Creating a Roadmap for Data Science


Share on LinkedIn

Creating an Analytics Center of Excellence is demanding. Success certainly requires careful thought around resourcing and technology, but, to my mind, the most difficult decisions and the biggest challenges involve creating the right set of roles and the right processes for your CoE. In last week’s post, I laid out a set of principles for standing up a truly differentiating analytics capability:

  1. Business driven not exploratory analytics
  2. Investing in analytics as a portfolio
  3. Agile methods for analytics
  4. Creating operational strategies for every analysis
  5. Coupling analytics and reporting to drive data democratization
  6. Making collaboration a core capability not an add-on
  7. Creating virtuous cycles inside analytics

In that post, I described the first point as non-controversial, yet there’s a strong and fairly natural tendency to believe that analytics should be left to analysts. You stand-up the technology, you deliver clean-data, and you stand back and let the experts do their thing. After all, if you’re a 49er’s fan, you don’t want Jed York (the owner) calling the plays.

It’s not really that simple, though. In professional football, the goal of the team is pre-defined and unambiguous. That isn’t the case in business and analytics. I’ve sat in countless meetings with stakeholders and analysts at every level of seniority and tried to explore analytics goals and projects. It isn’t easy.

At the highest level, goals like improved profitability, growth, and customer satisfaction may be simple and unambiguous. But these goals don’t translate directly into analytics projects. There is no direct analytics project for “growth”. It’s when you take these goals down one or more levels that a more direct path to analysis begins to appear. Profitability, for example, can be improved by improving pricing strategies, by changing customer mix, by growth or by reducing costs. There are analysis projects targeted specifically to optimizing pricing strategies, but that isn’t the case in these other areas.

To tackle cost reduction, for example, you need to drill further down into the realms of supply chain, call-center, paperless, and a veritable host of operational areas. Nor is the relationship at this level typically one-to-one.

The Website is just one piece – sometimes large, sometimes modest – of enterprise operations. In our practice, we have somewhere between 10-20 commonly-deployed methodologies for tackling Website optimization and there must be hundreds of one-off optimization opportunities for most sites.

The implication is clear. There are, quite literally, many hundreds or thousands of analytic opportunities in the enterprise even if you limit your scope to problems around which there are clearly developed analytics methods and approaches.

Nor is it clear that the right strategy is always to focus on the top goal. I’ve seen many an organization where growth is the #1 strategic imperative but the best analytics opportunities are around cost reduction. It’s not that surprising. When everyone is focused on growth, costs aren’t carefully controlled and opportunities for analytics improvement generally abound. Conversely, in organizations that aren’t aggressively growing, static marketing programs often leave significant analytic opportunities for growth and market share improvement on the table.

There are host of factors to consider when choosing an analytics project: everything from the top organizational goals to the likelihood of success, from the opportunities for operationalization to the complexity of the analysis and to the short-term availability of the data. Who decides?

There isn’t one right person. If you look at factors listed above, some are clearly the domain of the analyst. Only the analyst or data scientist is likely to have a real understanding of the analytics opportunities that exist and the complexity or likelihood of success of the analytics projects to address those opportunities (though sometimes business stakeholders have strong intuitions about the likelihood of success that should be respected). On the other hand, the potential for operationalizing an analysis and the determination of business priorities are best evaluated by fairly senior business stakeholders.

Like so many significant enterprise problems, choosing analytics projects likely requires collaboration and iteration. Business stakeholders are responsible for setting the agenda and identifying particular problem areas or gaps in knowledge. Analysts must be able to guide stakeholders from high-level goals to a level of detail where analytic opportunities begin to emerge and then must be able to identify the methods available to tackle specific opportunities. From this buffet of analytics, the analyst needs to be able to explain the potential impacts, the difficulty, costs, and data requirements of each project. With these, business stakeholders need to consider both the potential for and the roadblocks to operationalizing the fruits of the analysis to actually change and improve the business. Finally, someone has to be willing and empowered to make difficult decisions about which projects to invest in.

Messy as it sounds, that’s how it should work.

What about data exploration? What about visualization? Shouldn’t analysts be left to explore the data to find patterns and opportunities that no one anticipated?

Not really.

Visualization is a potentially useful technique for data analysis when it’s problem directed. For certain types of problems, visual identification of patterns may actually be easier than identifying those patterns statistically. This shouldn’t be confused with unguided exploration. You’re still looking for something quite specific, you’re just using visualization techniques to reveal the patterns.

It’s no different with machine learning techniques. It’s possible to let your machine look at the relationship between thousands of variables, but you still have to a problem in mind when you choose the data and the type of relationships you’re searching for.

From my perspective, unguided exploration of data simply isn’t a plausible method. I’ve never seen it work and I have no expectation of ever seeing it work. There just aren’t many serious activities where purposeful intent doesn’t trump random activity.

Your analysts should be able – at any point in time – to explain what business problem they are trying to solve, how the analytics methods they are using are intended to solve that problem, and why they’ve chosen both the problem and the method.

That just makes sense.

Next week, I’ll talk about a closely related topic – why the methods for selecting analytics projects should include a focus on building a portfolio of analytics projects and what the portfolio concept adds to your Center of Excellence.

Happy Thanksgiving!!

Republished with author's permission from original post.

Gary Angel
Gary is the CEO of Digital Mortar. DM is the leading platform for in-store customer journey analytics. It provides near real-time reporting and analysis of how stores performed including full in-store funnel analysis, segmented customer journey analysis, staff evaluation and optimization, and compliance reporting. Prior to founding Digital Mortar, Gary led Ernst & Young's Digital Analytics practice. His previous company, Semphonic, was acquired by EY in 2013.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here