WATCH // Google Series Talks at Vivatechnology (activate English subtitles in the toolbar)

Vincent Luciani, co-CEO of Artefact debates with Dounia Zouine (CDO, Unify) about ” Data Intelligence: turning data into value “. The interview is animated by Pierre-Louis Corteel, Manager at Google Marketing Platform France

The potential for transforming data into value through product, service or process innovation is immense, and companies that succeed in doing so outperform their competitors. Yet, only 30% of data-driven transformation projects achieve their objectives, often stumbling on scaling up.

At VivaTech 2021, I interviewed Dounia Zouine, Data and Digital Director at Unify (a 100% digital subsidiary of the TF1 group, which includes aufeminin, Marmiton, Les Numériques, etc.), and Vincent Luciani, co-founder and co-CEO of Artefact (a data consulting firm and data marketing agency), about the potential of data, conditions for scaling up, and underlying issues.

Vincent, Artefact’s slogan is “Value by Data”. How do you think brands can succeed in creating value through data?

Vincent: From our experience, we can identify three main areas where we’re seeing significant returns on investment, though this isn’t an exhaustive list.
Marketing is first, particularly digital marketing, through better budget allocation, campaign optimisation and targeting. Improving these factors can double sales.

Next, supply chain, which is becoming more and more data-driven. Today it’s possible to more accurately predict final demand by analysing signals that help refine probabilities, such as marketing investments, distribution network constraints or weather. We estimate that a 10-point improvement in forecast can result in a 1% increase in turnover. This is the type of ratio we saw with one of our CAC 40 food industry clients, where we improved forecasts by 2 to 10%, depending on the business unit.

Finally, customer relations. A good example is what we call “augmented advice”. With language recognition systems (Natural Language Processing), it’s possible to give call centre operators pertinent information during a real time conversation.
Today, leading groups are setting ambitious goals for transformation programmes using data and AI, and the Covid crisis has only accelerated this dynamic?

What is your strategic data at Unify, and how has its use shaped your products and services?

Dounia: At Unify, we develop a whole galaxy of media, content and e-commerce sites, all of which are data sources in terms of quantity and diversity (interest, intent, semantics, sociological profile). This data improves our audience knowledge and enables internal and external uses:

  • Internal uses to develop our content and audience expertise, and to improve our interfaces, advertising and traffic acquisition. We know, for example, that certain content works at such and such a time of the week and that a piece of content reaches its peak performance 7 days after it goes online, so it has to be optimised before that date.

  • External uses by qualifying our audiences and by improving their experiences as well as our advertising targeting capabilities. But also, by creating new businesses: for example, our data enabled us to become a Research Institute in September 2020, under the Unify Insight Lab brand.

The real challenge behind all this work, beyond collecting the data, is how to use it effectively, in a controlled, accountable way, while scaling up. For Unify, this has meant setting up the right organisation and investments in line with our ambition and our data assets.

Scaling up is often an obstacle in these projects for many reasons, because we’re talking about a real transformation that’s difficult to implement.

Vincent: It’s true that too few data projects reach their full potential. Over the last five years, however, a major effort to professionalise data has been made using two key concepts: “data as an asset” and “data as a product”.

Data as an asset means managing one’s data as an actual asset. But to do so, we first need to reduce the “debt” that accumulates around data: the multiplicity of complex IT systems that are superimposed over time, with data sources that are often poorly documented, difficult to access, and sometimes discordant.

To eliminate this “debt “, we need to identify the most important data domains, and create organised layers that improve quality, standardise, and make data accessible. New profiles of “data domain managers” or “data stewards” are emerging, who are responsible for maintaining and connecting data sources in order to secure their use.

Data as a product means creating real software that allows businesses to visualise, use and manipulate data independently. Software that must be integrated into information systems and maintained over time.

This is more difficult than it sounds because it involves fundamentally rethinking how data scientists work, and requires dedicated approaches (such as MLOps), product owners, and tools for steering, orchestrating and maintaining algorithms (such as Vertex, the new orchestration tool from Google).

“Too few data projects reach their full potential.”

Making data a usable asset, transforming it into “products” that can be used by business lines, then maintaining these products over time… A project like this requires new skills. How does your organisation adapt?

Dounia: To be completely frank, the subject of organisation – especially data governance – is one of the greatest challenges we face, and we’re trying to move forward by working closely with our IT department. It’s all the more difficult because with data projects, we often see costs before profits.

That’s why we began with the use cases most likely to improve Unify’s revenue, gradually shifting mindsets and building trust. Without trust, there is no change.

In the end, what helped us most was having in-house data expertise, and also a proactive organisation. We centralised the tools, skills and technical base.
Now that we’ve reached maturity, and can clearly see the value, we realise that to keep scaling up, we need skills in data quality, security and governance.

Scaling up implies a certain maturity of organisational data collection, security, provision and use. It also raises new ethical questions.

Vincent: Yes, with the spread of artificial intelligence, one has to be particularly careful, as unintended biases can arise when using historical data that is itself biased.

For example, if a credit institution’s database contains more men than women, an algorithm could interpret gender as an eligibility criterion, which no one wants.
This is why decisions made by algorithms must be documented and monitored (explicability) to ensure they comply with the institution’s code of ethics, and with ethics in the broadest sense.

Without ethics, there is no trust. Without trust, there is no change.

Artefact Newsletter

Interested in Data Consulting | Data Marketing | Digital Activation?
Read our monthly newsletter to get actionable advice, insights, business cases, from all our data experts around the world!

Newsletter Sign Up