Treating Data as a Product

Article written by Pete Crawford

Approaching analytics with a product orientation.

Even prior to COVID, the development of mature data and analytics capabilities was regarded by over 75% of organisations as a ‘mission-critical function’ as central as IT, HR and finance units.[i] Now, there is even greater urgency to formalise data as a cornerstone of digital transformation. However, it is estimated that between 60-85% of analytical initiatives fail to be operationalised in support of wider business goals.[ii] In part, this can be attributed to conventional issues such as lack of executive commitment, talent, investment or adherence to the principles of a clear data strategy. But a major obstacle is simply that building or growing a data function is reduced to delivering a set of applications, features and capabilities – regardless of common agreement over data accessibility and actionable analytical insights.

“Insights from Pete Crawford | Head of Data, Analytics & AI, Pete Crawford spends his day-to-day leading strategy, governance and execution over enterprise data platforms, data science and AI capabilities. Speaking at leading industry AI and data events, Pete is experienced in forming and directing multi-disciplined teams to manage enterprise information assets and deliver business outcomes through advanced analytics.”

The value of data is too important to be managed as a project and organised along the lines of an IT program. Data is a strategic asset better served by adopting a product-team model.

Problems with Data Projects

• Investment is attached to a pre-defined scope of work instead of funding a team

• Work is focused on data acquisition and infrastructure development leaving teams removed from owning solutions to actual customer problem – unfortunately, this exemplifies the classic ‘mercenary’ versus ‘missionary’ debate that typically leads to an inability to retain talent

• Data governance is assumed to be a ‘one-size-fits-all’ model more obsessed with reporting structures than responding to nuances such as the diversity of data-sharing interests across an organisation or competing stakeholder incentives

• Teams of specialised data engineers and data scientists operate without appropriate understanding or application of a suitable product prioritisation framework – compromising effective analytical use case development

• The operating model with which the data team engages with ‘the business’ fails to change in response to the dynamics of data requirements or data consumption

• There are dependencies associated with maintenance – such as analytical models which dynamically influence customer behaviour – which contribute to future arguments over funding and resources after handover

In contrast, a product-oriented approach recognises that the outcome of data and analytics activities must produce a product which solves a customer or business problem and will evolve once the product interacts with users. Success or failure starts with whether or not it is adopted by users.

Organisations growing their data function should be encouraged to borrow processes from product, people and financial management disciplines and incorporate them into formalised data management best practice.

Recommendations

1.     Focus on context

A discovery period comprising interviews and audits to understand the context of the organisation from the perspective of the analytics value chain. This is the bridge between data and strategic corporate direction.

2.     Understand alignment challenges by assessing data maturity

This exercise uses the discovery phase to map the current state of capabilities (i.e. data architecture; data security etc.) against target state. To be effective, the target state must align with strategic OKRs and feed into an outcome-oriented data roadmap. Most importantly, the selection of metrics must reflect how the maturity of these capabilities inform the digital transformation of the business toward customer-centric innovation.

3.     Communicate a clear change narrative

A fundamental principle of a data strategy involves integrating data and eliminating silos. This course of action is typically disruptive to existing operational patterns and team structures. Productivity, collaboration and morale suffers without a change narrative and segmented communications plan crafted to the dispirit issues and ambitions of the organisation. Defining and mapping stakeholder interests along with a commitment chart are valuable tools in monitoring the effectiveness of these changes.

4.     Determine the most suitable operating model

Balancing the assorted skills sets of a growing data team against the complexity of data requirements of the rest of the organisation requires flexibility and continuous optimisation. Baseline data maturity and relationships with other business functions will dictate the initial operating structure – typically a ‘Centralised’ model – and how it evolves over time.

5.     Have a rigorous prioritisation process

Inevitably, data teams need to make considered choices as to which experiments to run and how to juggle competing data product options with ad hoc internal requests or unexpected challenges. An understanding of prioritisation frameworks is an essential product management capability. It stands outside Agile and DataOps methodologies and must be sensitive to time, visibility, type of quantitative or qualitative inputs and a culture of data-based decisions. Having dedicated product managers (ideally with a background in engineering work) is highly preferable for building bridges with business units and tracking execution. This is even more critical with AI/ML focused use cases.

6.     Own the customer problem

To be clear, an effective data product roadmap is not a series of features or applications with projected timelines meaningful only within the data team. High performing data teams are trusted to own a set of customer problems. The success of candidate initiatives across discovery, validation and delivery cycles is measured against end states (the problem to be solved from the customer perspective) and metrics (OKRs) alignment to top-down business strategy.

7.     Take responsibility for improving enterprise data literacy

Solely focusing on the strength of data capabilities within a specialised group inside an organisation is no longer viable. In this case, knowledge will not scale and a data literacy gap widens.

Enterprise-wide literacy with data and analytics needs to recognised as an explicit driver of business value. It requires formal inclusion in a data strategy and change management program. A data team will be expected to play a significant role as translators and ambassadors leading education workshop programs to enable measurable improvements in how the rest of the organisation access, interpret and deploy relevant, task-specific analytical models.

8.     Obsess on tracking data events

Data events need to be tracked with the rigor of financial assets. Managing and communicating analytical decisions across the organisation can only occur by capturing events, properties, triggers, values and pathways as data is acquired, transformed, distributed and operationalised. This process must be simple and visual. It augments inputs to a prioritisation framework and OKRs. Moreover, a growing data function needs to look beyond reporting to analysing data patterns that help internal teams such as sales, finance, HR, marketing or product make actionable decisions.

Article by Pete Crawford

[i] Gartner, Fifth Annual CDO Survey, March 2020

[ii] TechRepublic, 85% of Big Data Projects Fail, 10 November 2017