Building an Experimental Enterprise

February 22nd, 2018 Comments off

Originally published on LinkedIn in December 2015.

Building an Experimental Enterprise

At Silicon Valley Data Science, we believe in building an Experimental Enterprise that is capable of fantastic learning and growth. In my work with our team of data experts, as I interact with the senior management of our current and future clients, we discuss important considerations and traps to avoid as they become more sophisticated at using data to drive their businesses. We focus on business decisions and philosophy, rather than on individual technology choices. This post provides a window into some of the frequent topics of these discussions.

Questions to Ask

If you are interested in building your own Experimental Enterprise, then consider the following questions. If you answer “no” to any of them, then that’s an area you’ll need to work on.

Do I avoid making gut decisions?

In facing any business decision, your first instinct should be to look for relevant data in your business systems to help you evaluate the options. Take a look at past decisions that you may have made on instinct alone, and examine the data that was available at the time. How might that data have given you insight into the eventual outcome? Use this research to understand how you can better inform yourself in the future, so that you can make decisions based on information, rather than on luck. Especially when you’re considering complex or controversial business decisions, using data and a fact-based view to help discuss objections or obstacles will keep the deliberation from being derailed by competing anecdotes.

Do I truly value failure?

Paying lip service to fast failure doesn’t help. Ultimately, the reason that failure is quite hard on companies and individuals is that it challenges the established mindset that success is the only desired outcome. To test a hypothesis properly in business, you must have a course of action in mind for both success and failure. Plan ahead so that, if you don’t succeed, you know what you need to do next in order to try again, instead of throwing up your hands. As long as you have a good mindset in this way, then even if your experiment fails, you can still move forward to the next branch in the decision tree. You may have to revisit your assumptions, or even take a first-principles view on evaluating an opportunity, which means: don’t be constrained by the things everyone has established as constraints. But have a plan for failure and understand what you need to do next to make that failure a valuable learning opportunity.

Do I evaluate my own ideas?

Bubba Murarka notes that the ultimate skill of a good product manager is to have a high rate of idea validation, not just a high rate of execution or experimentation. Achieving this requires a mindset shift from aimless exploration (“What is the data telling me?”) to purposeful fact-finding (“What do I want to learn and how do I collect or generate the data to support my thinking?”). In addition to evaluating the outcomes from business experiments, you must also evaluate your specific hypotheses and what led you to create them. To have a high rate of idea validation, you have to question your questions, not just your data.

Pitfalls to Avoid

In becoming an Experimental Enterprise, there are numerous traps that may hinder your progress or force you into sub-optimal decisions. Watch out for the following.

Starting with a particular technology solution.

There are no magic tools or platforms that solve everything. We recommend starting with clear intent on a set of business objectives (what do you actually need to accomplish in order to move your business forward?). Then move with purpose to understand how the data you have now, or will have in the future, can assist you in accomplishing those objectives — and what tools you’ll need to wrangle that data. Your choices of technology and analytic platforms should be driven by current and future business needs, not by current fads or legacy constraints.

Looking for perfect answers (they don’t exist).

We live in a world of probabilistic data. Data elements are generated from so many sources and with so much inherent lack of quality that attempting to construct perfect views is a fool’s errand. We used to talk about “analysis paralysis” in the past, but it manifests even more strongly now that data is more ubiquitous. You must avoid becoming paralyzed while waiting for perfect data to arrive to validate any decision. Perfect answers don’t exist. Instead, use the data you have to make a more informed decision — as long as you’re also careful not to conveniently tailor the facts to your hypothesis.

Assuming that data alone will help you innovate your business.

Fifteen years ago, I used to ask clients to just give us their data in the hopes of finding analytic gold. Though this approach can be helpful at times, I’ve come to the conclusion that it’s usually not. Your data can’t provide innovation for your business on its own. It can provide insight into what is happening, or help you predict what might happen in the future, but data only leads to benefit if you have a plan. You must independently develop business intent and ask good questions. There will be times when it’s worth mining your data to inspire yourself, though: sometimes starting small (with an incremental increase in sales, improvements to margin, or a lift in customer engagement) will help you to understand bigger potential actions as well as where your data can help, before you start to dream even bigger.

In Conclusion

The world of data is changing. The skillsets of the people who use it are changing; consumer expectations are changing; even employee expectations are changing. In the conversations I have with current and potential clients, they recognize this: the recruits they’re talking to are eager to work with data in their decision-making processes. The fundamental relationship of people to data has changed. To truly take advantage of this fact — to thrive — your business must adapt.

An Experimental Enterprise is, fundamentally, an organization that thrives on change, and that uses data as a catalyst. Becoming an Experimental Enterprise means reshaping the way you and your company see things like failure, the role of technology, and your own gut instinct. But the benefits are potentially limitless.

Categories: Uncategorized Tags:

Big Data in the Boardroom

February 22nd, 2018 Comments off

Originally published in September 2014 at the blog and LinkedIn.

What your Board of Directors wants to know about Big Data.

I recently spoke about “Unlocking Business Opportunity from Big Data” to a group of former CEOs and senior business executives. Glen Matsumoto, Partner and head of the NY office at the private equity firm EQT, invited me to speak with EQT’s group of Industrial Advisors who serve as Board members for a variety of companies in the industrial and infrastructure industries.

