What if you could predict which of your donors is a prime candidate for increasing the amount they’re giving your organization?
Moreover, what if you could find answers to questions you didn’t even know to ask?
There are patterns present in your donor data that you’re not even aware of. They may be invisible to the naked eye, but machine learning can spot them at lightning speed.
Machine learning is a vast, hard to understand idea. But you don’t need to understand all the ins and outs of machine learning to be able to understand the use cases and begin to apply machine learning concepts.
With a couple of examples from real-world applications of machine learning, and ruminations on how you could incorporate techniques or ideas informed by machine learning in your NPO, this guide will ease you into machine learning.
So What is Machine Learning?
Techmergence defines machine learning like this:
“Machine Learning is the science of getting computers to learn and act like humans do, and improve their learning over time in autonomous fashion, by feeding them data and information in the form of observations and real-world interactions.”
Most of the time humans need to “program” a computer to tell it what to do. On a simple level, this means writing computer code in a language the computer understands. This code tells the computer how to respond when a human interacts with the computer program.
Take Amazon’s recommendation engine as an example. There are two ways this could have been built:
The non-machine learning way…
Developers would program the website to make product recommendations. For example their developers would code, “For all females who recently purchase shampoo show other hair care products.”
As you can imagine, this is very time consuming. The developers would end up writing millions of lines of code to cover all the potential different use cases and furthermore, it probably wouldn’t be that accurate.
If a buyer purchased shampoo are they really likely to also buy other hair care products? Or is that just what we think they would buy?
The machine learning way…
Rather than programming the website, data scientists would provide historical information and ask the website to make future predictions based on the historical data.
Data scientists would analyze the purchase habits of a ‘training set’ of previous buyers and their purchase history, enabling them to identify correlations between certain groups of users and their purchasing habits thereby enabling them to make future predictions based off of this historical data.
For example, imagine their training data showed them that for all females buying shampoo who were over the age of 34, lived in the Northeast and were Prime members, 75% of them also bought a set of pillows. Using this method, when Amazon identified a user who met this criteria it would recommend a set of new pillows rather than other hair care products.
This method not only saves time but it also is more accurate because it is based on historical data. Over time as the model is given more data and the data scientists fine-tune it, it becomes smarter and smarter. In essence, this means humans aren’t telling machines what to find, they’re teaching them how to find whatever exists. And the machine learning part is that the machine gets better and better at finding whatever exists, with the help of human teachers.
To sum up, machine learning can be thought of as a tool to automate analysis we used to have to do ourselves. Given an amount of training data, humans train a computer to classify the kind of information in the data. After this training, the computer is able to classify new information based on the training data. Over time, you can continue to feed it new information which it will classify, becoming more precise and more accurate over time.
That’s basically it!
But What Does Machine Learning Look Like?
Here are some real-life examples of machine learning:
- Online recommendation functions, such as Hulu’s “top picks” section or Amazon’s purchase suggestions
- Fraud detection software used in credit monitoring settings
- Our technological friends Siri and Alexa
- Gmail’s recommended responses
Let’s turn to Netflix for a more in-depth example. Every time you watch an episode, or even part of an episode, Netflix is learning about you and what you like and don’t like. And Netflix has hundreds of millions of users, most of whom aren’t going to be anything like you. They’re going to live in different places, like gory horror movies over your preferred workplace sitcoms, and watch at different times, among other things.
The important thing though is that some of these users are going to be extremely similar to you–at least when it comes to Netflix! Netflix learns about you and everyone else who uses Netflix, and can use this information to tailor their recommendations to you based on what other people like you like.
Isn’t this similar to what we do in real-life? If a friend asks for a recommendation on what they should watch next, we’ll quickly assess what we know about how this person is similar to ourselves. Then we’ll recommend something we enjoyed specific to how we think this person is similar to us. So if you know you share a love for true crime, you’re not going to recommend the rom-com you watched last night. You’re going to recommend the amazing murder mystery you watched a couple weeks ago. This is what Netflix is doing but on a far more sophisticated scale. Netflix has often recommended shows to me that friends in real-life have already told me I’d love. And I did! This is machine learning at work.
But Netflix’s machine learning goes way beyond tailoring recommended shows to your specific preferences. Netflix recently published their work on personalizing the artwork each viewer sees for different titles, based (somewhat) on the reason Netflix is recommending that title. Furthermore, Netflix is using unbiased machine learning to accomplish this by starting the machine off cold, i.e. without training data, and allowing it to learn the best artwork for each viewer for each title all on its own. Imagine if you could go beyond your expert utilization of the identifiable victim effect and show each of your website visitors an image that would connect specifically with them. That future is here because of machine learning.
Why Does Machine Learning Matter for NPOs?
You don’t need to adopt machine learning right this second. In fact, you can wait until everyone else is doing it and then be in the last wave to catch on. But we think incorporating an understanding of machine learning into what you’re doing now is definitely do-able, and it’s costless aside from your time.
There are two big advantages to machine learning:
First, machine learning allows you to process huge amounts of data. Utilizing data is imperative for a successful nonprofit, but having a lot of it can be daunting. Machine learning can process large swaths of data rapidly which cuts down on both time and processing needs.
Second, machine learning can show you correlations and patterns that the human eye won’t catch. The donor segmentation NPOs do now will seem extremely rudimentary the further into the machine learning age we get. There are patterns in your donor database that you don’t even know to look for–that no human would know to look for! You need a machine on your side to carve out more specialized donor segments.
Sure, you may know that someone who gave in the last 6 months is more likely to give again than someone who never gave before. But you could also find out more sophisticated insight. Like that females with at least a high school education between the ages of 40-50 with 2 kids are twice as likely to give to you as a single female with a master’s degree. Specificity like this comes from machine learning and allows you to tailor your approach and use your resources more efficiently than you’ve ever done before.
By 2020, 57% of consumers will expect companies to anticipate their needs. The expectations of nonprofit donors won’t be far behind. This requires robust prediction capabilities that are impossible to create without the help of machine learning.
Luckily, now is a great time to introduce a machine learning mindset into your nonprofit organization. By focusing on data and how you can use it to your advantage, you are harnessing some of the potential machine learning brings to NPOs. If you’d like to do even more, get in touch with us so we can help!