Cloud solutions and data lakes
For organisations undertaking digital transformation, it is simply not possible to recentralise everything. Departments will guard their own software and technology, figuring they understand their own requirements better than the IT department does.
Despite this, there is room for IT to reassert its role in the business through ownership of data. The cloud makes this easier, with open application programming interfaces (APIs) and virtually limitless storage and computing power.
To achieve this, IT needs to make sure departments are using enterprise-wide systems that encompass each department’s point solutions. These can include enterprise resource planning, finance, customer relationship management and, most importantly, disaster recovery and security – each contributes to the ultimate centralisation and standardisation of data collected across an organisation’s many functions, departments and silos.
With a vast trove of cloud-stored data, traditional IT can gain a comprehensive understanding of the way technology is utilised throughout an organisation. This empowers specialists to ensure each department is operating as efficiently and effectively as possible, while not siloing data and introducing new bottlenecks.
Presenting at Telstra Vantage™ 2017, software and analytics company Readify’s Managing Director, Tatham Oddie, outlined the company’s outcomes-driven methodology for deriving real-world data insights.
Unlike exploratory data analytics, which seeks to derive insights simply by sifting through a dataset for patterns, trends and correlations, an outcomes-driven methodology begins with understanding a business and the challenges it faces. It then develops clear key performance indicators and metrics for success, ensuring those involved don’t lose focus on their overall objective.
“This is where you have to go out and talk to people,” Oddie says. “People on the front line often have the best understanding of where the problem originates.”
“A lot of organisations get this one-time amazing insight, and they put it in a report and go ‘Wow, that’s amazing!’ but then they have no way to actually run that regularly or have that available in their core business systems – a typical integration challenge.”Tatham Oddie, Managing Director, Readify
With a clear understanding of their purpose, the team then takes stock of the datasets available to them, selecting only those that contain information directly pertinent to the problem to sharpen the investigation’s focus and allocate technical resources.
“Often, you might only have point data – but it’s possible to extrapolate to fill in the gaps,” Oddie says. “For example, house prices don’t tend to suddenly jump up and down, they move in trends. So if there are missing years, we’ll go and extrapolate that out.”
A model is then developed and tested, often with the aid of a machine learning platform. Critically, the model is only developed using a subset of the available data – enabling the last portion to act as a proof to see if the model can accurately predict outcomes.
“A lot of organisations get this one-time amazing insight, and they put it in a report and go ‘Wow, that’s amazing!’ but then they have no way to actually run that regularly or have that available in their core business systems – a typical integration challenge, which can be resolved by optimisation and recommender algorithms,” Oddie says.
Continuous learning driven by the model’s real-world successes and failures allows revision and refinement as your dataset expands and exceptions are encountered.