Expanding the scope of digital advertising
Every once in a while, you may have heard (or wondered it yourself) that Meta must be listening to us given how well targeted its ads are compared to almost any other platform.
Indeed, compared to most other platforms (other websites, apps, or TV, radio, podcast ads), Meta’s ads are much better. But still, it feels like the “dumbest smart algorithm” out there. I have always thought that one of the enduring bull cases for Meta is that there must be a long way until we reach an asymptote of good personalized ads relevant only for an individual user.
Let’s think for a moment how many ads an average user on Meta’s Family of Apps (FOA) properties see everyday. Say, they spend 60 minutes per day. Within this 60-minute, they spend 20 minutes scrolling the feed, 25 minutes watching reels and stories, and 15 minutes on messaging (likely higher for a global user than US user for messaging). Meta doesn’t disclose these details specifically, and there are different estimates out there for all these variables; for the point I’m trying to make here, I am not striving to be accurate, rather just some ball-park estimates.
Feeds have the highest ad density (1 ad every 3-4 posts). If you assume each post takes 5 seconds to consume, you will see 12 piece of content per minute, ~3 of which will be ads (25% ad load). Assuming the user spends 20 minutes on the feed per day, they will see 60 ads per day.
Ads do not appear within private conversations. However, display ads are present within the Messenger inbox. Exposure depends on how often a user returns to the inbox (e.g., to switch conversations) rather than a continuous scroll. Let’s assume minimal exposure during this time, estimating 2 ad exposures from navigating the inbox.
Ad load is increasing in video formats but likely remains lower than the feed. Video requires slightly longer viewing time than static posts, balanced by quick skips. Let’s assume the user views 6 pieces of short-form video content per minute which means they will watch 150 videos or stories per day. Assuming 20% ad load, that’s 30 ads per day.
If you add it all, an average Daily Active User (DAU) is seeing 92 ads per day or ~34k ads per year. If you assume Click-through-rate (CTR) is ~1-2%, a DAU is clicking 510 ads every year at mid-point of that CTR assumption.
Following the CTR, let’s assume ~5-10% ads actually convert (CVR) to sale which implies only 38 ads (at mid-point of CVR estimates) converted to sale from the initial ~34k ads that the user has seen over a whole year. Despite the sheer volume of ads that a DAU sees on Meta’s properties, it basically is only able to convert just three ads per user per month!
UK CMA’s report on Google also bolstered my belief that there should be a long runway for Meta to improve CTR and CVR. Compared to Meta, Google’s search queries have much higher intent and hence, targeting and attribution are lot more straightforward. Despite that, the report suggests improving CTR and CVR were the primary growth drivers for Google search in the last decade or so.
The CMA report mentioned while commercial query growth was below 3% during 2015 to 2024, real search revenue growth (not nominal) in the UK was still growing low double digit CAGR primarily through higher conversion and click-through rates. If higher CTR/CVR contributed so disproportionately for Google, Meta should have a decent runway for revenue growth once they run out of users to grow their userbase. ATT was a major headwind to this story, but AI has resurfaced the possibility of durable ad revenue growth.
For Meta, AI should be a durable tailwind in improving their ad infrastructure. Eric Seufert has written about how AI efficiency could shape ad loads:
“Generative tools, coupled with AI-empowered automation, will result in ads being better targeted and more resonant with consumer preferences.
As ads become better aligned with consumer tastes, the probability that any given ad results in a conversion increases.”
In fact, as targeting improves, ad load may decrease on Meta over time as disposable income can be the real constraint. In that case, it may make more sense to “entertain” the user without showing too many ads:
“…ads do not follow a Bernoulli process: each trial (ad exposure resulting in a conversion or not) is not independent. The success of an ad exposure is influenced by a consumer’s available disposable income, which serves as a hard constraint on conversion. The success of any ad exposure influences subsequent ad exposures, since an ad-driven purchase reduces a consumer’s disposable income.
if a consumer’s disposable income is depleted, the conversion probability for all future ad exposures falls sharply.
As a result, ad platforms may reduce ad load to better align the volume and timing of ad impressions with a consumer’s capacity to make purchases, moderated by higher conversion rates in the ads they see.”
It is also possible that instead of decreasing ad load, more and more ads can move to brand advertising once disposable income is depleted for a particular period:
“Another potential adaptation that ad platforms could make to navigate the disposable income constraint that becomes more acute with improved conversion rates is to shift more impressions to awareness ads that aren’t intended to incite an immediate response. This could result in ad load mostly remaining constant, with the mix between direct response and brand changing.”
In a more recent piece, Seufert made another interesting point about scope of digital advertising potentially expanding as the ad infrastructure is improving over time, especially for SMBs:
I’d contend that the businesses — almost entirely SMBs — onboarded to the advertising economy will bring customers who were likewise not being targeted by advertising with them. In its Q2 report, the Census Bureau of the Department of Commerce estimates that eCommerce comprised 16.3% of total sales. Capital One estimates that while 84% of US consumers shop online, only 30% of US consumers shop primarily online.
In other words, the local component of the real economy still captures the overwhelming majority of consumer spend. But it’s eCommerce and the app economy that provide the data inputs for advertising targeting: online conversions are captured and propagated in the digital sphere and aggregated for advertising targeting based on digital identity. While offline retail sales can be attributed to digital campaigns (for instance, Meta’s CAPI can receive offline events), this almost certainly represents a small fraction of the overall data used for targeting.
What this means is that some proportion of consumers are currently undermonetized through digital advertising, given that their retail engagement takes place outside of the scope of traditional digital advertising targeting. In other words, these consumers are undervalued by eCommerce advertisers — they either have no data footprint or their footprint is too sparse to be useful for targeting. If the local retailers they do engage with are onboarded to digital advertising platforms, those users become targetable for those retailers given their geographic and interest-based relevance. This is expansionary.
Note that my argument isn’t that these consumers are not social media users currently, or that they don’t generally use the internet. Rather, my point is that the number of consumers who receive targeted, direct response ads will increase as more local, SMB retailers are onboarded to advertising platforms.
In addition to “Daily Dose” (yes, DAILY) like this, MBI Deep Dives publishes one Deep Dive on a publicly listed company every month. You can find all the 63 Deep Dives here.
Current Portfolio:
Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.
My current portfolio is disclosed below: