Sainsbury's 'Guttenberg moment' - applying data science to bolster customer service levels

Profile picture for user Mark Samuels By Mark Samuels September 13, 2019
Summary:
Group Chief Data Officer Helen Hunter says modern businesses in all sectors must be able to manipulate information at scale.

Gutenberg Bible

Retail giant Sainsbury’s is using data science to help it anticipate and then meet the fast-changing demands of its customers.

According to Helen Hunter, Group Chief Data Officer at Sainsbury’s, the firm has developed internal capability to help it take advantage of the huge amount of information it collects. The aim, said Hunter, is to help employees develop innovative solutions to customer challenges:

As we look to the future at Sainsbury's, we’re focused on quickly, easily and cheaply distributing accurate information. We're really aiming for a future in which there's absolutely no limits on our colleagues’ ability to share their curiosity about our multi-channel and multi-factor business.

Hunter says businesses in all sectors are currently experiencing what she calls a “Gutenberg moment”. As with the introduction of the printing press by Johannes Gutenberg in the 15th century, she said a fundamental disruption is being driven by the widespread use of cloud computing and high-powered computation, which makes it possible to manipulate data at enterprise scale.

Taking the lead on data integration

Sainsbury’s, which has been in business for 150 years, runs 1,400 shops across the UK. Its business interests include high-street brands, such as its supermarket chain, Habitat, Argos and Sainsbury’s Bank. Hunter says the firm’s vision is to better anticipate and meet the needs of customers across all its bands:

Think for a second about the amount of data our enterprise creates. We have thousands of shops, hundreds of thousands of colleagues, and millions of customers. My role is to catch that data, to collate it, and make it accessible to colleagues across our business, to enable them to creatively use that data to unlock business decisions in different ways and think about how we can best serve our customers.

Hunter's role is to think about how the business can create an ecosystem of technologies to help the firm turn the data it collects into insight that helps boost customer service. To create a single version of the truth, rather than patchy data silos held by individual business units, she is driving the development of a strategic platform called Aspire:

We are building this ecosystem of connected capabilities and technologies because we have reached the point of understanding that there's no way forward in our legacy tech. We know that the realisation of our vision is absolutely dependent on the data. And that is why we are driving extremely hard on the plan to ensure that the many transactional systems across our enterprise are publishing data to this ecosystem.

To that end, she is leading efforts to ensure pipelines are being built that will ensure enterprise systems publish data to a single location. Tagged data is dropped into rule-based storage in AWS S3 and curated as it moves into a Snowflake cloud-based data warehouse. Curated data then moves into a presentation layer through tools such as MicroStrategy, Power BI and Tableau, so that employees can make the most of the firm’s integrated data sources:

We've realised it’s imperative that all of those transactional systems publish data. And that's really the business transformation we’ve been on over the past 18 months. We’re making it so stakeholders understand that if they publish data, that’s historically been held within their individual business domains, that bad things won't happen. We want the revolution of Gutenberg, the freedom of information, the spread of ideas, the standardisation of information.

Making data-driven business decisions

Hunter argues that access to this vast resource of integrated information means Sainsbury’s can now understand demands for its products in greater detail than ever before. It's no longer good enough for a retail business to simply look at availability for a product subcategory, she states. Rather, availability must be understood by individual product SKUs (Stock Keeping Units)  across separate stores:

Now, even three or four years ago, that would have been an impossible dream. And again, to emphasise the freedom of the cloud, this is now possible – we now have the scale of computation necessary to understand over 13,000 SKUs across 1,400 shops with granularity. What my team is doing – in the science community, particularly – is ensuring that we can provide truly scientific, systemic, automated and data-driven decisions to complex, recurring challenges.

Hunter cites the example of dealing with pallet deliveries at store. About 130,000 pallets of products are delivered to the back doors of Sainsbury’s shops every day. These pallets include a complex range of products and it can be tough for staff to decide how quickly the pallet should be broken down to get stock onto the shelves. Store workers can walk hundreds of kilometres in the course of a working week doing that task. So Hunter’s team created a machine-readable format of every single store plan and the location of every single product on every single shelf to help expedite the process:

We started tracking those inbound deliveries in real time to the back door. And we built an algorithm to optimise for colleagues the decision of whether to break down the pallet or not in order to create the shortest possible walk time between the back door and getting the product to the shelf. The algorithm decides when and how the delivery should be broken down.

The data-science operating model that makes this high-level analysis possible has “significantly matured” during the past 12 months. The retailer’s algorithms are hosted centrally and draw data form the firm’s integrated data lake. Crucially, the algorithms also publish their findings back to the data lake, helping to ensure everyone around the business is able to benefit from the knowledge that’s created. Hunter concludes: 

This is very interesting because we’re always generating new data that the organisation has never had before. So, we get more information continually on whether the colleague accepted or declined the recommendation of the algorithm about whether they should break down the pallet delivery. So we are generating new data, but also making it possible for analytics to generate new leads. That sets up new lines of inquiry that we’ve never had access to before.