You May Be Building a Data Case No One Wants

Massage chairs can cost several thousands of dollars, but massage oils are a completely different product and sell for a fraction of that amount.

Understandably, the people in the business units at global beauty products giant L’Oreal know the difference between the two items. Still, the two were being captured in the same category in an external database for some reason.

The result wholly skewed data, giving L’Oreal’s data team a distorted view of the market and the company’s market share.

The anecdote was shared at the CDOTrends Chief Digital and Data Officer Summit in Singapore last week by Christelle Young, the chief data and analytics officer at L'Oréal South Asia Pacific, Middle East and North Africa.

Young used the example to show the importance of data integrity and how having the proper mechanisms in place to do quality checks on data veracity is critical.

“We didn’t know what was going on in the data team, so it is important that the business is educated and understands that it’s also their responsibility to call out something like this,” she said.

Cleaning out the data mess

L’Oreal is a vast global business with 34 brands and a head count of around 80,000.

Sales are made through various channels, from large pharmacy chains like Watsons and Guardian and e-commerce platforms like Amazon to small individual “mom and pop” stalls worldwide.

In addition to internal sources, Young said L’Oreal had 235 external data providers in her region alone, many using different data formats.

Many of these data sources are “messy” and tedious to ingest, but this needs to be done, and the data must be standardized within a data governance framework so that analytics and automation can be applied.

“It’s really important that we don’t have a data case for something which isn’t a business problem”

The data and analytics team has two objectives, she said. To be a center of excellence in creating use cases to scale across markets and to drive the transformation of in-country units to be more data-driven.

“So, this is everything from the tech infrastructure, the talent and the learning, the pipelines and the strategy on how data and analytics are enabling the business,” Young said.

All data projects have to come from the business first.

“It’s really important that we don’t have a data case for something which isn’t a business problem,” she says.

“We need a 360-degree understanding of the problem and then ask what potential data sources relate to the different issues in that problem. Once we have that full 360 view of understanding the business problem, we can then understand what external data we need to actually run those models for something like what is driving sales.”

Business lessons for IT

Young gave one example of the data team putting together an advanced machine learning model for HR on employee retention without scoping it with the business.

When presented with the insights from the data model, the HR team was asked if they were surprised.

“They said they knew exactly and already had an action plan in place, so we really didn’t need a model,” said Young.

“So, it’s really important for us to say, ‘is this going to change the decision we make as an organization, do we invest the time and energy into running this model, and if we have the answer, is it going to fundamentally change the way we operate,’” she said.

“So, the scoping part is really important because otherwise, we waste a lot of time.”

The need for data to serve the business also meant that data quality was the responsibility of everyone, not just the data team.

The tower of stewards

While the company had invested significantly in a “data governance tower” in the global headquarters in Paris, there were data stewards worldwide who worked with data governance officers.

These stewardships are not standalone roles.

“You could be a key account manager, and you’ve got a data steward hat on your head because you take in data, or it's part of your remit,” said Young.

“So it’s a shared responsibility across our entire organization.”

Data quality also varied across regions, and there were different levels of granularity and standards.

“There are just some things we cannot do in certain regions, and a model is better than no model, but it’s only indicative, and it’s taken with a grain of salt in some areas,” said Young.

“We have a lot of estimates, and we do a lot of calculations in regions with very different levels of granularity.”

In terms of return on investment, a performance analysis structure is set up under the company’s chief financial officer, which looks at the value that analytics and data bring back into the organization.

“A lot of our data is acquired, so like any asset within our organization, we need to understand the value of it and how that value is increasingly deteriorating over time and be able to monitor that quite closely,” said Young.

“So, it’s not just about cost savings; it’s also about revenue growth. Where some data models were to improve the supply chain efficiency and reduce costs, others were focused on recommendations engines, understanding our customer segmentation and how to get more customers on board, how to make sure our sales force goes out to the right dermatologist and doctor and understanding where the high valued offers are,” she added.

“So all those are about growth too.”

Lachlan Colquhoun is the Australia and New Zealand correspondent for CDOTrends and the NextGenConnectivity editor. He remains fascinated with how businesses reinvent themselves through digital technology to solve existing issues and change their entire business models. You can reach him at [email protected]. 

Image credit: iStockphoto/nazarkru