Machine learning for nonprofits and why it’s important
What is machine learning and how does it impact your nonprofit?
What comes to mind when I say “machine learning”?
Some people think about numbers. Some think about job security. Some think about this guy:
Whatever the thought was, undoubtedly, something crossed your mind when we brought up the term. So we thought it would be a good opportunity to dive into what machine learning actually is — and how it’s going to be a game changer for the non-profit sector (especially the ones who use Keela!)
At last month’s Capacity Conference, Keela’s Nejeed Kassam and Lee Sutton broke down everything the sector needs to know about machine learning.
There are three levels of data literacy that are useful to data scientists:
- Hindsight data
- Diagnostic data
- Predictive data
We will explain each of these types of data, and why they are important to the non-profit sector.
What is hindsight data?
Hindsight data answers the question, “what happened?”
It takes a look backwards and helps to give context to a situation. Examples of hindsight data for the non-profit sector include:
- How many donations did I get this year?
- Which eblast that I sent was more effective?
This is the first level of data literacy. It is one of the easier forms of data to interpret, because it already happened and the numbers usually tell a clear story.
What is diagnostic data?
Diagnostic data answers the question, “why did this happen?”
It takes a look at a situation, and tries to offer explanation using the data collected. Examples of hindsight data for the non-profit sector include:
- Donations decreased because the ask was not made at an ideal time of the year.
- Your eblast was not successful because the message was not properly targeted to the audience.
This is the second level of data literacy. There are tons of variables that go into explaining why a certain outcome happens. But diagnostic data helps to give some of those explanations a bit more of a foundation. When you use and understand diagnostic data, you are able to understand your roadblocks better — helping you make more strategic decisions in the future.
What is predictive data?
Predictive data answers the question, “what could happen?”
It takes a look into the future and predicts possible outcomes based on what has happened with the data in the past. Examples of predictive data in the non-profit sector include:
- Predicting the best time of year to ask for donations, based on when most donations have come in, looking at historical data.
- Predicting what tasks or programs to suggest to an individual based on what they have participated in, looking at historical data.
This is the third (and most exciting) level of data literacy. This is also where the majority of machine learning concepts come into play. This is where computers and software can make predictions to help you make strategic decisions for your organization, based on data.
What is machine learning?
The best definition of machine learning is that it is an algorithm that predicts an output based on an input (or multiple inputs), whose performance improves with more data. Or in other words, a program that helps you predict a result based off of the data it has received in the past. These predictions get better and better, the more data it has.
You can immediately see why this would be interesting in the non-profit sector. As fundraisers, we are constantly trying to build stronger relationships with individuals, and eliminate the barriers we face to make that happen.
But often, we can’t see those barriers because there are so many factors to consider all at once:
- Time of donation
- Size of donation
- Whether or not a donor participated in a program or initiative in the past
- If that donor is subscribed to our newsletter
- Where that donor lives
- How old that donor is
These are just a few variables that contribute to whether or not an individual will give — and conversely, these are variables that can contribute to various barriers that could potentially exist to building a stronger relationship and soliciting a gift in the future.
But do you see the problem?
It’s quite easy to visualize and analyze one, two or even three of those pieces of data at once. But what happens when there are more strands of data that we have to interpret simultaneously? It gets confusing, and we get lost.
Machine learning is an algorithm that predicts ant output based on an input (or multiple inputs), whose performance improves with more data.
This is the big problem that a lot of people have with data. It’s useful, but it quickly gets complicated. Your team does not have data scientists on the payroll — so what are we supposed to do?
Our answer is simple: let the robots do it.
Of course, this is a bit of a tongue-in-cheek response, but the sentiment is basically true. With the advance of machine learning into the non-profit sector, you are able to lean more on software to help you make predictions and steer fundraising strategies.
It’s important to note that machine learning in the non-profit space is not meant to replace workers or their expertise and relationships. Machine learning and predictive modelling is intended to be an aid that compliments your own expertise. Think of it like your own personal fundraising consultant: They will make suggestions based on past results — and you have the ultimate decision of whether or not to follow those suggestions.
At the end of the day, you know your donors best. So look at machine learning and predictive modelling as the tool that it is, and use it effectively.
******
READ MORE >>>
******
