Analytics, Strategy, and Agriculture

Month: July 2020

Top 5 Considerations for IoT Sensors in Farming

Sensor-based data collection in farming using data collection sensors (Referred to as IoT or “Internet of Things” devices) is growing in all types of crops, as more and more AgTech companies release products to track everything from irrigation levels to crop vigor. In previous posts, I’ve talked about the importance of tracking data to make better decisions on the farm.[LINK] Like with any new technology though, it’s sometimes hard to evaluate its effectiveness. 

  • Did that thing I buy really help the bottom line, or was it something else I did? 
  • How much better was it than the one that cost half as much? 

Along with that, determining what to consider up front before fully committing capital to implement and scale the technology can also be tricky.

Based on working with several sensor vendors over the past several years, I’ve come to view every new AgTech solution through a set of five criteria, or considerations.  

1. Concentration & Cost

When thinking about rolling out sensor-based systems across an orchard or field, it’s important to consider what the ideal level of concentration would be and the cost that goes with that. If a weather station is being sold as the means to track data for micro-climates, how many weather stations do I need per acre? If the real benefit comes from one station per acre or block for example, what will the benefit be if I roll the solutions out over time? Will I get value from just one sensor to start? Or is there a minimal number of devices I need to get any kind of meaningful output out of the devices? It’s important to start slow with any new approach towards farming, but knowing the minimal required concentration to be successful, along with tracking the return on investment over time, is important in justifying the upfront time and money. If it takes ten devices to start, how long till I see results that could justify buying the next ten? And what does that benefit really look like at the end of the day? 

2. Connectivity 

Most sensors require either wifi or a cellular signal to transmit data, so it’s important you have the necessary connectivity across the entire area you plan to deploy sensors, both your initial and eventual planned area. It’s important to know upfront if you need additional fixed wireless towers, a mesh wifi network, or sign up for a cellular plan on top of purchasing and deploying the sensors. 

3. Utilization of Data

So you have installed the sensors, and now they’re connected and tracking information. What do you plan on using the data for? Is it mildly interesting information, or something that will really drive impact to your yield? Some people will track steps or calories with a FitBit, but interest will drop off after 4 weeks. It’s important to make sure your implementation won’t suffer the same fate. Thinking through the use cases for what to use the data for is important to maximize the value of what you’re deploying. It’s also important to discuss with the vendor how that data can be made available. 

Oftentimes, there will be a web-based interface with fixed reporting provided, depending on the sensors. However, discussing means to collect the data (ie connecting to the data in an application, and extracting it) is important if you want to perform your own analysis on the data outside of the fixed set of metrics provided. 

Furthermore, combining that data with other types of data you’re collecting is useful to determine the holistic state of a given acre vs having to bounce between multiple systems to get a full picture of what’s going on. 

4. Scalability 

I touched on this earlier in #1, but you want to get a clear idea of what it would take to scale the sensor system across your entire orchard in terms of cost, connectivity, etc. If you find your initial sensors are useful, make sure you have a clear idea on the roadblocks in front of you, if any, to continue buying and installing more sensors across your farm. 

5. Diversification 

Your final consideration involves identifying what different types of sensors you’ll need to capture a full picture of what’s happening. If you are just investing in soil moisture for the time being, will you eventually need a weather station nearby to track evapotranspiration? If you are using sensors to collect / track bug counts, will you also need to track plant vigor & growth to determine the impact of pest infestations? 

It’s certainly not required to track everything at once, but having a good idea on the types of things you want to track will help you assess what sensor combinations would make the most sense. It will also help you understand what vendors either provide multiple sensors or work together seamlessly to help give you more of a complete picture. 

If you’re only tracking one thing today, are you going with an option that’ll easily let you expand into other things later on? Or will you be stuck using / supporting several different point solutions and trying to make sense of the data yourself across all these platforms? APIs become an important ingredient to extract and combine data for a given acre. 

Though there may be additional things to consider when rolling out sensor-based solutions in your farm, these five points will help you get a clear idea of what you’re getting into and how to maximize the value of your investment. 

If you have any questions about sensors, or IoT, you can leave a message for me under the CONTACT ME above and I’d be happy to discuss your specific scenario in greater detail.

The 4 Parts to a Data-Driven Farm

Data collection, and the tools and services that allow it, provide farmers with possibly the greatest technology benefit across a farming operation. Like calorie counting can help folks lose weight by simply gaining better visibility over what they are eating, data collection can reveal where resources such as labor and capital are being spent to better optimize. This allows for cost savings and greater yields as growers move closer to precision agriculture.

Regardless of the type of crop, farming inherently has always been a data-driven operation. The methods used to collect and analyze data, however, have been changing at a rapid pace. As a result, growers have started to go from clipboards and Excel spreadsheets to ‘Internet of Things’ data collection and business intelligence. Row crops have certainly led the way on automated data collection, given the large number of acres per farm compared to other types of crops. Specialty crops, however, have been catching up in the past couple of years with new AgTech companies forming that focus on crops such as wine grapes and tree fruit.

Where To Start

Before jumping into the mess of AgTech vendors though, it’s important to identify the four pieces you’re going to need to modernize data collection and analysis across your agriculture operations. This will help ensure you have a strong foundation from which to build from.