I told this group of executives about our philosophies at Silicon Valley Data Science, including becoming a data-driven businesses, how to approach an emerging technology like Big Data by setting the right high-level agenda via a data strategy, understanding the building blocks of an Experimental Enterprise, and how they can be marshaled to create results. Finally, I shared some industry examples, highlighting successful initiatives at other companies that have created lasting business results from big data.

It was fascinating to engage in dialogue with this group during the Q&A portion, and to hear their questions about data science and big data—which largely reflected Board-level concerns around strategic growth.


The first question was: “Should we be concerned about big data and privacy?”

This has been a hotly-debated topic within the data community for years, but I was intrigued that this question was also top-of-mind for this group of senior business leaders. As I shared with them, we believe that businesses have to fundamentally center the conversation on trust. I encouraged the group to focus on creating trusted relationships with their employees and customers, building upon a foundation of clear and transparent communication about how data is being used by their businesses.

Big data technologies and the data science algorithms applied to that data do have the power to create incredibly detailed understanding of customers and businesses. By working to create a trusted relationship with customers, companies can change the conversation from one about fear and distrust of technology, to one about the benefits (or not) that a customer can expect from the use of their data.

Because most companies evolve their use of data over time, privacy policies and corporate communication must be clear about how data can be used, how the customer may (or may not) benefit, and the available mechanisms to opt-in or opt-out of data collection or analytics. Companies must also focus on more visibility for customers—no one is going to read through a 30-page privacy policy, so summaries of key concepts are a must.

Separately, we recommend that all of companies be prepared for a data breach. Investments in security and protection of data are clearly important and should be a high priority for any CIO. However, with large corporate data breaches highlighted in the news almost every week, privacy failures have come to feel inevitable. Having a clear strategy and plan in place for how to deal with a data breach is an important facet of building—and retaining—a trusted relationship with customers and business partners.

Business Value

The second principal question was: “How do we ensure that we get business value out of big data investments?”

I shared our view on how to prioritize business investment in big data by creating an effective data strategy that focuses on how to use data to enable your business, as opposed to figuring out what to do to data as the end point. I advised the group to make sure that business objectives were well understood, and to then understand what technologies could be used to support those objectives.

We highly recommend that technology organizations understand the generic patterns (i.e., technical workloads) of how technology can be applied to address specific business objectives and use cases. With this understanding, companies can take an agile and iterative approach to deriving business value from new technology investment—to ensure that the question is not whether there is benefit in investment in big data, but rather how quickly benefit can accrue to the business.

(My co-founder and CTO, John Akred, and our VP, Advisory Services, Scott Kurth, recently gave a seminar on our approach to data strategy. If you’re interested in learning more, you can sign up to be notified the next time they offer the seminar.)

Strategic Planning

There were several other things discussed, but the last question I will highlight was: “In what areas of strategic planning should we make sure to consider when looking at big data?”

We discussed five major areas in which data science and big data should be considered when doing strategic planning:

  1. Business expansion. Using larger data sets and better analytics to support market analysis for new business areas, whether taking new offerings to existing customers or opening new markets for new or existing products.
  2. Operational efficiency. Especially in the industrial sector represented by this audience, the internet-of-things combined with general improvements and cost reduction in sensors have led to greater data-oriented visibility into the operations of a business. Big data can be used to improve operational understanding and create new efficiencies to drive growth.
  3. Understanding customers. Most companies gather information about their customers only to leave that information in a CRM or ERP silo. Developing more comprehensive views of customers, and using data from across the enterprise, can lead to better decision-making.
  4. Improved marketing. Marketing efforts have already been changed dramatically by the evolution in data processing and analytics. Companies should continue to look for opportunities in this area.
  5. Data services. Finally, I advised the audience to think about the data they are generating internally, and to think about what value it may have to trusted business partners or other companies. Every data-driven business has the ability to syndicate their own data outwards for mutual benefit (and within the constraints of their own trust and privacy approach).

What about your board?

If you haven’t yet talked with your own Board of Directors about what you’re currently doing with data, now is the time to answer the questions that may be on their minds.

Start a conversation about trust, but also discuss how to communicate your privacy policy clearly, and what you’ll do if there’s a data breach. Make your data strategy explicit (and if you don’t have one yet, we’d love to help!). Explore the areas of strategic planning in which you may not yet be using your data to its fullest potential.

What other questions about big data have you gotten from your board?

Categories: Uncategorized Tags:

Tweets for week of 2018-01-28

February 4th, 2018 Comments off
Categories: Twitter Tags:

Tweets for week of 2018-01-21

January 28th, 2018 Comments off
Categories: Twitter Tags:

Tweets for week of 2017-11-19

November 26th, 2017 Comments off
Categories: Twitter Tags:

Tweets for week of 2017-11-12

November 19th, 2017 Comments off
Categories: Twitter Tags:

Tweets for week of 2017-10-29

November 5th, 2017 Comments off
Categories: Twitter Tags:

Tweets for week of 2017-10-22

October 29th, 2017 Comments off
  • RT @levie: Amazon's next innovation: Amazon House. You literally just live inside of their warehouse and get instant delivery. 2017-10-25
Categories: Twitter Tags:

Tweets for week of 2017-09-03

September 10th, 2017 Comments off
Categories: Twitter Tags:

Tweets for week of 2017-08-20

August 27th, 2017 Comments off
  • RT @SVDataScience: In this post, @tomeff dissects an online case study & highlights a common mistake even professional data scis make: http… 2017-08-24
Categories: Twitter Tags: