REACHING MATURITY: ANALYTICS IS ONLY AS GOOD AS ITS DATA SIMON JAMES
There has never been a better time to be a data analyst. There has never been more data. There has never been more boardroom attention. There has never been more human interest in both our own data and that of other people. So, for all the talk of geeks inheriting the earth and Harvard Business School proclaiming data science to be the sexiest job on the planet, why is it that little physical evidence exists of brands mastering the application of data analysis to their business practices? Even Tesco, perhaps the brand most synonymous with data analytics, has not been immune to market forces driving down both its market share and stock price over the past five years. Creating competitive advantage through analytics is difficult both to achieve and sustain. Leadership is a critical issue. The person accountable for analytics within an organization often does not hold a decision-empowered position or is not connected enough to make an impact on business strategy. A second issue is one of adoption. Having the correct strategy and gaining broad-based adoption of it are two separate issues. A third issue is one of transformation. Few companies can communicate the process of how they turn data into information, information into insight, and insight into action.
Analytics is a means to an end Data analytics is a means to the desired outcome of positive, commercial impact. If the investment in analytics cannot be tied back to real business impact, then something is likely wrong. Interestingly, people expect data analytics to prove the return on investment on marketing activities, but few measure it on the data analytics themselves. For those students of calculus, you will appreciate that this is a second order differential problem. For everyone else, this explains why companies data analytics functions are often underinvested, understaffed, and misunderstood. A model for success It starts with consultancy. Before anybody starts adding up all the ones and zeros, or working out how to tag a mobile app with tracking code, SapientNitro works collaboratively with our clients to determine the scope and shape of analytics services that will succeed within their corporate culture and service their business needs. Broadly speaking, there are three target operating models that data analytics functions might form (see Figure 1). Each has its strengths and weaknesses, and it is often the culture, not strategy, of the business that dictates which solution is most appropriate. The three models are: Centralized: One big analytics group servicing the entire organization. Hub and Spoke: A center of excellence supported by champions embedded within each business unit. Decentralized: Each department or business line conducts analytics independently and in a selfsufficient manner. FIGURE01 CENTRALIZED PROS Single point of ownership Center of excellence Clarity of responsibility CONS Risks ghettoization Not embedded in business Bottleneck HUB & SPOKE PROS Hub provides leadership Spokes provide self-service option/ownership Flexible and scalable CONS Requires matrix management DECENTRALIZED PROS High degree of local ownership Embedded in business Hard to govern CONS No centralized learning Hard to act consistently
The centralized model benefits from having only one point of failure, easy knowledge sharing, and the most latitude for capacity planning. However, centralized teams risk creating a bottleneck and making prioritization of projects more political. Conversely, the decentralized model benefits from implanting knowledge at the point of use and giving individual business units full autonomy. However, this model hinders knowledge sharing and risks effort duplication. In the middle stands the hub and spoke model, which is usually the most popular, trading off the positives of the other options while minimizing the negatives. Having a clear target operating model with a widely communicated purpose is a key driver of success. Yet leaders in the field of analytics go further. They understand their investments in analytics, from direct labor costs and software to licensing and agency fees. They also know what commercial impact this work delivers and can, therefore, set their level of investment intelligently. Without this appreciation, data analytics is doomed to remain a cost center, missing out on the link to revenue growth. This explains why many client analytics departments are fundamentally understaffed, and why companies like us provide analytics services in such volume. LEARNING FROM THE LEADERS Todd Yellin, Vice President of Product Innovation at Netflix, was a bright spot at 2015 s SXSW Interactive Festival. He shared highlights of how Netflix leverages data to produce insights and drive product innovation. For example, Netflix design and algorithm teams have run more than one thousand A/B tests with tens of millions of users to continually improve the customer experience. They have had many successes and, even more importantly, many more failures. He talked about "mountain testing" an approach to testing major experience changes as one technique to counterbalance incremental testing and optimization. While incremental testing can get you to the top of your current mountain faster and faster, there are risks. Sometimes, the focus on optimization can inadvertently lead you to forget about, or ignore, any attempt to climb a higher peak in the range. 1 For example, Netflix spent many months designing and building a beautiful, colorful experience that their research said kids should love. But it couldn t be tested piecemeal they had to build the new experience and then test it across millions of visitors. In this case, the experience simply didn t perform. And, although they scrapped it, they were able to use the data to avoid a major investment in a complex redesign when they launched in Japan. There's an important lesson in this for us as marketers: Know when to test incrementally and when to look for new mountains to scale. Be bold, take risks, fail, and learn. Netflix created mountain testing to test major experience changes, not just incremental tweaks. #SXSW 1 IPA UK. "Netflix and Buzzfeed on Test and Learn. http://austin.ipa.co.uk/post/113699851588/netflix-and-buzzfeedon-test-and-learn.
