Avoid Data Transformation Disaster: Essential Learnings for Enterprises

Context

68% of data transformation programs fail to realise value…

With so many organisations now actively looking to AI to fuel their next wave of innovation and optimisation, the underlying need to utilise data better & appropriately is becoming increasingly important. Over the last decade I’ve worked with a number of large enterprises to help recover or execute data and digital transformation initiatives. When I read this statistic in an Accenture paper, I decided to spend 2023 understanding the common success and failure modes in FSI customers across the US, Europe and Australia. The high level learnings from my preliminary study are outlined in this article.

A large number of organisations I work with will describe themselves & their transformations as ‘data driven’. Before I get into key learnings, let’s consider why this focus on data utilisation and its impact on enterprises can lead to success or failure. Most functions of enterprise organisations depend on the right information flowing to them in a timely manner to execute correctly. As data volumes & varieties increase, opportunities emerge to innovate in functional areas through augmentation with new data & rich ML/AI service. Conversely, constraining the information flow that business functions depend on can have adverse impacts, delaying the realisation of business value whilst increasing costs.

The initial study cohort was limited to 16 organisations spanning the USA, Australia, UK, Germany, Netherlands. For each organisation, data points were captured across multiple dimensions which included: users, use cases, organisational structures, cultural norms, communication flow, implementation approaches, investment strategies, modelling approaches, tech paradigms, governance, and operating models. The following strong correlations were found across the study cohort which will be discussed further in this article

1. The top indicator of performance is the level of Business engagement
2. Technology strategies focused on centralisation were dominant in low performers
3. Higher performers favoured product management centric investment and delivery approaches over traditional project investments

Organisational Performance Delivering Value

In the absence of a better baseline, I defined three initial performance categories as outlined below to sort organisations into groups based on how effectively they were able to realise strategic value following/during transformation initiatives:

1. Low Performers: Failed to deliver benefits in 18 months of build commencing
2. Moderate Performers: Delivered benefits within 18 months of build commencing, with data workers using new services as part of key job tasks
3. High Performers: Delivered benefits within 9 months of build commencing, with data workers using new services as part of key job tasks

Once key data points were captured, the observed performance showed a significantly higher gap between low and high performers as follows

Figure 1: Average time to realise value across performance groups

Low performers accounted for 25% of organisations with each still unable to realise value having elapsed 18–24 months post build commencing. In most cases significant investment had been made.

Moderate performers accounted for 56% of organisations with each having delivered at least one value adding business use case within 12–18 months. In these organisations multiple data workers (10s to 100s) actively used new technology and processes in their key job tasks

High Performers accounted for 19% of organisations, with each delivering multiple use cases on the platform with material & reportable benefits within 6–9 months. In these organisations the majority of all data workers (100s to 1000s) conducted daily work activities utilising the new technologies and processes

5 Factors That Drive Success or Failure

Although a number of factors impacted overall performance, five key learnings are outlined below that correlated strongly within the study cohort.

1. Level of Business Engagement

The level of business engagement has a binary correlation with performance, i.e all organisations where data transformation was run as an IT initiative with low business engagement fell into the low performance cohort. For the purposes of this discussion, let’s define effective business management as sponsorship from a P/L owner who stands to directly benefit from use case outcomes

Figure 2: Business engagement levels across performance groups

Moderate and high performers both showed increased levels of business sponsorship and active participation in data transformation initiatives. This typically manifested as business executive sponsorship of use cases, ownership of benefits realisation, and the embedding of SMEs into transformation programs to provide critical understanding around business use cases, processes and data requirements

High performers further improved through centering transformation efforts around key user personas and tasks with active business data worker participation in development & test phases

2. Federated strategies drove better business outcomes

Figure 3: Architecture paradigms across performance groups

In 2020 I conducted a pulse check across our global community and found only two customers pursuing federated technology strategies to drive their data transformations. At the time the industry was dominated primarily by centralised IT approaches (data lakes) made popular during the ‘big data’ era. As part of this study in 2023 a significant shift was observed:

Centralised IT Strategies — 38% of organisations reviewed

This strategy was primarily observed as delivery of an enterprise data lake and associated enterprise data models; 100% of low performers selected this architecture paradigm

Interestingly 22% of moderate performers also adopted this strategy but were able to realise value. Observationally this was likely due to increased business involvement leading to more relevant, use case oriented data models

Federated IT Strategies — 62% of organisations reviewed

This strategy was observed in most moderate and all high performers, manifesting equally across two main approaches

1. Domain oriented federation: 31% of organisations provide autonomy to joint business & IT domains to make appropriate data analytics technology & process transformation decisions to drive business objectives — multiple technologies observed across domains
2. Planned federation: 31% of organisations planned for enterprise wide federation of the data analytics ecosystem, tailoring a data-mesh style approach on a common platform that standardised policy enforcement and user experience

In both approaches teams typically opted for the use of native ‘no-ops’ platform offerings (i.e. BigQuery, VertexAI, PubSub, Dataflow) with direct integrations enabling users to onboard, process & activate data at low cost and with low friction

