Our Blog

Information, analyses and news for agribusiness go-to-market

    Leverage Data Science in Your Buyer Persona Research (Part 2 of 3)

    Posted by FMiD Team on Mar 12, 2020
    Leverage Data Science in Your Buyer Persona Research

    If you want to achieve one-to-one communications at scale, a buyer persona is one of the most powerful tools you can have at your disposal.

    A solid set of personas can help you with a host of challenges that all agribusinesses face:

    • Strategy and market share. A persona can help you incorporate behavioral and attitudinal requirements to the typical parameters of product, price, and geography. This gives you a clearer idea of your addressable market and, by extension, your share of that market.
    • Marketing communications. When you can distill your market into a handful of personas, you can effectively communicate with each segment as though they were a single person. This is a powerful tool when creating targeted digital marketing campaigns.
    • Sales outreach. The more you understand about an individual grower and how they think, the better able you’ll be to make decisions about how to approach them for sales conversations.

    With this many potential use cases, it makes sense for you to truly invest in your personas and prioritize buyer persona research.

    First, there’s the question of accuracy. An inaccurate persona can lead to a bad market definition, a marketing message that doesn’t land, or a sales conversation that takes a wrong turn. All of these can be counterproductive and even harmful.

    Then there’s the question of time. In order to leverage personas in your agribusiness marketing strategy, you need to correlate a persona to each of your growers -- both prospects and customers -- for accurate analysis and segmentation. When you’re dealing with a market of 300,000 growers, that’s just not possible to do one-by-one.

    That’s where data science and analysis come in.

    This post -- part two of our series on buyer personas -- will dive into the data science operation needed to achieve this level of detail, and how to conduct the best buyer persona research possible.

    (This post is part two of a series on a data-driven approach to buyer personas. Click here for part one: What is a Buyer Persona and How Do You Build One?)

     

    New call-to-action

     

    The Need for Data Science in Buyer Persona Research

    No matter the size of your agribusiness, you need data science to inform your strategic and tactical decisions. Without it, you’re basing decisions merely on assumptions.

    In an ever-shifting agriculture market, making assumptions is a risky proposition.

    No one predicted that massive floods would wipe out 20 million acres of cropland. Trade policies flipped the market prices on their head. The massive growth of cropland in Brazil and Argentina is driving commodity prices down -- and they likely won’t return to previous levels.

    Intuition works great when it’s the status quo. But in times of change, it could lead you in the wrong direction.

    No matter what changes, however, we know that farmers are going to adapt and evolve. They’ll improve the productivity of their operations. They’ll continue to meet the demands of feeding a growing population. They’re already embracing technology in ways that no one was talking about five years ago.

    As an agribusiness, you have to understand how farmers are making their decisions, and adapt your processes to fit their behaviors. This why data science is a critical part of any business process, but especially your buyer persona research.

    Our buyer persona model is designed to account for the things that you may not be considering. That’s why we don’t limit the demographics that we analyze in our predictive model. We don’t just go for the obvious. We look at all the information at our disposal to figure out from all angles for who your customers actually are.

    While no method is perfect 100% of the time, the more data and the less subjectivity you have, the closer you’re going to be to getting it right -- and gathering insights that are truly actionable.

    How the Predictive Model Works

    Predictive models require two datasets in order to function. There’s a series of known data -- commonly called training data -- that serve to train the model on how to make predictions. The second is unknown data, against which the model is tested.

    In our case, a small subset of growers -- usually from a customer file -- is used as the training data. We append as many demographics to that as possible to gain the fullest picture that we can -- crops, acres, consumer data, financial, real estate, etc.

    The predictive model looks at this training data and uses the demographic correlations to predict grower personas in the unknown dataset.

    Buyer persona research -- creating your training data

    After correlating personas to the growers in your training data, that’s when we start doing demographic targeting.

    We take that small group of growers in the training segment and attach every variable that we have in our database to that. We don’t just go with the variables that are obvious and intuitive, because there are always correlations that aren’t necessarily obvious. That’s we throw everything at the model, and let the machine learning crunch the data to tell the story.

    buyer persona research -- predicting other personas

    Another key part of building the model is only use the information for the small subset that you have for the entire universe of growers. If you don’t have the data on the broader audience, the model is going to be less effective and possibly inaccurate when it comes to making predictions.

    For instance, if we’re building your model based on customers and then using that customer data to make predictions on the market overall, we wouldn’t want to use sales data to help score the model, because you won’t have that information on the broader market.

    The model itself organizes and ranks all these demographics, so see which demographics are correlative indicators of a grower's likelihood to match to a given persona. The model then looks at the covariant, the multivariate regression codependent variables, and many more factors that are commonly used in regression and statistics.

    It's a complicated process, but the results are simple. We train the model to predict what a given grower's persona is going to be based on all the information we have. That way, you can be confident in the results.

    Analyzing & Evaluating the Results

    After building the model, it’s time to see where the rubber meets the road. Does the model accurately assign the right persona to the right grower?

    Run the data through the model and see how accurate it is. If you’re happy with the results, the model is ready. If not, then we need to go back and refine it.

    No predictive model is going to be 100% accurate. So the testing and evaluation phase is focused on getting the predictions as accurate as possible -- or as accurate as you are comfortable with. We've found most clients are comfortable with the 80-85% accuracy range.

    There are many ways to refine the model, from ranking certain demographics differently to combining personas that are too similar -- or creating new personas when there’s a wide variance.

    But with the right testing and adjusting, the model becomes excellent at predicting personas and then, you’re ready to use it to score any new prospects that come in. When you know their attitudinal and behavioral preferences, you can more effectively market and sell to them.

    The final post in this series will delve into what to do once you’ve built your personas and how to utilize them in marketing and sales. Until then, if you’d like to get started building your own persona model, click here to get in touch with our team.

    New call-to-action