When Barry Panayi first joined John Lewis Partnership (JLP) around 2.5 years ago, it became clear very quickly that no-one could easily access the data they needed or wanted. Staff at the organization, which operates department store John Lewis and supermarket Waitrose, were wasting too much time trying to get hold of a grid of data served through an old-fashioned tool.
Moreover, the data might not have been exactly what they wanted, and they were spinning it around in spreadsheets. Panayi, Chief Data and Insight Officer at JLP, says:
I saw it everywhere I went. I imagine there's always been some ambition for people to have the right information at JLP. But when I joined, I noticed that it just wasn't good enough at all.
Within Panayi’s first three months, his new leadership team decided to overcome this data challenge. He explains:
It was one of the biggest pain points in the whole organization, of which there are many, but we could control this one ourselves and we thought it would be one of the most transformative ones.
One of the first steps for the team was to define what self-service meant, and how they could achieve it. Ultimately, they wanted people to be able to ask very simple questions and get the answer very quickly, with data that was good enough for them to use in a consistent way that enabled Panayi’s team to control it. The next job was to pick a tool and move very quickly on that.
Panayi had used Tableau in all his previous organizations, going back to one of the very first releases:
I liked it and could see how it would fit. I consulted my team and then the real people that were driving it forward, like Libby [Libby Hickey, Tableau Product Manager at JLP], tested it out themselves in a small-scale way. That gave us the confidence just to go for it.
There was no thirst for a long consultation with lots of different options for achieving self-service data. Instead, the team wanted to get something in place as quickly as possible, he adds:
We just wanted to get on with it because speed isn't something that we usually are that good at at JLP.
The team began testing Tableau technology in 2021, and it was up and running within six months. One-and-a-half years on, 10,000 JLP partners now have a Tableau license, and JLP is working towards opening this up to all 76,000 partners (as JLP calls its employees).
When the retailer initially made the decision to invest in the technology, it didn’t fully understand the potential benefits of the Tableau/Salesforce tie-up. Salesforce had acquired Tableau in 2019, but JLP’s decision was independent of this. Panayi notes:
But as we implemented it, we could see why it made even more sense. We use Marketing Cloud, Sales Cloud, Genie, the WhatsApp chats that triage people to get them the right answer are powered by Salesforce, as is our CRM. The fact it all joins together is more than just helpful. We would not be very good at spinning lots of plates with different vendors handing off data to different places.
Thanks to the integration of the Salesforce and Tableau technologies, JLP’s thousand or so contact center staff using Service Cloud for reporting and analytics can now visualize the analytics and make better decisions from within the Tableau platform; likewise its CRM teams get the output of Marketing Cloud data in their Tableau dashboards.
Another factor in the Tableau buying decision was that JLP had invested heavily in Snowflake. It wanted a tool that would work well with that data platform and facilitate a reduction in the number of tools in use across the organization. Panayi says:
Now that Tableau is able to sit on top of all the Salesforce stuff, it's just another reason why it will be a lot less painful. It's one less reason for people to deviate from our strategy of trying to use the tools that we've got.
However, the Snowflake integration still needs some work, something Panayi has discussed with Salesforce. In theory, Tableau works well with Snowflake on the Salesforce side, but there are some complexities around the transfer of data. Panayi explains:
It works with Snowflake at the top level but in reality, it’s still got potential to be a little bit more optimized. Sharing data, especially having to pull data out of Snowflake to then re-ingest into Salesforce, isn't quite right. The APIs aren't there yet, but they [Salesforce] know that. The Tableau side is nice but the heritage Salesforce products, they're still catching up. If we were on AWS or Azure, it would be a lot easier with Salesforce.
Ahead of rolling out the Tableau system to staff, JLP ran a series of in-person awareness sessions last year, as well as a road show around the business to outline the benefits using Tableau would have for them personally. The retailer has also run boot camps and lunch and learns.
These have been followed by the launch this year of a bespoke data fluency program, built natively into the JLP learning platform, aimed at helping partners have the confidence to use the tool. The objective of all these efforts was to create a self-sustaining, engaged and decentralized team of data enthusiasts, and JLP now has a growing community of Tableau users and champions. Libby Hickey, Tableau Product Manager at JLP, says:
We knew that if we wanted to land Tableau as the strategic and trusted self-serve tool, we had to enable the partnership by upskilling partners, building proficiency and cultivating that community. That's what has driven the central team: what are the different activities we needed to work on to enable 10,000 people to confidently use the tool and more importantly, maximize the value of the data. The tool is just the tool, the tool surfaces the data, it's the people that actually take the actions off the insight.
The actions and insights enabled by the Tableau technology have led to a multitude of benefits. JLP is able to visualize the impact of changing prices; how much stock should go on a lorry; how many facings there should be on the shelf of certain items; and how much it would cost to transport certain items versus others, as examples. Panayi says:
All these little operational decisions are optimization problems and quite difficult because they operate in different dimensions. When you see them on a Tableau dashboard, our commercial teams can make some pretty big calls, like how many people need to be on the shop floor at a particular time based on what stock levels are like and historical patterns. They couldn't have all that information before.
What all this means is that partners are in the right place doing the right job at the right time, and products are in the right warehouse or in the right place in the shop or in the right storage unit. Panayi adds:
All of the big use cases that are driving money are around understanding a number of variables that go into some sort of optimization, whether it be price or a process. We're talking dozens of millions of pounds each time on those. We wouldn’t have been able to do it without the way we are presenting the data in Tableau because otherwise data analysts and scientists would be writing code, producing a model and trying to sell that model to our colleagues. Now, partners can play with the parameters themselves on the dashboard and see the effect of their decisions.
For people to buy in, not just to self-serve but to buy into the recommendations and analytics, they've got to see it and feel it. There's only so far an email with some bullet points telling you how good our model is will work. Instead, it’s open up the Tableau dashboard and they can see exactly what the patterns are, and it takes 10 seconds.”
One of the biggest successes so far is the bakery optimization tool. The JLP data science team have developed a sophisticated model that predicts which bakes should be prepared in each branch at what time. The tool provides in-branch bakers with a comprehensive bake plan through Tableau, leveraging data and predictive analytics to anticipate customer demand, minimize waste and maximize sales. The implementation has enhanced customer experience and improved profitability.
One of the insights garnered was to do a second bake at 3PM on certain days, which goes against the current convention of baking patterns. Panayi notes:
Our bakers believed the algorithm enough to do something that felt counterintuitive. Then, they could see the bread flying off the shelves and they've contributed to the profit of that branch.
This level of trust was fostered by the data scientists and analysts going into the shops, speaking to the people that work at the bakery and asking them what they wanted, Panayi adds, rather than having those discussions at head office and trying to predict the requirements.