Market Snapshot
In 2025, the online advertising marketing size was ~500B. Notables:
- North America has the largest revenue share of over 34%.
- By platform, mobile leads at 59% of revenue share.
- The top 3 are: Google ($214B), Meta ($196B), and Amazon ($68B)
eMarket’s digital ads projection for 2026. Meta is expected to surpass Google for the first time. Didn’t expect to see MSFT so low, and even even, signficantly buoyed by LinkedIn. Also, Pinterest over Reddit is surprising.
Targeting Ads Based on Mood
NYT’s nytDEMO launched perspective targeting, an ad product that allows advertisers to place ads based on the sentiment that the content is predicted to evoke . ESPN, USA Today, and Oracle’s Grapeshot have similar products too.
Available emotions include positive ones (optimistic, inspired, self-confident, amused, adventurous, in the mood to spend, love, interest, hope, happiness, nostalgic, indulgent, competitive, informed) and negative ones too (sadness, boredom, fear, hate). On whether the newsroom will be prodded for more emotional stories, NYT claims that it already produces many emotional stories. Furthermore, the ads are less privacy-intrusive.
Less privacy-invasive, true, but maybe more predatory? Tracking is bad, but at least the user can recognize that the ad is following them around and up their guard.
The major challenge is proving that mood-based targeting works and at scale. In areas like music, mood-based targeting’s effectiveness is more plausible.
Is that I’m more likely to play sad songs when I’m sad, rather than the sad songs making me sad?
Online Advertising is Ineffective
eBay paid $20m to MSN for sponsored links for people that searched for ‘eBay’. Consultants viewed it as a $245.6m return. Asked whether the ad clickers were going to use eBay anyway, the consultants invoked fancy math like Lagrange multipliers. When Prof. Tadelis halted Google ads on ‘eBay’, ‘shoes’, ‘shirts’ and ‘glassware’, the effect was insignificant. However, it’s not in the best interest of eBay’s marketers to know whether its campaigns are profitable.
P&G, Chase and Uber observed similar results: cutting digital ad spending did not affect business outcomes. The vast majority of impressions and clicks come from bot activity.
Warrants more investigation. If ad provider knows that the traffic is a bot, then are they not scummy because they know that the bot won’t buy anything?
Attribution to an ad is easy to measure. The question of incrementality, how many of those “conversions” would have happened anyway, is harder to answer. Google’s Ad division doesn’t seem to have convenient tools for determining incrementality.
Earnings from Fraudulent Ads
Meta projected that it would earn $16B (~10%) of its annual revenue from running ads for scams and banned goods. Meta rebutted that the 10% estimate was rough and included many legitimate ads as well.
Meta’s internal systems predict if a advertiser is committing fraud. If less than 95% sure, Meta charges a higher ad rate instead of banning them. Even then, smaller advertisers get 8 strikes, and high value accounts can accrue more than 500 strikes. Meta reported that the higher ad rate showed both a decline in scam reports and a slight decrease in overall ad revenue.
“Slight decrease in overall ad revenue”. Is the penalty enough to make fraud unprofitable? Maybe it’s like the $1B fine for $16B – it’s still profitable to keep spamming.
Meta anticipates penalties of up to $1B. These fines are smaller than Meta’s “violating revenue”. Meta decided to do more in response to impending regulatory action.
The team responsible for vetting questionable advertisers weren’t allowed to cost Meta more than 0.15% of the company’s total revenue.
In 2023, users were filing ~100K valid reports per week of fraudsters messaging them. Meta incorrectly rejected 96% of them. Meta’s internal target was to reduce false dismissals to at most 75%. Meta’s user base is exposed to 22b organic (not involving paid ads) scam attempts every day.
Undescores the fact that Meta’s real customer is the advertiser, and not the users. If advertisers were complaining about something, then they’d probably get company attention because there is a more straight line to revenue.
After police in Singapore provided a list of 146 scams targeting its citizens, Meta found that 77% violate the spirit of its policies, but not the letter.
References
- The New York Times Sells Premium Ads Based on How an Article Makes You Feel. Rick Edmonds. www.poynter.org . Apr 10, 2019.
- The New Dot Com Bubble is Here: It’s Called Online Advertising. Jesse Frederick; Maurits Martijn. thecorrespondent.com . Nov 6, 2019.
- The New York Times Advertising & Marketing Solutions Group Introduces 'nytDEMO': A Cross-Functional Team Focused on Bringing Insights and Data Solutions to Brands. Danielle Rhoades Ha; Angela He. investors.nytco.com . www.businesswire.com . Feb 15, 2018. Accessed Aug 30, 2021.
- Project Feels: How USA Today, ESPN and The New York Times are targeting ads to mood. Lucia Moses. digiday.com . Sep 19, 2018. Accessed Aug 30, 2021.
- Want to improve your measurement? Get a grip on incrementality. Avinash Kaushick. www.thinkwithgoogle.com . Jan 1, 2021. Accessed Oct 14, 2021.
- When Big Brands Stopped Spending On Digital Ads, Nothing Happened. Why? Augustine Fou. www.forbes.com . Jan 2, 2021. Accessed Oct 14, 2021.
- Meta is Earning a Fortune on a Deluge of Fraudulent Ads, Documents Show. Jeff Horwitz. www.reuters.com . Nov 6, 2025. Accessed May 8, 2026.
- Online Advertising Market Size, Share | Industry Report, 2033. www.grandviewresearch.com . Accessed May 8, 2026.
- Meta to Surpass Google in Digital Ad Revenues for First Time Ever. www.emarketer.com . Accessed May 8, 2026.
How successful is this approach? Google, the largest online advertiser, does not mention it in their ads targeting help page , but Google is not in the publishing business either. Did not find much success for “mood-based advertising” on Google Scholar.