Inside the Dentsu Aegis Network India Data Sciences division with Gautamm Mehra

In an exclusive chat, Gautamm Mehra, Chief Data Officer at the Dentsu Aegis Network Data Sciences division shares how his team is building data driven process flows to make business decisions smarter and reliable

gautamm_mehra

“You show me data and we make a decision,” says Gautamm Mehra, a true believer of data when it comes to decision-making. The serial entrepreneur added that he has always been a data guy. His last venture Ultraviolet Digital Solutions got acquired by iProspect, which later got acquired by Dentsu Aegis Network (DAN).

Today he is the Chief Data Officer, leading the Data Science division in India, a small team of smart people who are helping business decisions to be smarter, quicker and consistent.

But we live in times when every other agency wants to own a lab and mine data, to serve their clients better. However, there are few agencies who are solving business problems with data. Lighthouse Insights was informed that DAN’s India Data Science division is not only solving business problems but is also building products every month that either create new revenue streams or decrease costs significantly.

In my recent trip to Mumbai, Gautamm was keen to sit down and walk me through the business problems the data lab has been solving. This was my first meeting with the data guy.

Travelling early morning from Pune to Mumbai, I landed straight at the DAN Worli building. Post gulping down a much-needed glass of water, I made myself comfortable in his open cabin with its back facing the beautiful Worli sea face.

Before starting my recorder app, I asked him: Isn’t this view a perfect work motivation? “I rarely use the cabin (sit mostly with the team) but at times such scenic views can be work distracters,” Gautamm admitted, dressed in a comfort fit striped t-shirt and jeans.

Without wasting time, we got down to business. “We don’t have PPT’s at meetings, we walk in with product demos. We also follow a startup culture. Which means that we have a running model ready in 48 hours and we also believe in make fast and break fast,” he shared as he fired up his Macbook Air.

To start with I was given a product demo that creates a media plan in seconds after taking the required parameters. For testing, we decided to build a media plan for an auto giant that wants to target Indian males within the age range of 30-35 because they are interested in one of its affluent car. The objective of the campaign was to get website clicks with a maximum budget of one million.

In the next 30 seconds, the product pulled out data for the particular target audience from Facebook. “It has built a complete media plan with data sets like audience sizes, bids, actions, among other data sets. Along with a graph that shows you what other competing cars are of interest for the targeted audience.”

Subsequently, the product also creates a Facebook ad campaign with the given parameters, budgets, and interests. The only bit remaining is uploading the creative for the campaign and the entire plan can go live in a minute.

“At present the same process takes 2-3 days and with this product built in just 48 hours we can make media plans live on the go.”

The next product has the power to bring Google keyword optimization for Facebook data. One of the problems with Facebook is that we only get bucket level data and not interest level data. This means that a user has to create a large bucket so that she can scale but it also leads to wastage. On the other hand, if you create a small bucket there is an increase of cost. To solve this problem the team built a product called the Laser Site. Based on an ad id, Laser Site will pull up interests that are working and that are not working well. “With this product we get to know what is not working, we fix that and what is working really well we maximize it.”

After giving two product examples that run on Facebook data sources, I was shown another product that builds over the power of email ids and other social data that is available on the Internet. To find how effective the product is, I typed in my personal email id: my Facebook id was the first public data to be picked up, along with all my other social network profiles.

In addition, it also pulled my professional and location details, all in less than a minute. “We don’t want to look at your bio but we want to find out how a particular audience is active on Facebook vs Twitter vs LinkedIn. Besides we are also trying to add a layer of validation for our ORM process, so that we know who the person is before the interaction,” Gautamm informed.

The team is also working on building a psychographic profile of a person. This will help in customizing responses. “This is a long term vision where we wish to build a psychological model of the person just based on the public data available. This will provide us the required details on how we should communicate with a certain customer.”

Ecommerce or telecom companies who are slowly evolving their social media communications can look for such intelligent products. We are living in the world of personalized communication where brands have started to think beyond a regular hello on social media. For that a psychographic profile of a customer about what he likes to talk and how he loves to be greeted will be really handy.

“90% of communication on social media doesn’t need a personalized reply but there is that 10% who take the effort in making a brand aware of a certain issue or situation. It needs to be taken care of and addressed.”

Another product that the team has built recently is doing the job of reading images. The tool reads the image and tells the color combinations used and what it consists of. Now for a brand this becomes interesting insight, as the tool can tell when a certain image with a certain color, and with a human face performs better than all other posts that don’t contain a human face. “Not only are we doing analysis of text  but we are also building capabilities for visuals while bringing insights.”

Not just digital, the team has built a product that is providing insights for brands running TV as well as digital campaigns. The product is capable enough to tell a brand based on the supplied parameters, how much it should be investing on TV and digital to get the desired objectives. “For one of our brands we used the product to cut their TV budget down, re-invest on digital and the results were phenomenal. Their cost per lead was down by 88% and this was during a period when all the competing brands were running aggressive campaigns.”

Going further plans are to integrate programmatic videos to the product along with Facebook and TV. Subsequently, talks are on with video streaming players like Hotstar, Voot, among others to be part of the product. Additionally, there are products that also tell clients how wisely they should invest in different platforms, without increasing the budget limits.

“Whatever we build should be explainable in five words. If it is taking long to explain then there is something wrong with it.”

After the demos, I was keen to understand Gautamm’s vision from hereon. “We want to build data driven process flows and it could be in any field not just limited to digital only. And this is what we want to keep doing.”

dentsu-aegis-network-data-lab-team
The team

Elaborating on what is in store; I was given another demo of a bot sitting on Facebook Messenger. So in the coming months, media planners or brand guys wouldn’t need to hunt for campaign results. The bot will provide all the required data in real-time on the mobile screen. Discussions are on to make the bot smart enough to run the media plans, pause and optimize on the go.

While it was cool to see the bot working, for me it was scary as tomorrow these bots might just eat away the jobs of media planners in an agency. Worse, you might just find bots running a media agency or at least a wing! Gautamm laughed, he said he wasn’t sure if bots will kill jobs or not but one can’t leave large jobs for manual operation. “Today railway track changes are done automatically, you don’t want it to happen manually. There are things that need to be automated but when it comes to decision-making it will still remain with humans. Bots will help in making decisions smarter, consistent and faster.”

Before bidding me farewell, he told me he is flying to the US to demo these tools at the DAN offices. “I am really looking forward to this trip as it will give our team an opportunity to showcase the products we have built and tested here to the US team. Usually we have been using products that have been built and tested in the US. And it has been my personal goal to reverse the cycle by pushing innovations from India.”

Furthermore, he stated:

“I don’t want us to be an outsourced company for data analytics. India should be seen as the exporter of data driven products.”