KISS the 288 View of Your Customer
Much has been written about the power of our massive data collections to enable the 360 view of our customers, our business, our employees, and our processes. When our numerous disparate heterogeneous data collections are aggregated and joined in our data lake or data cloud or data fabric or wherever, with appropriate data tagging, data discovery and data integration tools in place, then we can reach for that ideal: the 360 view of our domain!
But is the “360 view” really the right goal? It is definitely a good target and we should incentivize productive work toward that ambition, but should we go all the way to achieving that full 360 view in all projects, at all times? In some cases, yes we should, such as in critical application domains such as healthcare (360 view of the patient) and cybersecurity (360 view of our cyber environment), and in other critical systems involving human and operational safety and security.
Avoid analysis paralysis
Most of us have probably learned by now the truth in the statement “the perfect is the enemy of good enough.” Another way of expressing that is in the 80–20 “rule,” which represents the common experience that 80% of the value of any project is achieved in the first 20% of effort, while the final 20% of value is painstakingly squeezed out (if at all) with much more effort - with the other 80% of your project’s time and resources! Of course, this “rule” might actually look like 90–10 or 75–25 or 60–40 in any real situation, but the underlying principle is clear. Striving to achieve 100% perfection on a data analytics modeling project is often the wrong end goal. Trying to reach that goal may lead to a failure mode known as analysis paralysis.
Some of the mistakes that could be avoided when we are willing to accept a little less than perfection in our machine learning and data science models include: overfitting, bias, and missing the low-hanging fruit.
A majority of business analytics applications are focused on forecasting: predictive analytics. Of course, we want our predictions and forecasts to be as accurate as possible, since the impact of such models is felt all across the business. See this list for examples:
Predictive analytics is fueled by good training data. But good training data really means data that is rich in behavioral patterns and insights — it doesn’t mean “piles of data” (which we often try to build up in service to the “360 view” mantra). If you can tap into key patterns and actionable insights in the “right data”, and subsequently achieve excellent outcomes from that approach, then don’t equate “excellent” with “100% perfect”, especially if the extra cost associated with “perfect” is 4X the cost of “excellent”.
Keep it simple and smart!
Consequently, let’s focus on another common principle: the KISS principle! No, not the one you might be thinking of, but this one: “Keep it Simple and Smart!”
In this context, our application of the KISS principle means to focus intelligently (informed by data) on the first 80% of value in our project, which requires only 20% of the effort (time, money, resources) that we would otherwise have to spend in order to achieve perfection (= adding on the other 20% of value).
Therefore, I am here to advocate for a new business goal: The 288 View! Why 288? Because 288 is 80% of 360! Simple math, right?
288 vs 360
The rationale for this modification of the traditional goal of achieving a 360 view is not in trying to achieve that specific number (288), but in achieving a successful business outcome: KISS! In other words, focus your data analytics implementations on achieving your MVP (minimum viable product) using an iterative “fail-fast to learn fast” agile process. Focus on value first, and on perfection last. These approaches effectively and efficiently enable value-driven “data-to-action”: data productization, data operationalization, and data monetization. In this context, both the KISS principle and the “288 View” idea find a natural home.
A useful example application of the 288 View is in digital marketing analytics. If 80% of your personalization efforts (such as product recommendations and “Segment of One” hyper-personalization marketing campaigns) are successful (as verified through standard A/B testing and other key performance metrics), that would certainly validate the 80–20 approach to your marketing campaign research, design, and implementation investments. How many of our marketing campaigns already fall short of achieving 80% conversion, cross-sell, or upsell? So I say: spend less, and achieve more! That’s the “288 View” approach.
Digital marketing is a valuable example for another reason: we all have stakeholders. Well, any of us who work for a living has stakeholders. Those might be customers, investors, employees, our employer, board members, citizens, our local community, or the global community. In all cases, we want to deliver the right (and best) services and products at the right time to the right person. Thus, achieving success in those efforts both efficiently and effectively enables us to achieve even more successes. For example, we see recommender engines in places beyond e-commerce online stores, including e-government websites, which recommend services or information resources to site visitors who are already there to obtain some other assistance. This form of stakeholder engagement and CX (customer experience) enhancement goes a long way to building loyalty, trust, and communication with your other MVPs (most valuable partners).
288, KISS, and digital marketing
So, how do the 288 View and KISS principle look in practice in digital marketing analytics? When faced with marketing in the data-rich digital world, we are tempted to exploit every drop of insight from our data collections. We want our marketing campaigns to be data-driven and consequently as automated as possible. Striving for the “360 View” is one way to reach that data-driven automation nirvana. However, the challenges quickly become enormous when we consider the velocity and variety (as well as the volume) of incoming data. For example, we have data from multiple sources (internal and external to your organization’s servers), from multiple channels (call center records, web logs, purchase transactions, customer service logs, customer reviews and comments, #hashtags that call out your business products on social media), and in multiple dimensions (context, content, sentiment, time, and location).
When faced with these challenges, we can reach for the most appropriate data that enable us to achieve that 80% value proposition and to materialize “data-to-action,” simply and smartly. Even our data science models should follow this principle. To paraphrase Albert Einstein: “Things should be made as simple as possible, but not simpler.” That viewpoint is entirely consistent with this statement made by famous statistician George Box: “All models are wrong, but some are useful.”
The four stages of “data-to-action” are empowered by the 288 View and the Kiss principle. Those four “data-to-action” stages are the steps in the MIPS workflow (Measurement-Inference-Prediction-Steering).
They are:
- Measurement (data collection, aggregation, and blending).
- Inference (pattern discovery, machine learning, and hypothesis generation).
- Prediction (model building, deployment, and validation).
- Steering (moving people, products, and processes into the right place with the right action in the right context at the right time, as informed by the data).
Focusing on value (and not perfection) in each of these stages is a winning strategy.
Conclusion
In summary, our best goals are to become more outcome-driven, to be less perfection-driven, and to overcome analysis paralysis with a little KISS applied to the “288 View” of your domain. This is the simple and smart thing to do in the era of massive data, and it will become even more imperative as we enter the hyperconnected data era of the internet of things!
Follow me on Twitter at @KirkDBorne
Learn more about my freelance consulting / training business: Data Leadership Group LLC
See what we are doing at AI startup DataPrime.ai