3. User centricity led to higher performance

Product Management practices were observed in all high performers who designed and delivered change to accelerate specific user personas and job tasks, creating streamlined user journeys, and consistent experiences that accelerated use case onboarding. High performers typically worked with end users continuously from design stages and dedicated effort towards user onboarding and education

4. Business centric motivations led to better performance

Four key driving motivations for data transformation were observed across all organisation which in some cases directly correlated to performance outcomes, Figure 4 below outlines how each motivation correlated with organisational performance

Figure 4: Transformation motivations mapped to performance

Reduce Complexity: This motivation was observed in 100% of low performers with low business engagement. Organisations approached transformation with a migration and consolidation mindset, aiming to migrate key data sources to a target platform & create ‘single source of truth’ data models to enable decommissioning of data silos across business domains. Observationally these programs appeared to consistently stall when trying to establish common models for varying requirements across business domains

Increase Understanding: This motivation was observed across low and moderate performers. When organisations approached the problem with a “Reduce Complexity” lens with low business engagement, they formed part of the low performance cohort. Where stronger business sponsorship and SME knowledge was present, organisations achieved valuable outcomes and formed part of the moderate performance cohort.

Compete & Disrupt: This motivation was observed when organisations shared a cohesive vision across domain oriented business and technology teams, targeting specific user experiences to capture market share. This resulted in focused data transformation roadmaps to enable domain teams to quickly deliver well defined use cases. Higher performers were separated from moderate performers through owning more of the end to end value chain (see factor 5)

Optimise Outcomes: This motivation was observed in customers that wanted to measurably improve the performance of key business use cases to compete more effectively. Transformation teams worked directly with users to define success measures and optimise user tasks across critical user journeys. These organisations all formed part of the high performance cohort

5. High performers focused on end to end value delivery

Consider the data value chain as described in figure 5. Data is collected, stored and then turned into information across the first 3 states. Following this a valuable data intervention e.g. a metric, insight or recommendation can be produced using BI, ML, or AI services to land into a specific channel/process at the right moment to create a new user experience, automate a step, or help bring the right insight to a decision maker at the right time.

Figure 5: Ownership across the data value chain within performance groups

A direct correlation was observed between performance and how much of the value chain data transformation initiatives focused on. In the case of low performers, investment was focused on the storage and management of data, with limited focus placed on usage. In contrast high performers focused across the value chain, ensuring appropriate taxonomies & coding of data at collection, as well as ensuring generated interventions could seamlessly be introduced into business/customer tasks through either organisational constructs like value stream based teams, or technical frameworks that enabled multiple teams to release changes to channels

Strategies for Success

This article outlines some initial strong correlations and early findings, and although there’s more to learn, if you’re about to start a data oriented transformation or are struggling to execute one now, here are some thoughts to consider

1. Focus on Value: If your data strategy looks more like an architecture diagram than a set of measurable business outcomes, start working with business teams to map the key use cases & associated KPIs that will drive your business strategy as your first priority
2. Business engagement is vital: If you’re a business leader and you’re seeing a lot of spreadsheets running key business processes, this is likely because the flow of information your teams depend on is either being restricted, is inflexible or incomplete. Work with IT in a ‘shared fate’ approach sponsoring changes, measuring impact and ensuring SMEs are involved from the outset to help your teams truly realise benefits
3. Be user focused: “Build it and they will come” approaches have long been frowned upon but still dominate in low performers where architectural and technology selection choices still occur abstract from users. Integrate end users into transformation programs from day zero to ensure clarity around user personas, user tasks and processes; create simple, low friction user experiences over architectural marvels
4. Empower: Provide autonomy, resources, and the right tools to domain teams equipped to deliver focused use cases. Where possible empower these teams to own as much of the value chain as possible and where not, focus on removing friction & improving communication between teams
5. Invest in change management: Remember the technology is a small part of the company wide change to enable. Invest in user enablement programs & process optimisation to quickly onboard users and help them confidently (and with appropriate business context) read, work with, analyse, innovate and communicate with data.
6. The ‘sugar hit’ mentality can derail even the best-funded data initiatives. I’ve observed companies spend millions on centralising data only to discover it was unusable, or worse, launch prejudiced AI bots built upon poorly understood data & models. Beware of simply focusing on centralising data or hoping the AI will miraculously find the right insights. To give yourselves the best chance of building viable propositions, start by mapping key user journeys together. IT and business teams should jointly identify where data innovations can improve those journeys. Remember, technology alone won’t transform your business; a commitment to cross-functional dialogue and user-centric solutions will.

Final Notes

In a digital world the relationships between organisations and their customers are increasingly dependent on the data they are founded on.

Data transformation is not a simple endeavour but the potential benefits are immense. As organisations move to adopt rich AI services, increased focus will be placed on ensuring the underlying data ecosystem enables data workers to quickly deliver ethical, relevant and appropriate propositions.

Encouragingly in this initial study 25% of organisations transforming with us fell into the low performance category versus the 68% industry indicator. We’ve already been able to apply some of our early findings & recommendations to help more than half these low performers start realising value, so If you’re about to start your journey or are struggling with some of the challenges outlined in this paper, reach out to your Google account team and we’ll see how we can help.