Coles Group has a clear purpose: to sustainably feed all Australians to help them lead healthier, happier lives. As one of Australia’s leading retailers, Coles employs over 100,000 staff members across a broad set of businesses including supermarkets, fuel, and insurance. The technology footprint that underpins these operations is massive, with a particular focus on supporting time-critical and seasonally driven supply chains. Being in a highly competitive space there is also a constant pressure to innovate, to simplify customer transactions, and to bring new products to market. Coles is a company with a rich history, operating its retail businesses continuously for over 100 years, and has made significant investments in traditional on-premises infrastructure.
Managing this infrastructure is complex; involvement from multiple independent teams required each time hardware is provisioned for new applications. After getting agreement on business requirements, this complexity saw application teams having to engage with five separate teams across concerns like architecture and networks. Coles’ technology leaders were facing an increasingly difficult challenge — finding ways to deliver solutions to customers (internal or external) quickly, especially with a shift to smaller, Agile delivery teams working in parallel.
Two unrelated events gave Coles a unique opportunity to address this complexity and establish new ways of delivering quickly. First, in late November 2018, Coles demerged from the Wesfarmers conglomerate. Amongst other impacts, this led to greater independence in operations. Second, Coles was about to enter its next five-year investment cycle, with dedicated resources available to modernise infrastructure and processes.
Overcoming these obstacles to rapid software delivery came down to two decidedly different options. Coles could either hire a lot more operations specialists to staff each delivery team — competing for talent with the public cloud providers — or adopt public cloud wholesale and enable the teams to self-provision with the support of an internal platform team. Coles chose the second option. The company set an aggressive target of exiting their data centre for 85% of existing workloads, primarily moving these to Microsoft Azure and leveraging SaaS solutions where practical.
It was apparent from the start that transitioning from a near-zero cloud spend to 85% of workloads being cloud-hosted was financially significant and would require disciplined financial governance. As Paul Dubaich, senior finance manager of technology, states “…you cannot spend that much money as a public company and not have a very good financial and transparent cost model”. Besides the demands of being a public company, Paul’s perspective was that you can’t spend that much of the company’s money without being able to justify how it’s being spent, why it’s being spent, and whether it’s being spent well. Knowing that they would inevitably have to do that, rather than facing the pain of retrofitting financial governance the approach was to “…start day one with some amount of cost visibility and we’ll make the visibility more and more granular as we scale up.” With this in mind, Coles implemented Cloudability very early in year one to monitor the growing Azure spend and begin categorising it in business-meaningful ways.
The move to public cloud also gave Coles the opportunity to attack a long-term cultural problem faced by most organisations running traditional infrastructure. The multiple degrees of separation between where value is delivered and where cost is incurred means “…no one can be sure who’s usage growth caused which additional costs, and whether that was a good decision for the company or a bad decision for the company.”
Closing this gap on public cloud, across over 170 teams and 100 Azure subscriptions, was achieved using a programmatic, structured, and software engineering driven approach. For the primary cost identifier, Coles chose to enforce a tagging standard on resource groups. Resource groups had the double advantage of already being a well-used construct and offering full support of tags (even if the usage within doesn’t support tags). For any cost item where engineers had neglected to tag the resource group, the cost would be tied to an official cost centre based on its subscription, providing a handy fall-back mechanism. If individual teams wished to do their own internal cost allocation, they were welcome to do this through regular resource tags. With this strategy in place, Coles then used the Business Mapping capability within Apptio’s Cloudability solution to represent these business rules so that all costs are automatically classified each time Azure refreshes its billing exports.
Coles are approximately one third through the planned workload migration, but already the results have been truly impressive. In addition to speeding up application delivery, entire new use case possibilities have opened up, best highlighted by Coles’ data science team. The main data science workloads, critical for supply chain predictions, were running at the upper limits of their on-premises kit. These workloads were migrated to Azure Batch, and now add and then remove about 800 machines (over 8000 cores), for an hour’s worth of processing. That is roughly 25% of Coles’ entire data centre capacity from 2018, and its instantiated and torn down four times a day! This has allowed the data science team to focus on the business problems, enabling their people to concentrate on building better machine learning models, rather than being worried about maxing out hardware resources. Cloudability provided important validation for the Coles team that the cost profile of running four large processing cycles each day was sensible.
The disciplined approach to cost allocation has also paid dividends. “With Cloudability, we have been able to directly attribute costs to our general ledger (GL) codes in near-real-time, as Azure’s billing exports refresh. This represents a phenomenal degree of automation for such a complex domain and company of Coles’ size”, states Paul. One aspect of this implementation is that it makes it very clear that costs are either assigned to a CapEx or OpEx code — for capital or operational expenses respectively. Both of these code types are associated with business value — either a project business case for CapEx or aligned to a product team who has a value story for OpEx.
This new cost allocation methodology has opened up new levels of team ownership. Coles has run a significant education and enablement program to help cost centre owners (stakeholders previously far removed from managing infrastructure costs) to understand this new type of billing data and the categories within. These stakeholders and their teams now engage with cloud cost information on a daily basis, either through a pre-prepared IT finance pack (this includes broader information on the health of the cost centre) or through Cloudability itself. Particularly useful are curated financial dashboards that are experienced with views specific to each cost centre. The response from cost centre owners has been remarkable. “These stakeholders have become sensitive to costs falling through to their default cost centres due to being untagged. It’s almost like someone grabbed their corporate credit card and took it out for lunch! They want to know what was behind that cost and assign it to the correct GL code”, states Paul.
The adoption of public cloud is just one component of Coles’ modernisation efforts, and from an IT finance perspective, is part of a bigger IT finance transformation. At the heart of this transformation is unwinding the capital-heavy project capitalization model that large enterprises have traditionally relied on. With this model, all software development costs were capitalised whereas operations team costs were treated as OpEx. With the move to DevOps and continuous product development, it is no longer straightforward to split these efforts. In the new model, only R&D initiatives are capitalised, while all other costs are treated as OpEx. While this change might not sound that interesting, as Paul states, “A lot of how the organisational structure is built including the governance models, the approval structures, and organisational behaviours, is all driven by the finance model. From a project manager’s perspective, the dynamic completely changes, with new priorities and incentives”.
Seeing the successful implementation of Apptio’s cloud cost management solution, Cloudability, the IT Finance team saw an opportunity to extend this SaaS-based financial governance by adopting the IT Financial Management solution called ApptioOne. Coles have identified that it is important to have a dedicated management solution in place as they progress through the IT finance transformation. After quickly identifying initial priority use cases, ApptioOne will be ready for use within a few months of starting. Paul and the team at Coles are excited as this initiative kicks off: “we are going to use ApptioOne to help us measure the status quo now, measure changes as they occur, and calculate the overall benefit of this major transformation”.
It’s difficult not to be impressed with the disciplined approach Coles is applying to IT transformation and how the company is leveraging new technologies to deliver tangible benefits to its customers. At Apptio, we are honoured to help support this continued innovation and look forward to another 100 years of Coles helping Australians lead healthier and happier lives.