Being data-driven is not enough — data-driven how, and why?
In the data strategy and data management world, the proverbial dot on the horizon towards which data leaders are attempting to travel, is that of a data- or insight-driven organization. When a company employs a “data-driven” approach, it means it makes strategic decisions based on data analysis and interpretation, as opposed to based on intuition or set precedents.
This is not controversial. Deloitte, KPMG, EY, PwC, IBM, McKinsey, BCG, AWS, NewVantage Partners, and many other thought-leading organizations have a remarkably similar message. Becoming data-driven is the goal, which implies growth across various dimensions, including obtaining executive buy-in, nursing a data culture, growing the right talent and skills, deploying the right technologies, building ecosystems, treating data-as-a-product/asset, and driving digital transformation.
Data-driven how?
I am as big of a proponent of organizations becoming data-driven like any other data strategy aficionado. But with one key difference, namely that it is understood and articulated to what end organizations are becoming data-driven. Saying you want to become data-driven is not enough — it is equivalent to a business leader saying she wants to generate profits, or an Olympic team coach saying he aims to win. That much is obvious — but how? Focusing on what?
Depending on the organization and its unique objectives and set of constraints, becoming “data-driven” can mean very different things. Size matters, of course — if you’re a local bakery, you probably won’t need an in-house data platform. More importantly, specific business goals can drive where you may want to become data-driven.
Compare and contrast car manufacturing companies like Tesla, Rolls-Royce, and Tata. Tesla’s competitive advantages relies on absolute product leadership — they enjoy first-mover advantages through their innovations in electric batteries, autonomous driving, and revolutionary design. Tesla does not advertise and the options to customize are limited. Rolls-Royce cars are extremely exclusive, comfortable, and can be tailored to the individual owner’s taste. Where Tesla is product-centric, Rolls-Royce is customer-centric. Tata Motors, to take another extreme, produces the most affordable car in the world. Their key competitive advantage is that they know how to make a minimally viable car at the lowest possible price.
All three of these companies claim to be data-driven, yet obviously given their completely different strategic outlooks, how they do so is radically different.
Strategic profiles
It is helpful to keep a few strategic archetypes in mind (there are more, but let’s keep it to these 4 for the purpose of our discussion):
Product leadership
A company can differentiate itself by making products or services stand out from those of competitors by adding high perceived value. A company that excels in this area clearly states and shows that it offers ‘the best’ product or service on the market, through innovative products of the highest quality that keep up with trends.
If a company puts product innovation at the center, what could this mean for the desired data capabilities? Of course it depends on the ‘product’, but here a few examples that often hold up:
- Data scientists and product innovators need access to a safe and secure environment in which they can experiment and ‘fail fast.’ This should support both directed hypothesis-testing, as well as more general data exploration.
- A data platform can be used to design, execute, and learn from experiments, encouraging data scientists, product managers, data engineers, and business analysts to actively collaborate.
- Knowledge of and access to external data sources can fuel ideation and enable externally validated hypotheses, for example by incorporating third party data into product offerings.
- AI and data science capabilities can be used to create descriptive, diagnostic, predictive, and prescriptive models that can drive a deeper understanding (e.g., understand what constitutes a hazardous situation on the road) and recommend actions (e.g., make an evasive maneuver to avoid a collision).
- A business-oriented data catalog is critical especially in larger, complex organizations with a large number of data sources. It enables data scientists to find the data they need and interpret it correctly through the accompanying metadata.
Let’s consider Tesla:
We have expertise in developing technologies, systems and software to enable self-driving vehicles using primarily vision-based sensors. Our FSD Computer runs our neural networks in our vehicles, and we are also developing additional computer hardware to better enable the massive amounts of field data captured by our vehicles to continually train and improve these neural networks for real-world performance.
The use case of enabling autonomous driving is critical for Tesla. It’s beyond the scope of this article to unpack exactly what the scientific process is that is being followed, but suffice to say that from the above quote alone, at least three data capabilities appear critical: data integration bandwidth to gather IoT-generated data, mass storage for it to be kept, and data science applications to train and deploy the neural networks.
Even from the very beginning, Tesla installed devices on its cars that generated data, and it ensured to gather and store this data for further analysis. As data accumulated over the years (for example, over 3 billion recorded miles, of which 35 million completely autonomous), models were built to scan and understand the direct environment of driving cars, recognize hazardous conditions, and avoid collisions. With basic data integration capabilities in place, it also became possible to offer games and movies directly through the car’s console.
Although Elon Musk had always had visions of producing electric cars that can drive autonomously, a lot of the the incremental innovations that would follow later were not specifically contemplated in the beginning. They became possible because of the enormous amount of gathered, quality-controlled data. Data scientists and engineers with access to that mountain of clean, well-labeled data could then expose it to AI and analytics tools in a safe environment — fertile grounds for data-fueled innovation to thrive.
