Please (PLEASE!) Define Proprietary Algorithm
Several months ago, I was approached by a startup CEO who was offering to pair his company’s analytical tools with ours. The firm’s senior leadership team told me fantastic stories about the power of their company’s proprietary algorithms. I would never need to hire a new data scientist or statistician they assured me. They confidently threw around adjectives that suggested their models were powerful and unique and sophisticated – though I admit having to consult a dictionary after every discussion.
Over a period of several months, this prospective partner nearly convinced me that his company’sproprietary algorithms could solve all of life’s problems (presumably right up to and including world peace). Our senior leadership team was also on board. We had checked most of the journalist’s boxes – who, what, when, where, why – in our due diligence process.
With every conversation though, I became increasingly aware that they were deftly avoiding the question of “how the algorithms worked.” It was a simple question that we asked 100 times in 100 different ways. At that point, we weren’t even looking for a copy of their code. We just wanted to be educated on their approach to data science and exactly how they could help us with predictive sales analytics. Every time we asked this prospective partner how their analytical tools worked, they avoided a deeper discussion of their scientific approach and only spoke broadly of the power behind their proprietary algorithms.
While that discussion eventually wound down, it alerted me to the fact that this term – proprietary algorithm – is overused to the point that it has become utterly meaningless absent a deeper discussion of what it does and how it works.
Since that time, I have likely heard 50 pitches from tech companies looking to sell us something or partner with us in which this term proprietary algorithm was used. Learning from past mistakes, I stop every conversation dead in its tracks when I hear the phrase and politely request the complete definition of what it does and how it works. More often than not, I am met with a look of stunned confusion, followed by an “Um”-filled definition that includes the very phrase in question.
Ultimately, though, I have discovered that people use this term for three reasons:
- The quick and dirty overview: They are trying to respect my time and want to give me a high-level picture of their offering to gauge my level of interest and where my interest lies. These people can give a sensible response to my “how it works” question in two minutes or less. Typically, these people do understand the product and actually have something of value to present.
- A crutch: They don’t understand the product or what is powering the product. It is never clear to me if there is actually something of value or not, but a second conversation never happens.
- Avoiding the truth: There really isn’t anything there. These folks seem to use proprietary algorithm in place of “Trust me: It works.” If they actually show the offering, it is instantly clear that there is almost nothing special about the logic underlying their platform.
Pilytix happens to be a company that has complex algorithms at the core of our product. These algorithms also happen to be proprietary, but they aren’t magical. We aren’t a business built on crystal balls or tarot cards. Just solid mathematical models.
As a rapidly growing provider of predictive sales analytics tools, we too had fallen into the trap of boiling down thousands of lines of statistical code into this empty, two-word characterization. However, we have never welcomed a client or partner who didn’t demand an answer to the “how” question. I am quite certain that we never will.