The Analytics Value Chain The diversity of types of analytics from advertising effectiveness to experience optimization to business cases for digital transformation is significant, as is the breadth of techniques applied to solve a vast array of commercial problems. SapientNitro has developed a unifying framework of analytics that we refer to as The Analytics Value Chain (see Figure 2). In this value chain, there are five incremental steps to achieving the highest levels of analytics maturity: INSTRUMENTATION How to capture data from networks, interactions, and behavior. REPORTING How to organize data into information. ANALYSIS How to generate insights from data. OPTIMIZATION How to programmatically apply insights and fine-tune the process. ADVANCED ANALYTICS How to innovate and apply more advanced concepts (multi-variate statistics, machine learning, algorithmic work) to push analytics into new territories. Everyone wants to maximize the strategic and commercial value of their data. But we also understand that high-performing analytics is only possible with strong fundamentals in place because analytics is only as good as its data. When analytics falls short, it is rarely for reasons of technique or logic it is likely due to incomplete or erroneous data. Instrumentation: The foundation Instrumentation is fundamental to good data. Short-sighted instrumentation is the number one reason why companies fail at analytics it s a productivity killer. FIGURE02 The Analytics Value Chain The Analytics Value Chain emphasizes the fundamentals of instrumentation and reporting before optimization and advanced analytics. Analytical techniques are constantly evolving; this innovation helps brands move from lower value to higher value positions. But as advanced techniques become widespread, they are commoditized and the process starts again. CLIENT VALUE Innovation CONSULTING INSTRUMENTATION REPORTING ANALYSIS OPTIMIZATION ADVANCED ANALYTICS Commoditization
Poor instrumentation creates problems down the line, and wastes valuable analytics resources creating workarounds and post-hoc remedies. With the rise of the Internet of Things, the number of touchpoints that can be instrumented to capture data is growing exponentially. This expansion makes data capture more difficult and stresses the need to automate data capture wherever possible. Reporting: Too often a placebo Most client briefs end at the reporting stage. This often proves to be a critical mistake. The world is full of reports that no one reads. They can be like an industry placebo a sugar pill of numbers. The very fact that reports are being physically produced often gives the illusion that all is well. Sadly, as Churchill might have framed it, reporting is not the beginning of the end, but the end of the beginning. Once you have accurate and timely reports, the fun really starts. Everything to this point is preamble. For a number of retailers, daily reports are used to set priorities for the next twenty-four to forty-eight hours. For example, we generate a daily scrum report and align the team based on what happened the previous day. This information then feeds retail operational decisions when to discount or move stock, what trends to jump on, and the impact of weather-related shifts. The way these retailers operationally tweak their business on a daily basis can be a valuable model for many other verticals e.g., financial services, service companies, and automotive. Analysis: Real value creation begins here The analysis phase is where real value is created. Drilling into the information we have, falsifying hunches, and thoroughly following disciplined trains of thought in our analysis, we look for unstated problems and try to solve them. It s worth reiterating that if you don t get instrumentation correct, you have neither the time nor the data to do the analysis (see Figure 3). For some clients, the analysis phase can reveal signals coming from the data. For example, imagine a scenario where 3 percent of customers through their behaviors can signal that they were likely to switch to another provider. Once you understand those signals, the analysis phase can help the business make decisions about whether to take steps to retain the client. FIGURE03 The problem with (big) data Big data is the problem, not the solution, for most companies. In this rough sketch, the author notes that while the amount of data increases rapidly, the amount of insight from that data does not. 35 Zettabytes 2 Data Insight 3 2020 2 Approximate numbers. 3 I made this up but trust me, I m a professional.
