By Biddappa Muthappa, Data science manager, MiQ
A blog series explaining some of the concepts, processes and technologies we need to do our jobs - in plain English.
It’s been said that ‘data scientist’ is the sexiest job title of the 21st Century. But why?
I like to think that data is the new electricity. It powers everything. And for businesses who want to make optimal strategic, tactical and operational decisions, they need the meaningful insights and actionable intelligence that data science provides.
That’s especially true for marketers and advertisers - and they’re paying attention.
According to research by YouAppi and Dimensional Research, over the past five years, 67% of marketers have significantly increased their focus on data and analysis. And a poll by Market Land suggests that data science and analytics will be the technical skills most needed at digital ad agencies worldwide in the next two years.
Why do they need the skills of data scientists? Well, a big part of it is moving towards programmatic omnichannel, the ‘ideal future state’ of marketing, where marketers can use the data they uncover in each channel to inform their targeting in all their other channels.
In a recent survey we conducted with Digiday - The Omnichannel Programmatic Promise - marketers were extremely optimistic about reaching the goal of programmatic omnichannel, but clearly stated that challenges remain. In this blog, we’ll look at three ways data science can help them get there faster.
1) Breaking down data silos
Traditionally, programmatic advertising has always been part of multichannel strategies, sitting alongside print, TV, out-of-home, radio and so on. But the world has changed. Phones have become smart, TVs are connected, iPods have given way to Spotify and even billboards are digital.
That means almost all screens and channels can be bought programmatically - and it means advertisers have to take an omnichannel approach to programmatic. This is a huge opportunity allowing marketers to target more people more efficiently with personalization at scale, improving cross-channel targeting and attribution.
But it’s not as simple as just flicking a switch. One of the big challenges for marketers in shifting to omnichannel is data silos - where only one team in an organization can access a set or source of data. In our report, an overwhelming majority of respondents - 80 percent - said they sometimes or frequently experience unsatisfactory measurement and attribution as a result of data silos.
To achieve an omnichannel approach, marketers need to integrate siloed data from different sources, to create more reliable attribution models, better cross-platform targeting and align budgets with optimal outcomes.
Data science can help by using machine learning attribution models - in other words, bringing the masses of disparate data together, and then using AI to analyse it - to help marketers get a clearer picture of things like:
- Who their most valuable customers are,
- What behaviours predict what outcomes,
- And what channels are most effective in reaching different audiences.
By adding in data from interconnect devices (eg a TV and a mobile belonging to the same household) and location data, we can then personalize messaging and optimize the kinds of inventory we buy to target them.
2) Better measurement
Another key challenge highlighted in our report was that of measurement. A third of marketers in the UK and Canada and a whopping two-thirds in the US said they were struggling with measurement and attribution as a result of data silos.
But again, data science can help, by closing the loop between analytics (finding the cool insights) and activation (putting those insights to use in campaigns). By developing mathematical models such as incrementality testing and multi-touch attribution, data science can help marketers use machine-driven insights to actively influence campaign performance, not just simply measure it.
That means that when it comes to running an omnichannel campaign, data scientists can provide traders with smarter algorithms for bid optimizations, making a big difference on margins for both advertisers and publishers. These models are developed by looking at historical patterns, campaign budgets, competitive pricing and audience data, then evaluating the right price for every ad opportunity based on the precise KPI the campaign is trying to solve. But this would be impossible without the data scientist’s ability to process and analyse vast quantities of data.
3) Less fraud
As the number of addressable channels grows, so does the number of criminals trying to defraud advertisers and publishers within those channels. Fortunately, data science can help with that too.
Whereas traditional measurement methods are unable to identify and detect fraudulent activity, data science fraud detection models use AI to help identify fraud (eg bot traffic versus real consumer clicks) by learning key browsing and behavioral patterns across all your data.
By examining statistical properties related to the click behaviour of regular users versus the click properties associated with that of malicious bots the two can be differentiated through a variety of machine learning approaches.
Omnichannel means data science
The promise of true omnichannel marketing is within reach - and it’s data science that will help marketers and advertisers get there. Whether it’s better cross-channel targeting, smarter attribution or end-to-end campaign optimization, the tools and techniques of data science are what turn the raw potential of data into the powerful insights that drive success.
Want to find out more? Download the MiQ and Digiday report The Omnichannel Programmatic Promise now.
References
