Up until a year ago, Einat Naveh’s days were spent knee-deep in spreadsheets.
Back then she was a senior manager working in Digital Ad Sales and Finance Business Intelligence at Viacom and was responsible for providing the company with data regarding its digital ad sales revenue.
Her team was tasked with creating ongoing management, sales, and financial reports for the ad ops team, as well as monthly and quarterly reports that rolled up to the CFO and were delivered as part of the quarterly earnings.
“On a monthly and quarterly basis, I had to pull and consolidate data from so many sources and my spreadsheets were really large with a lot of tabs,” explained Naveh. “It took me more than 40 hours, which is well over a week. I spent probably closer to two weeks wrangling data, doing vlookups, and ensuring the data is in the same format.”
The Race to Report
All of the reports required accurate revenue and impressions data across the multiple platforms that served ads to Viacom viewers. For instance, to gather display data, the team pulled data from the Google Ad Manager instances for BET, CBS, and the rest of Viacom’s properties. The team also pulled video data from Freewheel, as well as data from myriad SSPs and ad networks with whom Viacom worked, such as Rubicon, SpotX, and OpenX.
All of the manual data pulls and subsequent formatting and uploading to spreadsheets left Naveh, who majored in Business & Corporate Finance, with little time to do the things she went to school for, specifically analyzing the data and gleaning stories from its meaning.
“I spent so much time just getting the data that I didn’t even have much chance to think about it or ask questions. Obviously, I want to know why things happen, and what they mean for the business. But by the time the numbers were ready, it was like time to deliver them up, that was it.”
With the data due before she had any time to think about it, Naveh’s team never uncovered the core drivers behind the trends reported (e.g. “Why did this business behave this way this quarter?”).
More People Not the Answer
Would having a larger team have given Naveh the time she needed to analyze the data and tease out the stories she knew it contained?
“Not really,” she said. “Working with spreadsheets is a one-person job. Perhaps we could have hired an analyst to pull the needed files from the different sources and then another analyst to aggregate into a spreadsheet, but version control quickly becomes a nightmare. Which spreadsheet is the most accurate? What is our shared version of truth?”
Introduction to Burt Intelligence
In August 2019, CBS and Viacom announced a merger, and Naveh was trained on CBS processes. It was then that she discovered Burt Intelligence, a platform that harmonizes disparate advertising data, and applies machine learning to spot anomalies and emerging trends.
“I loved the user interface, and the ability to drag and drop new data into Burt, which was one of the biggest challenges we faced at Viacom,” Naveh said. What once took 20 hours now takes less than 30 minutes.
The visualizations she saw were also game-changing. Naveh had been using Power BI for a number of years but found the tool difficult to use. In fact, her team frequently hired consultants to work on visualizations projects for them. With Burt Intelligence, she saw the possibility of wholly democratized data science. Any team member can build a dashboard, visualize trends and get to the core drivers behind critical trends.
“When I was at Viacom, I saw a shift in revenue, from mostly display to 50% display, and eventually to mostly video. This is the kind of trend that would have been great to visualize. The same is true for the platforms our viewers use to consume content. Our audiences went from desktop to mobile to CTV, and being able to spot and validate those trends early on would have helped us make better decisions about the channels to invest in.”
A New Job; New Insights
Eventually, Einat Naveh left Viacom to join the Burt Intelligence product team, where she primarily focuses on Burt Advisor, a curated, daily email that prioritizes how ad ops professionals should make decisions to preserve as much revenue as possible and deliver the best results for brands.
She also helps clients build end-of-campaign wrap-up reports, but her goal is to empower the people who are in charge of a campaign to do all of the reporting themselves.
“This is an important step for the industry because it will mean that we have achieved a true democratization of data science,” Naveh explained. “If you work in this business, you have a natural curiosity and an impulse to answer questions about your campaigns. That requires easy and seamless access to data.”
Democratization of data requires pre-built templates and dashboards that serve as a starting point for analysis. The templates are based on a range of campaign criteria and goals, such as click-through rate, viewability, or video completion rate. The templates automatically pull information from the sources, whether it’s GAM or OpenX, and populate the fields with real-time data.
AI is then deployed to spot anomalies in the data, such as discrepancies in impressions served, that are difficult to spot due to the enormity of campaign datasets. Machine learning helps campaign managers assess which aspects of a campaign are performing better than anticipated, and which are delivering disappointing results so that they can redeploy media spend to optimize overall performance.
“AI is great for first finding things that don’t make sense, and then helping us get to the root of the issue.” Error prevention is a top concern for Naveh.
But AI isn’t just about problem-solving. It’s also a critical tool for understanding an audience, and how its consumption patterns shift from day to day. This is the kind of insight that can help an advertiser drive efficiency in their programmatic ad spend.
If a client’s top priority is viewability, is the media plan acquiring the desired number of impressions that are in view? Or is it time to swap out some publishers in order to meet the campaign’s criteria? Without AI, the right answers may not come until long after the campaign has ended.
But AI isn’t just about problem-solving. It’s also a critical tool for understanding an audience, and how its consumption patterns shift from day to day. This is the kind of insight that can help an advertiser drive efficiency in their programmatic ad spend.
“When trends are displayed daily on a dashboard, the campaign manager can pivot in time to materially affect a campaign’s performance,” Naveh explained.
Put another way, AI allows campaign managers to focus less on what happened, and more on the leading indicators that predict what will happen as a campaign unfolds. In the future, Naveh would like to see tools like Burt Advisor do more forecasting of trends that affect individual campaigns.
She envisions a scenario where AI helps a campaign manager predict where an audience is going, how the impression mix will change, and how best to spend a budget as the campaign progresses.
Tools that harmonize campaign data and apply machine learning to it are very much in demand. The digital advertising industry has always been data-heavy and has seen a plethora of point solutions and platforms that generate it in abundance. But data is only as good as the analytics that we can apply to it.
The next challenge is to find ways to harness that data so that regular people can make sense of it, and use it to make smart decisions ASAP. This will be the next frontier of decision intelligence.