Data is seen as the spearhead of digital transformation - the value creation stories that are centered around data are always easiest to drive - but the unlocking of value is challenging at best. Few best practices are emerging:
Harvesting Value from Data
Building an enterprise data platform with “truth” in data is key to unlocking its value across an enterprise. The before and after pictures of companies that started with fragmented data with ownership across distributed groups and moved to a single centralized enterprise data platform are strikingly clear. And the winning operating model that has emerged is “data at the center, analytics at the edge”.
Data products are becoming the best practice in terms of visibility and driving business outcomes. There are three main types of data products, depending on how the data is consumed and the level of transformation carried out on data - source data products, aggregated data products, and fit-for-purpose data products. The value is created by the network effect of all the data products in an organization, and together, they serve the purpose of making data from various contributors easily accessible. Data products can be horizontal or vertical - examples include simple product recommendation (aggregates data within various interactions), or semantic search and compliance management (aggregates across data sets). The best practice is the creation of data products that comes along with the formation of a data market and an intelligence factory – and consumption is pull-based in that usage triggers a charge back mechanism – so the build can be really outcome focused. Finally, data products are only as good as the context in which they are applied, so it is key to build integration into existing workflows and adoption into business processes.
Guidelines and frameworks are key to build in upfront - while many companies have access to rich data – the key is to work out what can really be used. A tried and tested approach starts with a comprehensive understanding and inventory of all data that is accessible and then reimaging and extrapolating value can be extracted either directly from that data or from insights generated from that data. To this, evolving regulations and company policy and digital ethics are applied to drive the outcome list of data products. Framework and structure also help drive business applicability of data products – many in the industry favor a four-layer approach in their architecture, starting with the data lake at the base layer, golden data sets as the next layer up, a set of data products that sit at the next layer up, and finally a layer of the catalog that can drive rapid identification and usage.
Data Quality is King
Data is no longer a technology play - the technology is in fact the easier piece, and data quality is increasingly the unsolved enterprise problem.
Many companies are well down their data platform journey with key decisions behind them on infrastructure and modern data stacks, but data cleansing continues to be “a journey - not a destination”. In reality, this is actually a two-part journey - first cleaning up the current data sets, and then designing the new data build – in order to get it first time right. This needs attention, and often failing operating processes can better highlight this problem - than the theory of the case (and report inaccuracies) – which have not proven compelling enough to fundamentally fix the quality problem.
Data is the food for AI, and the investment strategy to source data sets for AI is key. But data engineering (and the work around data management, data governance, data dictionaries, data ontologies and data quality) can be likened to data janitorial services – it can be very expensive - and often doesn't get the investment it needs to do it right. Best in class organizations have driven data quality by shifting all operating reviews into the operating tool only – and as a result, driving significant focus on fixing the data and what it is used for. How you organize for data is also important - great examples of scaled CDO organizations show that there is a data officer in each business reporting up to the CDO accountable for the data within each business unit.
And as the journey on data matures, vectorization and knowledge management become increasingly accelerated and drivers of significant outcome value. As an example of a resulting use case is the subsequent training of large language models - each model representing a defined sub segment of the market – and built by tuning a foundation model with this knowledge graph – resulting in a “synthetic human” of sorts. Collectively this is then used to drive accelerated focus group feedback results by reasoning and inferencing against these engines – and used to drive product designs, market strategies, and lead campaigns in a rapid time to market fashion.
Data rights and regulations is evolving as a key area to focus on – with both purposeful disinformation and unintentional misinformation equally concerning. Additionally, data can be inadvertently or maliciously poisoned and that can affect the AI models and resulting actions. Also, data sovereignty rules are both here and now, and fast changing - and existing hyper-scaler solutions need to be evaluated relative to sufficiency with sovereignty requirements of data storage – for instance, depending on classification, some data needs to not only be in the EU, but actually, in-country. Finally, as unstructured data now becomes discoverable with Generative AI solutions like CoPilot, the age of “security by obscurity” is over, and building a semantic layer for unstructured data within the corporation that allows for an overlay of classification based granular security rules will become the new norm.
One final consideration is lock-up risk, and needs to be thought through. As the data platform provider landscape on the infrastructure side becomes more concentrated and, in some ways, closer to monopolistic, how do you retain flexibility in a still-evolving landscape. Data-driven businesses need to balance the trade-offs between different technology providers and single data platforms - reducing complexity but mitigating risks sufficiently.
Culture is Key
A company-wide data-driven culture – that understands and sees data as a new asset class- within and across the business is key. Often there is a vacuum in leadership around data-driven cultures, and the best CIOs and CDOs play a very large and visible role in helping businesses understand how data can be used to achieve better outcomes. The role of the technology senior executive is increasingly to ask the questions the business doesn’t think to ask – and to encourage the business to become an active part of democratizing data.
A culture of data democratization is a culture of curiosity, accountability, ownership and change management – often clubbed together as “data literacy”. And accelerating data literacy programs across the corporation starts with new tools that are designed for business people instead of just data scientists. The second step is driving agile programs across the company that demonstrate the journey from ideation to visualization to outcomes for all to see. The final step is making data a first-class citizen - the best practice is driving mindsets top-down, not bottom-up - starting with the CEO’s office. All put together, companies that have a data driven culture are easy to spot since they have shifted away from an organizational dynamic that is focused on the hierarchy to one where people with data have a voice.
Creating data-driven businesses can and probably will cannibalize existing business, so this requires active change management and executive sponsorship. A critical success factor is culling ideas quickly and decisively and only developing and driving initiatives that will have an impact at scale within a reasonable timeframe.
In summary
How do you get to big outcomes with data? A few keys to success are important. (1) Develop very bold goals, (2) Find and hire people with Big Ideas (as an example, over the air updates for EVs came from a software mindset not from the automotive industry) (3) Publicize the metrics (4) Create a safe space for failures and hackathons, and (5) Nurture competitive teams to accelerate creativity.
Executive Technology Board (c) | North America & Europe