Stockland has modernised its data platform over three phases, and is now in the process of scaling up use cases and building internal skills.
The property group started the transformation program back in 2019, and had piloted and stood up a cloud-based data platform within the first six months.
It has spent the years since scaling up its use of the platform, under what it is calling phase three of the transformation, and this remains an active program of work.
For Stockland, which owns a large portfolio of residential communities, offices, logistics parks, and town centres, data analytics offers key opportunities to "differentiate from our competitors", according to general manager for data science and insights Jenny van Zyp.
Speaking at last month's AWS Summit Sydney, van Zyp said the company had, for years, used an on-premises data warehouse as the basis for its data analytics and reporting.
"Many businesses can relate to this story: you spend a lot of time energy and money in standing up your data warehouse, your single source of truth, the Holy Grail ... but we found that it was limited in its ability to allow us to extract the maximum value from data," she said.
"We found that the data warehouse was necessary, but not sufficient.
"With intriguing new data sets emerging at an exponential rate, our set of use cases and potential users was growing too.
"Social media data, mobility data, geospatial data - data was growing, and we knew that the technology to be able to work with this data was advancing, too, but at the time we didn't have the tools at Stockland to capitalise on this opportunity."
That led to the decision to shift from a data warehouse to a cloud-based data platform.
The architect of the transformation was Praveen Kumar, who left Stockland in 2020 and is now with AWS.
Kumar said the requirements for the new platform was that it take advantage of serverless technology, and be flexible and extensible to handle new use cases, ingest new data types, or support new analytics technologies and tools.
Stockland started with a four-week pilot project on a single "specific use case", before moving to the second phase, which encompassed the cloud migration.
Van Zyp said that rather than a lift-and-shift, the company sought to rearchitect its existing environment to take advantage of cloud-native optimisations.
An architectural overview
Architecturally, Stockland’s data platform uses S3 as its data lake service.
Data is ingested from a variety of sources, including from a SAP-based enterprise resource planning (ERP) system, a cloud-based CRM, social media channels, Google Analytics, and from Stockland’s web properties.
The company is also now bringing in “non-traditional datasets and geospatial information” to support data analytics work around the environment, Kumar said.
“Once data is in S3, [Stockland] use purpose-built tools to process the data,” he said.
“This could involve cleaning the dataset, flattening the dataset, removing personally-identifable information, as well as preparing the data set for analytics and machine learning.
“The team mainly uses two consumption services to generate insights: one is Redshift for BI and data warehousing, and [the other is] Amazon SageMaker for machine learning use cases.”
Kumar said Redshift was used “to scale and optimise the data warehousing environment.”
“So, for example, Stockland use elastic resizing, which allows you to adjust the compute capacity based on workload requirements, such as BI and ETL, and that results in cost savings,” he said.
“The team uses a lot of automation within Redshift to save on administrative costs, and they are also looking at implementing newer capabilities like data sharing and serverless to further optimise and scale their BI environment.”
With respect to SageMaker, Stockland is using it to “build, train and deploy machine learning models at scale for a large range of use cases,” Kumar saiod.
“The team has used Amazon SageMaker to deliver many of the solutions: for example, pricing analytics, propensity modeling, and they have recently implemented SageMaker Pipelines for end-to-end orchestration, which helps them further automate and standardise their machine learning lifecycle.”
Driving results
Van Zyp said Stockland has seen a marked shift in the value it is able to extract from data and analytics-based use cases since the program began.
"I'll put it this way - I've been at Stockland for over 13 years, and we've delivered more value from data over the past three-to-four years since we implemented the platform than in the previous 10 years combined," she said.
"The platform has definitely delivered on its promise of being scalable, and it's proven to pretty much handle every use case we've thrown its way."
She said it had led to “significant cost savings and efficiencies”, faster experimentation and time-to-insight, and had also been "a game changer in our ability to attract and retain talent”.
In the past 12-to-18 months, Stockland has created more sophisticated data-based outputs using the platform, in the form of "custom data-driven apps".
"We found that there are certain processes and teams that simply cannot be revolutionised with a dashboard or AI model alone," she said.
“There are certain parts of the business that require bespoke applications that can encapsulate and streamline every step of their workflow, from data capture through to decision support.
"Unleashing this type of capability selectively in your business can prove to be a real source of competitive advantage.”
Stockland is now focusing on a “tailored human-centric approach to lifting data skills”, enhancing insights “by leveraging non-traditional datasets and geospatial information” and extending insight capabilities to its customers.