Optimization: Programs of change Often, we formulate individual insights into programs of change. These programs represent closed-loop learning where hypotheses are tested and iterated upon, and improvements in performance are achieved through the accumulation of marginal gains. These optimization programs offer bottom-line impact on everything from advertising campaigns to digital experiences, and ultimately, commerce conversion. Finally, the optimum application of analytics is to see what isn t there. For one client, our analytics team identified more than 100 potential improvements to digital touchpoints, which resulted in over $46 million in additional, incremental revenue. Advanced analytics: Seeing what isn t there Finally, the optimum application of analytics is to see what isn t there. We apply our knowledge of bleedingedge technologies, techniques, and data to create solutions for problems that brands didn t yet realize they had. For example, our ability to make sense of unstructured data with speed and at scale is providing new insight into areas where data analytics had previously shone no light.
In one case, we helped a client identify spot pricing opportunities in the oil industry by identifying rigs, tankers, fields, and refineries along the predicted paths of hurricanes (see Figure 4). For a grocery retail client, we optimized the layout of their dark stores (stores that only take e-commerce orders) to minimize transaction times and pass those time savings onto consumers. These are examples of applying tried and trusted techniques to new sets of data in innovative ways. FIGURE04 Hurricane path prediction Advanced analytics combine new data sets with existing business problems to deliver innovative solutions. Using predictive models, real-time location data, and market information, we were able to identify spot pricing opportunities by predicting likely weather-related production changes. Realizing your potential The promise of data analytics is real. And data analytics as a driver of growth is here to stay. Bridging the gap between the promise and value realization is best achieved by a systematic approach. Linking effort to commercial outcome and implementing analytics across the entire value chain provide the best chances for success. But in order to achieve success, you need to set yourself up for it. This means not only ensuring that analytics is plugged into the highest levels of your organization, but also appropriating data-driven decision-making as a company-wide issue and adoption measurement as a key metric of success.
Simon James Vice President, Global Lead Performance Analytics, SapientNitro London sjames2@sapient.com Simon is the global lead for Performance Analytics. For the past 20 years, he has worked in marketing as a data analyst. His team is responsible for measuring and optimizing the effectiveness of our work. INSIGHTS WHERE TECHNOLOGY & STORY MEET The Insights publication features the marketing intelligence, trend forecasts, and innovative recommendations of boundary-breaking thought leaders. The SapientNitro Insights app brings that provocative collection now in its digital form to your on-the-go fingertips. Download the full report at sapientnitro.com/insights and, for additional interactive and related content, download the SapientNitro Insights app. SapientNitro, part of Publicis.Sapient, is a new breed of agency redefining storytelling for an always-on world. We re changing the way our clients engage today s connected consumers by uniquely creating integrated, immersive stories across brand communications, digital engagement, and omnichannel commerce. We call it our Storyscaping approach, where art and imagination meet the power and scale of systems thinking. SapientNitro s unique combination of creative, brand, and technology expertise results in one global team collaborating across disciplines, perspectives, and continents to create game-changing success for our Global 1000 clients, such as Chrysler, Citi, The Coca-Cola Company, Lufthansa, Target, and Vodafone, in thirty-one cities across The Americas, Europe, and Asia-Pacific. For more information, visit www.sapientnitro.com. SapientNitro and Storyscaping are registered service marks of Sapient Corporation. COPYRIGHT 2015 SAPIENT CORPORATION. ALL RIGHTS RESERVED.