Customer-centricity
Organizations can adopt a guiding principle to put their customers at the center of everything. Any decision or prioritization should be informed by what is perceived to be in the interest of, or most appreciated by, current or prospective customers.
Several data capabilities can be important:
- A so-called Customer 360-view can be built through an MDM system and/or expanded data platform to have a complete and holistic understanding of the customer across all interactions. Some use the term Customer 720-view to emphasize an even richer data-driven view of the customer by integrating external data, for example from social media sources.
- Using the data from the 360/720-views, customers can be segmented across dimensions that drive preferences, needs, and other characteristics. A good customer segmentation drives action — the marketing, sales, or service approach can be tailored depending on the segment.
- With the explosion of data around us, hundreds of organizations have popped up that offer data that contain so-called ‘data signals.’ Advanced capabilities include a data catalog that continuously scans the world for external datasets and translates those signals into possible customer-related use cases.
- Customer satisfaction can be measured, tracked, and enhanced by explicitly measuring it. The net promotor score (“NPS”) is probably the most well-known example.
- Basic data integration and visualization capabilities in combination with an intelligent (rule-based or AI-driven) recommendation engine can be important to enhance customer service. When a customer calls into a call center, walks into a branch, or goes through a life event, the respective company representative is not helped by being flooded with a mountain of 360/720-view data — rather, this should be translated into tactical recommendations (for example, to ask if last week’s issue has been resolved, confirm a phone number, or recommend a new product).
Let’s take a look at Rolls-Royce, through the words of its CEO who is describing its key customers:
These uniquely conscientious and prosperous personalities are not interested in convention and conformity. They are successful precisely because they play by their own rules. And because such individuals have fruitfully leveraged their personal brand with excellent results, it makes sense that they would be exceptionally discerning about the products they purchase and the lifestyles they lead. Their Rolls-Royce is more than just a reflection of their personal tastes; it is a unique branded luxury capsule that contains their hopes, dreams and achievements.
It is for this reason that so many successful individuals are attracted to the extreme customer-centric approach of Rolls-Royce, an approach that I have termed, Customer Hyper-centricity.
Rolls-Royce uses data- and insight-driven capabilities in various ways to optimize how they serve customers. For example, before designing its new Ghost model, Rolls-Royce interviewed and studied its customers in depth. Across its target segments insights were collected, which in this case led to the perhaps surprising conclusion that existing cars were perceived as “too flashy.”
Telematics are used to minimize customer inconvenience in the case of maintenance or car troubles. When maintenance is required, all relevant data can be sent automatically to the right dealer, so that they can schedule the maintenance check. And if roadside assistance is needed, with the click of a button the location and relevant car data will be transmitted to an assistance team that can then rush out to help. This requires IoT sensors, connectivity, and bandwidth, but also underlying master data to match customers to dealers and to be able to contact them through preferred channels.
The most differentiating factor, however, is personalization. Where a Tesla at the moment of writing is available in 5 colors, at Rolls-Royce there are 44,000 options — and if that isn’t enough, a customer can bring in things like flowers, pets, or jewelry, and a custom color will be created. The company put together a “Bespoke Collective” that includes designers and engineers to enable extreme individualism so that customers can co-create the car. This requires an extremely careful data flow across the end-to-end supply chain. Inputs may be gathered from customers in myriad ways, are then fed into the design process, which further downstream will trigger ordering of customized parts.
Cost leadership
Reducing costs through data can be done in two principal ways. First, whatever data capabilities are to be built, the cost consideration can be given a higher priority. For example, maybe you don’t need that Informatica data catalog if a more basic (or internal) version can meet minimum requirements as well. In your cloud environment, you can specify stricter rules on when data can be deleted or pushed to colder storage. Rationalizing visualization and data science tools can help to minimize licensing costs and allow you to negotiate down the total cost.
But second, and more in line with the purpose of this article, data capabilities can be prioritized and built that facilitate data-driven decision-making to reduce costs. Just a few examples:
- In process mining, data science is applied to event data to understand how processes are executed and how to improve them. It is perhaps the most targeted use of data science to manage and reduce cost.
- Supply chain analytics can provide insights into the operation and performance of the supply network, to reduce waste, and to predict the right supply at the right time. Specifically, determining the most cost-efficient way to ship goods is a proven data-driven use case.
- Companies that procure goods and services from a large base of suppliers can use data and analytics to inform vendor selection and price-setting, which is particularly relevant for raw materials.
With this in mind, let’s take a look at Tata Motors:
Throughout 2021, we increased our capability as an agile, fully data-driven, digital business with the creation of InDigital, a key pillar of our Refocus transformation programme. Our 250 specialists focusing on analytics, data science, data engineering, and automation have already supported initiatives that have delivered a return of over £300 million value to our business this fiscal year.
