Unilever teams up with Microsoft to deliver AI-assisted decision making to users

Profile picture for user ddpreez By Derek du Preez May 24, 2018
Summary:
Unilever CIO Jane Moran explains how the organisation is using Microsoft tools to bring reliability and trust back to its use of data.

Unilever sign, Mexico © Unilever
Unilever is one of the world’s largest consumer goods companies, operating in over 190 countries, with its products used by 2.5 billion consumers every day. With that scale, comes an enormous amount of valuable data. However, getting to grips with that data to deliver value for business functions isn’t straightforward, as CIO Jane Moran explained at a recent Microsoft event in London.

Unilever is now deploying a wide array of Microsoft tools, including its AI, machine learning and BI capabilities, in an attempt to bring trust back to the data and to put data in the hands of users. It is hoped that this will allow for AI-assisted decision making to become commonplace across the organisation.

Moran explained that Unilever is in the midst of a ‘digital transformation’. However, Moran doesn’t want to deploy technology for technology’s sake - and instead is dogmatic in focusing on the business value. She said:

One of the things I like to say to my team is, why are we doing this? For anyone that’s a CIO or CTO, you know you get requests for hundreds if not thousands of projects. You really need to look at that and ask yourself, what business value are you giving back?

So, let’s talk about the business value. What is the size of the opportunity? At Unilever, our BI systems have primarily over the years focused very well on looking back in history. And how we have the enormous opportunity to predict the future.

And with some of the new examples of this Microsoft technology that we are employing, we are improving the accuracy of our forecasting, which is a big deal. We have seen improvements with lean inventory metrics, we can even spot slow moving goods before they become slow moving.

Re-thinking data

Moran explained that Unilever is keen on deploying platforms, as these give Unilever consistency and the ability to scale quickly, at an economic price point. She said that whilst data is at the heart of what Unilever does, in the past data sources came from a number of disparate systems, where there were multiple interfaces, lots of APIs and lots of ETL activities taking place. This made it very hard to get access to the data.

Unilever is now deploying Microsoft platforms to change this. See the images below for an overview of the new data structure at the organisation.

8C8738DF-AA8C-4C7D-B21A-40076D0681C9

F9630A4A-AB15-4534-946F-1012213C43A7

Describing the new environment, Moran said:

These tools now help remove those siloes and democratise the data. It allows us to streamline and automate the data ingestion into our environment. Why is that important? Because we are trying to put data in the hands of our employees. How many of us have so many different reporting systems? You have reporting systems on top of reporting systems.

The source of data becomes suspect over time. And our employees sometimes don’t trust the data because they don’t know where the data lineage comes from. With some of these tools we are building out data catalogues, that helps our business community really trust the data. And trust is an important thing, to get your employees to really use the analytics you’re providing.

Moran said that in the past the IT department did all of Unilever’s reporting. It was the reporting engine. A couple of years ago, it moved to making use of business analysts who could translate and rework the data for business users. However, this meant lots of churn in all that work for the organisation. However, by making use of the data lake, as seen in the image above, Unilever is now able to deliver analytics to users that are trustworthy, using Microsoft’s Power BI tool. She said;

The data lake actually makes it possible with the Power BI tool on top to put the data in the hands of the employees that need to use the data. And it makes it easy for everyone in the organisation to consume the data in this self-service way.

We have done more in the last six months than we have done in the last six years. It’s amazing how fast this is going and it’s very simple. Our source data, structured and unstructured, internal and external, is the basis. This is your SAPs or your big ERP systems. All the systems you have in an organisation. In Unilever we have about 2,500 business applications.

That source is then sent to what we call our universal data lake. And it’s preserved in its native form and it’s time series. This is a really important concept, because we are preserving the source data. We put it in the data lake once and it can be used over and over again for different business purposes. So we curate the data in the business data lake. And that’s really looking at, what are the business metrics and the KPIs that we need to structure the data? So it could be for a geography, it could be for a product, it could be for a business function.

The Power BI tool also makes use of natural language processing and voice capabilities, and includes diagnostics, predictive and prescriptive analytics. Moran said this is enabling Unilever to think about using deep machine learning in the future to provide assisted and automated decision making. She said:

“We are using all of the services under AI services and AI infrastructure. In our roadmap we are going to extend. We started this journey in September. We have made amazing progress. We will begin to use some of the AI tools and deep learning frameworks.”

Examples

One example Moran provided was a HR chatbot that can be used by employees via a self-serve interface. It was an impressive demo, with an employee asking for a name change following getting married. Using natural language processing, the employee was able to chat to the bot, explain the situation and get the name changed in just a few short simple steps. Moran said:

It allows our employees to engage a little bit differently with all the questions that they have. We are building it out for HR, but we can also apply that to programmes that we have in our legal, finance and manufacturing divisions. We estimate that for HR we are reducing this manual churn we have had in the organisation - and we estimate that it’s about a 90% improvement in answering questions. And it frees up the HR professional’s time to really focus on productive ways to support the staff and the business. We can put X number of systems, with all the data sitting behind the front end, and translate. And with the AI sitting on top of it, it’s helping us make the decisions for us.

Moran also gave an example of Unilever’s LiveWire product, which provides business functions with insights into how products and business lines are performing globally. This was previously done via business analysts having to pull data from disparate sources, but is now in the hands of the users via a simpler interface. She said:

It provides brand sales and performance insights. And blends sales and market share data from internal sources, like SAP. But also external sources. The solution is built on the Microsoft data lake and the analytics are provided by Power BI. Key business stakeholders, the marketing department, demand planning, the finance organisation, they’re all very excited about this capability. It’s very easy to use.

It shows, is our brand winning competitively versus the competition? It shows in-market execution and innovation performance. Am I winning profitably? I mean, wouldn’t everyone want to know this easily? For us to produce this kind of data in the past, it was multiple systems with business analysts pulling it all together. Now the technology is sitting on top of the data lake and we can easily aggregate the data. It also shows if you’re meeting your business goals.

Learnings

Finally, whilst still a work in progress at Unilever, having only started the project in September, Moran said that the organisation had already made some significant learnings from its data activities. She said:

What have we learned? A number of things. Number one, you need to have trustworthy data. People will not listen to us if they don’t trust the data. And actually this technology let’s us track the data lineage right back to the source data. Our business users are now trusting that.

The second is around scaling. We need to provide solutions that can scale economically and use across different regions, across different business functions. This technology allows us to do that.

Finally, democratising the data. Putting data in the hands of the people that need the insights. This really gets us a step closer to automated and augmented decision making. And that’s where we want to be, to be an insights driven organisation.