The five minute CIO: Yael Cosset, Dunnhumby

17 Oct 2014

Yael Cosset, CIO of Dunnhumby

“Organisations need to move from data being this asset that is hidden and processed and managed by one team and make it an asset visible to the entire organisation,” says Yael Cosset, CIO of global customer science company Dunnhumby.

Originally from France, Cosset graduated as valedictorian with a master’s degree from École supérieure de commerce in Chambery.

At Dunnhumby, Cosset provides strategic leadership on all aspects of technology, infrastructure and product offerings.

As chief information officer, Cosset is focused on building innovative, world-class products and solutions, including an expanded portfolio in the personalisation area.

He also serves as CEO of KSS Retail, a Dunnhumby company and a global provider of price modelling and optimisation solutions for retailers worldwide, including retail giant Tesco. In this capacity, Cosset interacts with executives and clients from both companies in order to ensure access to the value and services they jointly deliver.

Prior to joining Dunnhumby, Cosset served as vice-president of operations, west region, of MicroStrategy, a global leader in business intelligence technology, where he managed one of the largest and fastest-growing segments of the business, including implementation projects with Fortune 500 companies, such as Safeway, The Gap, Nordstrom, Intel, Disney, Universal Studios, American Express and Visa, among others.

We caught up with Cosset at the recent Oracle OpenWorld convention in San Francisco, California, where he shared the stage with Oracle CEO Mark Hurd and gave his insights on how data science, analytics, big data and trends ranging from smartphones to wearable technology are transforming the retail experience.

Can you give us a sense of the scale of platform you use to analyse massive amounts of data and how important is latency in being able to ensure relevancy?

It really depends on the type of data we are talking about. Transactional data we typically only worry about absorbing that data and those details on a daily basis, sometimes extensively good and frequent but we find that the value of the insight we gather around customer engagement at a daily level is fine.

Some of the digital channel data is real-time because a lot of the data can be used for context-based insight and you can’t afford to have much latency. The data is typically shopping app data, website analytics, search data. You may be online on Tesco.com for example and you enter “milk.” I need to in real-time understand who you are, what your history is with Tesco because the next product I am going to display for you when looking for milk may be different than what I’m looking at. You may be more health-oriented, but I may not, so the right product to display for you would be organic.

For real information you can’t afford to wait.

The second thing I will say about latency – whether you bought a Pepsi brand for example for the past ten years, every single time you wanted a soft drink I probably don’t need to worry whether you bought a different brand today but if you engage with me on a digital channel the fact that you have been consistently loyal to a brand for 10 years and suddenly for the next six months you change your habit then I need to obviously adjust my insight around your brand preferences.

This adjustment doesn’t need to take place real-time but heavy lifting work that can be done in a more reflective way and you don’t need to spend hundreds of millions of dollars on infrastructure to process that.

Your behaviour online, dual location information, the weather information, how it may influence what you may decide to eat for dinner tonight, we need to see real-time.

Transactions might be relevant in real-time you may have searched for something five times on a digital channel and if I’m engaging with you again on digital channel and you are on your way home I need to be recommending offers but not on the same products. It’s an interesting balance

The last thing I’ll say about latency is the process of absorbing the data is expensive from a client perspective but what we’re finding is equally challenging is the ability to perform the computer science on algorithm or models at a pace that is acceptable.

When you are looking at targeting specific offers to provide value to consumers, the complexity of the models has increased dramatically. And the techniques we use are also quite a bit more sophisticated

That’s an area where in partnership with Oracle looking for opportunities to go beyond the traditional hardware software model.

We are looking at some platforms that are available around big data appliances but we’re also starting to think about are there better ways to leverage a technology that exists today and bundling it in a way to do more work around artificial intelligence at scale.

The era of wearable computing devices is dawning. What impact will this have on the retail experience and of course for businesses being able to gather this data, analyse it and gain real insights about what customers want?

I think it is a massive opportunity and it will probably take a while before we have enough critical mass of specific devices for the data to be really meaningful. But at an individual level for consumers going back to every customer is different. Consumers are becoming more and more health conscious – not just losing weight and feeling athletic – but people in the US are challenged by the cost of healthcare and want to take matters in their own hands and acknowledging having the ability to challenge at the source which starts with what you ingest is a big opportunity for consumers to change the outcomes of their behaviour.

The data is massively minimal, the absorption or interpretation of that data is not going to be that challenging because the reality is the Jawbone wristband, for example, will generate the same data from a structured perspective whether you are wearing it or I am.

Once you provide an app that connects to that Jawbone API you can start providing health and wellness related services back to the consumers who care about it and provide recommendations on meals.

If you are really into convenience and leading a healthy life and I have opportunity to provide recommendations for your family, I as a retailer can package a meal for you, put the ingredients on a tray for you to pick up on the way from home from the office. These are the services consumers are starting to ask for and the data is extremely valuable.

Oracle updated a massive amount of products at OpenWorld, which are the ones you are most likely to deploy?

I think it has been referred to as one of the worst named products so far: Zero Data Loss Recovery Appliance. I think it is the one I am definitely interested in.

In a world where we have to provide more and more real-time insight about consumers to engage with them in a personal way in digital channels losses of data can be quite damaging and quality of service is impacted immediately.

You cannot afford to have transactions go offline for an hour or two, it is not acceptable, but it is manageable because ultimately the science for the logic of the data doesn’t change every minute of the day. But when you stop looking at the contextually and digital data we want to make sure we have the right infrastructure, the right technology to protect that and ensure minimal downtime can happen.

A lot of the innovation around exadata and Oracle Database 12C seems to be very important to our database team. I couldn’t really articulate effectively why, but all I know is we are constantly looking for opportunities to increase the performance of the throughput of our analytics and the ability to move a lot of the data and analytics to a more robust environment.

12C seems to be a big opportunity and we are definitely looking at that now.

Our business systems team is continuing to look at innovation taking place around our cloud base of applications.

The CIO is becoming known as the chief digital officer in some organisations. At the same time companies are employing chief data officers. Is there any difference between these roles or is it just a moment in time?

It’s a very good question and a good statement because I do agree that I think one of the most critical misconceptions for CIOs today is to a better job at exposing the data and the insight the data can create for the organisation.

We need to move from data being this asset that is hidden and processed and managed by one team and make it an asset visible to the entire organisation

We have found that this has helped with innovation within company but also as we consult with clients and partners

We do have a very fluid structure with in Dumhummby where we have the best talent in the industry on a data innovation perspective.

All companies should have a separate CIO and chief data officer – it is really a case by case situation, but every organisation should really have somebody focused on insight.

Historically the CIO has moved from technology-related challenges, organisations have created the role of a chief data officer and looked at it from a data strategy perspective, what data should we worry about, go after and in what way.

As you move to cloud apps do you work with one vendor as a one-stop-shop or do you work with various providers?

I can’t really say much about the implementation of the cloud-based app, we are still only a few months into the implementation.

In terms of vendor and one stop shop, we are not committed to being a single vendor organisation. Everything we look at we always do so from culture of challenging what we’ve done in past and make it different.

Our role is to challenge exadata every time we look at new data work and so far there has not been any situation where the revised answer has been acceptable as the right one.

We have other vendors, we work with SAS and Microsoft on some aspects of their portfolio. From a data tech perspective and on exadata. Oracle has proven and continues to be the best solution for us.

John Kennedy is a journalist who served as editor of Silicon Republic for 17 years

editorial@siliconrepublic.com