My advice is to start with data analysis as the first step (even though it’s listed at the end of this post) in order to understand what you want to measure. It’s easy to incur a lot of costs quickly if you bite off too much at once when it comes to building up a modern day data ecosystem. Having a clear set of metrics will allow you to tie what you’re measuring to what you’re saving / gaining and get to a return on investment faster. From there, you can add to what you’ve built, and measure more as your use of data grows across the operation.

Part 1) Data Collection

Recently there has been a surge in sensors for farms that can track everything from soil moisture to pest counts. There’s two things to focus on when considering what tools to use for data collection. 

  1. How many sensors and what type are needed for collection.
  2. How “friendly” the sensors are to feeding data to the system you’ve chosen to consolidate and analyze it in. 

Most vendors offer web-based solutions for showing you their data, but it’s only really useful if you’re able to extract it into a central place (such as a data warehouse) because the real value of this data isn’t what’s happening to an acre of crops by itself, but what collectively is happening within that acre. This is important because you will need to pull together individual readings over a given acre to then compare to an output metric such as yield. As for concentration, without sensors the data is likely already being collected today using manual sampling methods. It’s important, then, to consider the ROI from automated data collection at a sufficient concentration. And of course, always start with just a couple sensors to validate the technology will work for you before scaling. If you only have 1 type of sensor though, to track something specific, then the vendor provided solution may end up working fine.

Part 2) Data Consolidation (or ELT)

Once you have your data collection sorted out, and you’ve begun to measure data in an automated way, you’re going to need to get the data into a central spot, such as a data warehouse, in order to put that data to best use. In that case, you’re probably going to need to pick a vendor to help with this. Though there are ways to do it yourself, it requires hiring one or more experienced data engineers / programmers that can build and maintain the system.

For most farming operations, picking an ELT (extract-load-transform) vendor is going to be a better bet, such as FiveTran or Dell Boomi. These tools will allow you to connect to a data source, such as your accounting system or field data collection, and extract / load / transform the data into a central place which then allows you access to all of your raw data across as many systems as you’re able to connect to. This is why, when considering data collection, you need to understand what vendors have capabilities such as an API which is a capability of a software platform that allows customers to connect to the platform and extract (and in some cases write) data.

Part 3) Data Warehousing

Having a central location for your data is the next foundational component. Depending on the volume of data, you may be able to use a standard database such as MySQL or SQL Server. If the volume of data you’re receiving is large enough (500 GB or more to start with) you may need to consider a Columnar based database such as Snowflake or Azure Synapse which is designed for large data sets. When you’re storing a lot of data from sensors across your operation, you are likely going to get larger data volumes sooner than later. The amount of data a sensor puts out varies quite a bit, so be sure and discuss with your vendor before signing up for the sensors. The key though, is to start with smaller data sets that are useful right away vs starting big by storing everything and having to support too much at once. You will be paying for storage monthly, so having too much unused data will just drive up the cost vs data you can put to work to justify what you’re investing (keep in mind though, storage is relatively cheap in the cloud so it takes a lot of data to drive up the cost of storage). 

It’s also recommended to host your data in the cloud, and to get as much of your ecosystem into the cloud as possible. Though you likely have data hosted locally today, which in some cases is required for business continuity, having a plan to get as much into the cloud as possible has a number of benefits. These include automated back-ups, easy scalability (you pay for what you use, and can increase the size of your server as needed), and streamlined pricing along with the convenience of having the services managed by the cloud provider vs handling onsite servers yourself among several other benefits.

Part 4) Data Analysis

Once you are measuring the right information, and then extracting / storing it in a central location, you need to consider what metrics you are going to track. Having ten or so metrics from which to measure your biggest cost drivers (such as labor or parts) means you can get actionable insights faster, versus guessing what might be useful. It will also help you understand what data to collect first, instead of boiling the ocean and gathering all the data at once. Using a business intelligence tool such as Tableau or PowerBI will allow you to point to your data warehouse and begin to construct these metrics to begin building automated reporting.

It’s recommended to work with an experienced IT resource that has worked with these tools in the past to get your reporting up and running, although if you’re technically inclined and on a budget, there are great BI tutorials available online. Once you have these metrics established, you can look at other capabilities under categories such as data science, artificial intelligence, or machine learning. However, that will require specialized expertise to build solutions using these advanced methods. The good news is these capabilities have been growing over the years, so it’s a lot easier to find expertise with these skills than it used to be a couple years ago. And the tools to roll your own advanced analytics system are getting better every day. 

Next Steps

Though this is a very simplified view of a data ecosystem for your farming operation, I hope this provides you with the 30,000 foot view of the pieces essential to a modern data-driven architecture. Understanding where AgTech vendors fit is critical to ensure you don’t have a bunch of isolated (or “silo’ed”) point solutions that you aren’t able to bring together into a single set of metrics.

The next step is to learn more about each, which I’ll be covering in future blog entries. Of course, if you have any questions, feel free to reach out directly or leave a comment.

© 2020 DanMaycock.com

Theme by Anders NorenUp ↑