Media Briefing: How Dow Jones is developing an AI model to help its planning team respond to advertisers’ RFPs
This article is part of Digiday’s coverage of its Digiday Publishing Summit. More from the series →
This week’s Media Briefing looks at how Dow Jones is incorporating generative AI tools into its ad ops workflows.
- AI-assisted ad ops
- Microsoft pays publishers, publishers put up paywalls and more
AI-assisted ad ops
Preparing a media plan in response to an advertiser’s proposal request takes time, takes thought and — given the stakes involved in terms of ad revenue — usually takes a human.
“The Dow Jones advertising business is a mid-nine-figures business. It is largely direct, and that direct business depends very heavily on media planning,” Kedar Prabhu, vp of ad product and technology at Dow Jones, said during an on-stage session at the Digiday Publishing Summit in Key Biscayne, Florida, last week.
But while media planning similarly depends very heavily on humans to create those plans, that is no longer exclusively the case. Dow Jones, for example, has begun incorporating generative AI technology into its media planning workflow.
“We started with this hypothesis about two years ago that we could build an AI-informed automation tool for media planning. The challenge there was that the process is non-trivial,” said Prabhu.
For starters, members of Dow Jones’ planning team are able to process advertisers’ RFPs, which can vary in language and format, and, effectively, distill their objectives and constraints. Then they need to come up with a media plan that satisfies those parameters, which can take a certain level of not simply puzzle-piecing but creative problem-solving. In other words, Dow Jones’ planning team can’t just copy-paste RFPs into ChatGPT and expect the chatbot to spit out a winning media plan.
Instead, Dow Jones met with at least a half-dozen B2B sales vendors to see what technology they had available to help automate the publisher’s media planning process. “But there was no vendor that actually did everything we needed this system to do,” Prabhu said. So Dow Jones needed to look into building the system itself. The company spoke with system integration vendors and determined that doing so would cost in the single-digit millions of dollars — not exactly chump change.
“The problem was that, in order to justify that investment, we needed the business to commit to either significant revenue growth or cost reductions or both,” Prabhu said. And compounding the problem was the chicken-or-egg problem of not knowing how much revenue or cost savings the automated system would generate without having some version of the automated system in place.
So Dow Jones decided to build a proof-of-concept to validate the effectiveness of an AI-based model for assisting in creating media plans. It took roughly six to eight weeks to compile a set of previous RFPs and media plans from emails and phone calls into a data set to train the AI model on. Then it took another four to six weeks to set up the model.
As of last week, Dow Jones was in the midst of putting its minimally viable product through its paces. Prabhu described it as a “zero-shot model,” which refers to a machine learning process that trains an AI model without providing examples of information and instead forces the model to develop its own categorization process. That can be a helpful method for ensuring that an eventual official iteration of the model can handle non-standardized data like advertisers’ proposal requests.
“There will be several subsequent models that we launch over the coming months, with the expectation that within six months we are able to demonstrate that the model — not the automation, but the model — is able to achieve the efficiency and effectiveness [key performance indicators] that we hope to see,” Prabhu said.
What we’ve heard
“As long as Google has the buy-side of its ads business nothing is going to change — and that’s the part of the business they’ll never give up.”
— Publisher on the potential for Google’s ad tech business to be broken up
Numbers to know
$3.99: Monthly subscription price to get around CNN’s new paywall.
£34,500 (~$45,832): Average annual salary for journalists in the U.K.
5 billion: Number of text messages that Advance Local’s Subtext will send this year on behalf of its customers, which include publishers such as BuzzFeed, Condé Nast and Gannett.
-80%: Percentage drop in X’s estimated value since Elon Musk acquired the platform formerly called Twitter.
What we’ve covered
How Disney is nearing its goal to automate 75% of ad sales by 2027:
- Disney’s svp of addressable sales Jamie Power is the guest on this week’s Digiday Podcast.
- More of than half of the streaming dollars that advertisers committed to Disney in this year’s upfront are set to be transacted programmatically.
Listen to the latest Digiday Podcast episode here.
Breaking point or breakup? Why publishers fear a Google ad breakup could backfire:
- Breaking up Google’s ad tech business could kneecap the ad dollars that flow to publishers.
- The concern revolves around the potential for Google to react by cutting off its demand-side platform from the rest of the market.
Read more about publishers’ fears here.
The art of negotiating AI deals, according to Time COO Mark Howard:
- Time has struck deals with six AI companies since the start of 2024.
- The publisher is in the process of negotiating with a half-dozen more.
Read more about Time here.
How Axios increased its pre-booked ad revenue by pitching niche audiences:
- Axios currently has three times as much revenue pre-booked for 2025 as this time last year.
- Existing clients have primarily been the sources of that pre-booked revenue.
Read more about Axios here.
What we’re reading
Microsoft is paying publishers, including Axel Springer, Hearst and The Financial Times, for providing content that will be distributed via an AI-powered daily digest from Microsoft’s Copilot AI assistant, according to TechCrunch.
More publishers put up paywalls:
CNN and Reuters are adopting paywalls in an effort to gin up subscription revenue, and Vox Media’s The Verge is weighing whether to do the same, according to The Wall Street Journal.
Snapchat unsettles publishers:
Snapchat’s decision to combine Stories and Spotlight into a single feed with friends’ posts prioritized is leading publishers to worry whether they’ll lose their audience reach on the platform, according to Business Insider.
Apple highlights narrative podcasts:
Apple is redesigning its Podcasts app to feature serialized shows in a new row on the app’s browse tab, according to The Hollywood Reporter.
The New York Times overhauls its mobile app:
The redesign of the Times’ app introduces a menu bar atop its home screen for users to swipe into alternative versions of the home screen based on topic or one of the company’s other publications, according to Nieman Lab.
More in Media
Media Briefing: Publishers’ Q3 earnings show revenue upticks despite election ad pullback
Q3 was a mixed bag for publishers, with some blaming the U.S. presidential election for an ad-spend pullback.
Workplace policies poised for seismic shakeup post-election
Topping the list of expected changes: a rollback of many health insurance reforms provided under the Affordable Care Act, better known as Obamacare.
News publishers didn’t sustain a traffic bump in the 2024 presidential election week like they did in 2020
Unlike the drawn out process of the presidential election in 2020, this year’s election quickly revealed that Donald Trump would be the winner – and that meant less of a sustained traffic bump to publishers.