to even faster adoption.

Got data?

  1. Off the top of your head, what types of variability do you think lie across a field?
  2. For what kinds of treatment would it make sense to run trials across an entire field?


Originally published in Corn and Soybean Digest.

Mind the Gap

If you’ve ever traveled to London and been a passenger on their subway system, the “tube,” you’ve heard the phrase “mind the gap.” It’s kind of funny because that’s not how Americans would say it. We’d probably say “watch your step.” The long, straight cars and the curves in the tracks cause gaps between the train cars and the loading platform; therefore, passengers are warned to “mind the gap.”

This phrase is relevant to all of us who are trying to create value from data. The gap we need to mind is the gap in time that exists between when the field is harvested and when results can be delivered to the grower. Historically, the gap exists for several reasons. Many growers tend to be focused on the task at hand in the fall — the physical work of getting the crop harvested, lime and nutrients applied, and in some cases, tillage operations. Downloading yield cards is something they can do later; the same goes for the decisions they will make from the data. Harvest season priorities and decisions tend to focus on where to take a bumper crop when the “wet bin” is full and the line at the elevator is backed up.

Technology offers an answer to close the gap, and the wireless transfer of data can eliminate the gap, but there is a catch — data quality.

I get challenged on data quality all the time. Frequently, when I present at a large conference, there are skeptics and “unbelievers” in the audience. Usually, someone will make a comment about “all the bad yield data out there.” My response is to defend growers. If I’m a grower and I’ve never made a decision using my yield data, after a few years I will likely quit caring about calibration. But Premier Crop’s experience is that when growers understand how they can use their data to make better decisions, they drive everyone involved crazy in getting their data perfect! If their grain cart scales say the field average is 236.7 bushels per acre and the yield file says 229.7 bushels per acre, they expect us to correct the yield file. So while it might be true that many farmers haven’t taken the time in the past to calibrate their yield monitor, there is a logical reason for their lack of attention.

Telematics — being able to send data wirelessly — offers advantages to all of us in agriculture. But in many cases, it can mean we have sped up how fast we move inaccurate data. And you know the saying: “garbage in, garbage out.”

At Premier Crop, our customers are in the thick of delivering 2014 data analysis results to growers. For our customers and staff that means a lot of “fixing” data. But data fixing doesn’t stop at the yield file. Why would data need to be fixed? Many times, it’s because something didn’t get logged correctly or as completely as needed.

Many companies are now offering data analysis solutions that rely heavily on telematics as their service’s backbone. I saw an example recently where the company was providing an independent benchmarking for genetic performance by soils driven by their telematics solution.
The problem is many times not all the data needed to provide quality analysis is logged. Five different growers might have logged that they were planting Channel 207-13. But that hybrid has five different trait combinations ranging from conventional to SmartStax (see table for examples). Averaging the yield for all five trait combinations into one number isn’t an accurate comparison.

Quality data analysis relies on quality data. And quality data requires hard work and attention to detail. Even though we can use technology to make the timing gap smaller, we all must still “mind the data quality gap.”


Originally published in Corn and Soybean Digest.