That’s right, you heard me. And this was not a one-time fluke. The highly precise and robust algorithms of machine learning analytics can successfully identify lapsed individuals that have not given in decades, repeatedly! I am going to go one better and tell you that this big data driven, super-computing powerhouse can also accurately identify and assist in targeting non-donors. Most non-profits have them. This is a mysterious list of folks that have no transactional or promotional history. We typically hear similar anecdotes, “I have a file of names and addresses that dates to the beginning of our organization. We have no idea where they came from, and they have never contributed to our cause. Anything you can do with these records?”
Without blinking an eye, the answer is unabashedly “YES.”
We’ve been taught the benefits of reactivating lapsed donors.
- Lapsed donors are great warm prospects
- The long-term value of lapsed donors is significantly higher than newly acquired
- Cost to reactivate is far less than acquiring
- Lapsed donors can be a lead generation for monthly or planned giving
While these are true statements, it’s not so easy to re-engage with this plentiful, yet elusive audience.
There are a multitude of reasons why donors lapse.
And it all begins with the very first touch.
- Are first time donors thanked in a timely, personal manner?
- Do you have a welcome kit that resonates with new donors?
- Are donors receiving too much mail, not enough?
- Have you developed a meaningful, personal relationship with the donor?
- Have you distinguished yourself completely from the “competition”?
Now that we have established that your organization has checked the boxes when it comes to communicating with your newly acquired donor, you are still faced with the lapsed dilemma.
How do you reactivate lapsed donors?
If you Google how to reactivate lapsed donors, hundreds if not thousands of articles will pop up giving you tips, tidbits, and tricks for many tactics that can be employed to successfully re-engage.
We know them, we’ve used them, we think we have re-invented them and sometimes we are successful — but not anywhere near the success, had we relied on technology that runs across high performance networks in Peta scales… More on that later…
We’ve tried everything written in the books! What’s left?
No matter the strategy or expense, non-profits will still have lapsed names available to reactivate. The good news is that there’s a better way to target lapsed records that have the greatest propensity to re-engage… machine learning!
The History of Machine Learning
We believe that machine learning is an innate human concept: the concept of using abstract belief structures in order to simulate future outcomes. The early Egyptians used this logic to optimize crop planting cycles using binary representations. Today, we have scaled this up into the many trillions or greater pattern representations per second using algorithms, many times modeled after knowledge of ourselves or the emergence in the world around us.
Conceptual Building Blocks of Machine Learning
- Machine learning is based on algorithms that learn from data without relying on rules-based programming
- It’s a scientific discipline made possible through advances in digitization, enabling data scientists to stop building models and instead train computers to do so
- The volume and complexity of the big data available is unmanageable and virtually worthless using traditional modeling and data mining techniques, highlighting the need for powerful machine learning tools
- Because machine learning’s emergence as a mainstream management tool is recent, it often raises questions. However, the competitive significance of business models turbocharged by machine learning will increase as these tools become more widely adopted
- Current machine learning techniques developed in response to modern “data surge” — i.e., the amount of information that we “know” as a species doubles every two years
Raw computational power is growing at an even faster rate, and machine learning is unconstrained by the preset assumptions of statistics, a primarily descriptive tool. Business management and operations (the human element), however, have not caught up with the enormous increase in machine learning capabilities.
Machine learning can incorporate “visual recognition” technology as a component of predictive algorithms, effectively converting raw data into multi-dimensional pictures (not 3-D, but rather >2.08 x 10^90 dimensions which is incomprehensible to humans). Visual recognition technology can process an unlimited amount of data and any conceivable combination of variables.
Machine learning applications analyze and predict the optimal combination of touch points based on many patterns at scale with very high precision, e.g., geography, time of day, demographic factors, credit quality, etc. This is a computationally effective process, often involves many trillions of calculations per second (TFLOPS/ranges or greater), and is controlled by distributed frameworks which manage data flows, formulate algorithms, and provide consistent statistically valid output.
Machine learning applications get “smarter” with more data/patterns and successive, automated searches (re-training) of their solution spaces, meaning predictive capability continuously improves over time.
Leveraging as much value from your data as possible, machine learning algorithms account for non-responders and responder information.
Machine learning analytics can be segmented into three stages:
Where do we go from here?
The goal was to get you excited about the advances in machine learning analytic techniques and how you can easily employ these complex methodologies to re-acquire those sought-after lapsed names. Lapsed models built using the principles of machine learning will beat every method of reactivation every single time. Unlike traditional modeling methods, machine learning gets smarter with more data: more mail files, more responder, and non-responder information. Machine-learning models will never stop working. Models are continually updated in real time using new solution spaces and automatically retrained with new vectors.
A few real-life machine learning examples:
| CAPTURING THOSE TWENTYSOMETHINGS
After decades of failed attempts to re-capture donors that had given 5, 10, and even 20 years ago, machine learning methods identified over 25,000 prospects from a pool of 170,000. Income to date from mailing to this group of high scoring long lapsed is >$250,000 net revenue in 18 months.
| TURNING WATER INTO WINE
Those non-donors we talked about earlier—1.60% response rate compared to .53% response rate for cold acquisition lists. The non-donor segment achieved positive net for not only this campaign, but are continually doing so with each acquisition effort.
| TOO MANY COOKS IN THE KITCHEN
A mature, national health organization was using several models concurrently to reactivate lapsed donors. The machine learning methodology beat initial projections in both response rate and average gift and replaced all other methods for selecting lapsed donors. Machine learning models allowed this nonprofit to successfully mail deeper into its lapsed pool thereby reducing total campaign costs.
| WORTH THE PRICE TAG?
Not all are firm believers in this relatively new field of machine learning. Here’s where a head to head proved the proof is in the pudding. A bake off between those cute, catch phrase segments and machine learning algorithms resulted in significant lifts in every segment: many upwards of 50%. Segmentation combined with giving history is no match when it comes to looking at data in 50,000,000 dimensions.
These are just a few of the many examples of how machine learning tools can surpass traditional methods of RFM, demographics, giving propensity, and basic regression models. Most marketers in the commercial space have been approaching reactivation through this network of machine algorithms and making sense of big data.
Competition in the non-profit space has increased exponentially. Since 2000, the number of public charities is up 50%. Fundraising costs continue to climb. Universes are shrinking. Despite your best efforts, retaining online donors and offline donors alike is difficult, and donors will lapse.
Look to new technology and methodologies to effectively re-engage and re-activate. Machine learning is by far the most powerful tool available today.