In the same annual report, details are provided on Tata’s Refocus program. It contains 4 foundational pillars, of which one aims to use “data and technology to power the transformation”:
As expected, there is a heavy focus on controlling costs — one of the foundational pillars, “Responsible Spend”, is even dedicated to it, aiming to remodel Tata Motors’ “approach to spend and investment, updating our purchasing processes, improving cost and time saving.” Of the 6 operational pillars, 3 are directly driving lower costs through reduced warranty spend and reduction in vehicle costs, and the 3 remaining ones do so indirectly through faster delivery times and increased digitized capabilities to drive profitability.
Indeed, Tata Motors is using various data-driven approaches and capabilities specifically keeping in mind their cost leadership objectives. For example, they successfully onboarded >500k vehicles onto a connected vehicle platform, which the company indicated to be a step towards a common standard technology stack to deliver scalability and to transform internal operations. It is built as a modular platform, which can be scaled up to offer a range of solutions with third-party applications through APIs. It therefore enables secure and selective access to the platform and relevant data with its wider partner ecosystem, including dealers and suppliers. This, in turn, will help original equipment manufacturers to convert data and insights into lower costs.
Governance and compliance
Across the world, expectations from customers and regulators are hardening as it relates to how data is secured, protected, and used ethically. Compliance with various regulatory guidance is relevant across themes like data privacy, ESG guidelines, and industry-specific regulations such as those related to risk reporting and anti-money laundering.
In addition to responding to external expectations, organizations can also prioritize data governance for internal reasons, for example to ensure data availability and reliability, and to protect itself against operational risks.
A few capabilities that drive data governance and compliance:
- Formal data governance practices establish rules for everyone in the organization to follow and to monitor adherence to these rules. Components typically include data policies and standards, data (governance) forums, and a compliance process (possibly with 3 lines of defense).
- Data quality controls can be built to define standards and implement controls to ensure that the data meets these standards. Data can be measured against various quality dimensions, such as completeness and accuracy.
- Data lineage can be documented, tracking how data is moved and transformed across the organization and its business processes.
- Data classification enables the identification of sensitive data that can subsequently be protected accordingly.
There is an important perceived tradeoff between defensively-oriented measures that drive Governance and Compliance and the remaining flexibility for business users. If you lock down data in strictly specified formats and limit access, this restricts to what extent data scientists and others can use this data, for example for use cases related to product leadership and customer-centricity.
I don’t believe that any of our 3 case studies — Tesla, Rolls-Royce, and Tata — are primarily focused on governance and compliance. That doesn’t mean they don’t pay any attention to it — in fact, they all do. All three are taking measures to ensure they comply with environmental, privacy, and other legislation. They all have implemented internal controls to protect against operational risks.
Strategic profiles should not oversimplify
Although the strategic profiles outlined in this point of view paint a clear and distinct picture, we should not oversimplify them — it does not mean that blinders are put on as it regards any secondary objectives.
Indeed, as argued above, all three car manufacturers have data capabilities in place to drive data governance and compliance with regulatory standards. For example, for Tata Motors it is noted that they engage their suppliers to exchange sustainability data to drive adherence to their Environmental and Social Requirements guide, which sets out expectations across ethics, environment, human rights, and working conditions.
In another example, although indeed secondary to their product leadership goal in electrical vehicles, Tesla indicates that controlling costs is important as well for long-term success:
We are particularly focused on vehicle cost during this period of macroeconomic uncertainty, high interest rates (thus higher cost of vehicle financing) and vehicle price deflation. We continue to focus on cost efficiencies while improving functionality and reliability. While cost efficient manufacturing of EVs is still rare across most of the industry, it is critical for profitability at scale and will ultimately determine long-term success of OEMs.
The takeaway
Returning to the initial observation that current data strategies often claim to be data-driven, I believe that data strategists should take some of their own medicine when they assert that data capabilities need to serve the business.
Organizational strategy should not be fed into the analysis somewhere down the line through a list of identified use cases — it is critical to spend time on this very early on. Challenge business leadership on how data capabilities can drive lasting competitive advantage and drive the hard vision-setting and decision-making. None of that is self-evident or can be lifted from a data management guidebook. Becoming data-driven is not enough — figure out exactly what data capabilities are critical, and why. You cannot have it all — if you direct resources towards product leadership, they cannot be used to reduce cost.
The stories of Tesla, Rolls-Royce, and Tata Motors show you what the result could be.
References and Recommended Reading
- https://tesla-cdn.thron.com/static/SVCPTV_2022_Q4_Quarterly_Update_6UDS97.pdf?xseo=&response-content-disposition=inline%3Bfilename%3D%22b7871185-dd6a-4d79-9c3b-19b497227f2a.pdf%22
- https://www.sec.gov/Archives/edgar/data/1318605/000095017022000796/tsla-20211231.htm
- https://www.ibm.com/garage/method/practices/culture/failing-fast/
- https://www.bmwgroup.com/en/report/2021/downloads/BMW-Group-Report-2021-en.pdf
