Sashi Marella, Senior Data Scientist at Viacom Weighs In

There’s a brewing skepticism around data, or as it’s ominously called ‘big data’ — what does it really tell us, can it replace human thinking, is it worthy of the hype? With attention-grabbing headlines like ‘Can Big Data Save the World?’, it’s no wonder a crop of non-believers have been critical about its value. Enter Sashi Marella, a data scientist at Viacom and mathematician with a Ph.D. in theoretical neuroscience. Despite his experience with big data, he’s attuned to people’s skepticism,even people within data science, mistrust the data; it’s not the data itself, they don’t trust the interpretations of the data. There’s a saying in the industry — I’m not afraid of the other person’s data, but their interpretation of it.”

Marella recently left the world of academia behind. He now uses his analytical modeling methods while working with clients to harness the power of data for more effective marketing — whether that be for a Snapchat campaign or a analyzing the best social influencers. He considers himself and his colleagues “doctors,” because “we give the diagnosis but it’s up to you [the client] to take our medical advice.”

With all of the uncertainty about the role of big data, V asked Marella to set the record straight and debunk some of the biggest misconceptions, myths, and assumptions that surround big data today.

Myth #1

V: Big data determines the best influencer for a campaign based on the size of their following, making it a numbers game. Do I really need big data to tell me that Justin Bieber tweeting about my product would be a win?

Marella: We look far beyond how “big” the influencer is, and focus on more impactful factors like the right (in target) followers, engagement, and the campaign goals.

People falsely assume that impressions matter most, when it’s really engagement. How engaged is this influencer with their community? What’s their posting cadence? The number of total followers doesn’t really matter. It’s the level of influence amongst a certain demographic.

V: OK, so it’s not quite as simple as people assume. If it’s not the number of followers, how do you know the best influencer for a given campaign?

M: To do this, we ingest all of the influencer’s social content, from all platforms — whether it’s video, their tweets, photos they’re sharing on Facebook, etc., and we analyze the statistics on how their posts perform. We then use this information to help us understand how much reach and engagement we would get if we were to place branded content on one of those channels.

We also look at the best match between the audience of the influencer and the audience of the brand. So if we were to find an influencer for Nike, we have created models that allow us to rank the top 10 potential influencers with an actual associated number for each — this is our proprietary Viacom index.

Micro-influencers are so important for today’s marketing strategies. We’re looking at influencers with very specific and granular expertise and credibility (e.g., a dodgeball personality with tremendous engagement among men aged 18–34, living in the suburbs, with an audience who enjoys MTV and NFL).

Myth #2

V: The capabilities of big data are overstated, look no further than the 2017 presidential election… we had all of the data in the world but no one predicted Donald Trump would win the presidency.

M: Big data did predict the popular vote. The Electoral College complicates the data.


V: Big data has no emotion.

M: Big data is filled with emotion! One of the most important things we do is measure sentiment. We even measure emojis — each one has a certain emotional value and specific set of emotions attached to it. If we’re looking at response to comedy content, several crying face/laughing emojis hold more weight than say, a heart.

We have a lot of text content we analyze — comments under posts for example — and all of the text we analyze goes into our sentiment and emotionality index. It extracts all of the indicators that give us an idea of how people respond to the content.

If a client is looking to reinvigorate their brand through branded content, we can tell them which social influencer has higher engagement on specific topics, e.g., politics and gender. It’s the way we determine how extreme a particular sentiment is on a scale.

V: But aren’t there are too many emotions to really quantify them into neat little buckets?

M: Actually, psychologists have recently proven that there are four basic emotions that everything falls into:

  • Anger
  • Happiness
  • Love
  • Fear

So we track emotionally charged words in each comment and create a percentage breakdown of how many of these words appear in each of the influencer’s social channels. This helps us predict engagement and sentiment.


V: TV doesn’t have the sophistication that digital does because there’s less data available.

M: The medium doesn’t make or break the case, you can absolutely have TV with data. We match TV viewing behavior with different data sets to create sophisticated targeting. You can have the best digital platform and no data, and your campaign won’t work or you can have great data on a TV campaign and your campaign will be great.


V: Big Data kills creativity when you have algorithms making decisions for you.

M: I agree, in a measured sense. For example, we cannot at this moment have an algorithm that would create an entire campaign because algorithms can give hard numbers and insights into how those numbers are related. Today, big data isn’t replacing the creative process, it’s just pointing us in the right direction.

The takeaway? Big data will only make creative work better and campaigns more efficient. Above all, big data allows us to understand and engage our fans better than we’ve ever been able to before.