Some fun stuff this week: We start with a breakdown of where current advertising dollars are being spent, move on to the arms race at detecting plagiarism (and evading detection), and end with an incredible video of solving the Rubix cube blindfolded (My favorite is the recursive joke :) ). Enjoy!
Keep it simple,
This Week’s Sponsor:
Raising your first round of funding can be daunting for any founder. How much should you raise? How do you find investors (and get them to give you the time of day)? How do you nail your pitch?
Join Rami Essaid, Venture Partner at IDEAfund and co-founder/CEO of Finmark, for a webinar on Thursday, February 24th at 1:00 pm EST to get the answers to these questions and more.
Rami has raised over $80 million in venture funding and has sat on both sides of the table in countless pitch meetings, and knows what investors need to hear to buy into your vision.
- How the VC ecosystem works (including how to find and connect with investors)
- How to effectively pitch your startup
- Which metrics investors want to see
- Free resources to help founders get funded
Roblox: Following-up on my essay on Roblox proprietary currency, the company announced that developers/creators on their platform earned more than $500M in 2021 (note this is in Robucks, which can be spent on-platform for not change, but has a hefty transaction fee to convert off-platform into USD$ you can spend for non-Roblox products and services).
Advertising Spend: This year Digiday projects that more than half of ALL advertising dollars (not digital advertising - ALL advertising) will be spent on Google, Meta and or Amazon. Related: Benedict Evans build this chart comparing the major channels. Note that Google alone is almost as big as all TV advertising now.
Personalized Advertising: Eric Seufert asks, “As platforms and browsers reduce the ability of advertisers to target people who are interested in their product, who bares the cost?”. Initially this will reduce the ROI for advertisers who will reduce previously-targeted advertising spend. This hurts the merchant, who either takes lower ROI or lower growth. It also hurts the platform (i.e., Facebook) because total demand for it’s inventory goes down. But then what? The second order effect is that advertising-driven businesses are less profitable. Which reduces the incentive for investors and product owners to invest in their platforms. If each user of your site goes from making you $100/year to $80/year any investment that gets you an additional annually-active user is worth 20% less than it was before. So some investments in the quality of the platform just won’t be made. Conclusion: So reducing the ability to target ads will make free-to-use products worse for consumers immediately (since ads will be less relevant. See my “In praise of Advertising” essay), but also reduce the number and quality of products they have access to in the future. This “invisible” cost is rarely talked about, but it does not mean it is not real.
Pricing: The WSJ (free link) writes about the rise of dynamic pricing. Selection effect bias, but one pricing software company says that 75% of retailers on their platform increase the frequency they updated prices last year with 1/3rd changing prices on produces more than once per day (up from 15% the year prior). Pricing is one of the most powerful marketing levers, and is very under-appreciated in the field (maybe because so many marketers do not have control of the pricing lever).
Vaccination Marketing: In this study 37 scientists tested 23 variables to see what types of text messages could influence Walmart customers to get vaccinated. The results:
- Very little impact, but at least not negative.
- Single best performing text was “waiting for you”
- Best results came from repeated reminders (this is exactly what we saw in a study I did for telecom churn reduction Africa in 2008)
- The scientists prior-to-test predictions on what would work were no better than chance. BUT “prolific survey respondents” did make directionally correct predictions
Here is a Twitter thread by one of the researchers with more details (assuming you don’t want to read the paper yourself)
Television Influence: Psychology Today on why television advertising “won’t change your mind”. Maybe some day this message will get through…
Second Life: Meta has often been compared to that “other” VR-world, Second Life. Given how badly Second Life flopped (after a lot of hoopla), it would be worthwhile to understand the differences between Second Life and what Meta is attempting to do now. Bloomberg interviews Wagner James Au, author of “The Making of Second Life,” on his take on the differences. Good read.
AI/GPT-3/Machine Learning/Analytics/Data Science
Plagiarism: It is common practice now for student essays to be put through software that looks for plagiarism. But now tools like GPT-3 can do things like, “paraphrase this essay in 1000 words”. The AI-created essay beats plagiarism identification tools because it IS new content. This academic writes about how he caught one student doing it (it was pretty blatant: “…the source text mentioned a "graph above", the essay mentioned (but did not include) a "chart above"”). Other (give-away) examples include referring to Adam Smith's classic work The Abundance of Countries; or changing the “separation of powers” to the “division of competences” or the “sundering of puissances”. It feels like a minimal amount of work from the cheater would catch these types of errors. And in a few years the AI will be good enough that even that courtesy glance won’t even be needed - unless the defensive software develops the ability to identify AI-created content. It will be an arms race.
Wordle: I have yet to complete it in two, but I had a streak of four 3s last week. For those who want to improve their ability here is a “Mathematicians Guide to Wordle”. While I enjoy the game, I think I enjoy the meta-game or figuring out what the ideal guesses are even more…
Runway: Runway uses machine learning in its video-editing software that allows removal of background images in a single click. All in-browser. Impressive.
Neural: Nueral is a photo-enhancement company. Their latest product allows you to convert art (painting, drawing, sculptures) into what looks like a modern-day photo. In order to use it you need to create an account. I did, and uploaded the cover of Amazing Fantasy #15 and a drawing of Thor. Then I had to wait for my results to be emailed (this took about 3 hours). The final results were not not quite photo-realistic quality… The AF15 did not work at all. Thor’s was… interesting? Maybe it works better if the photo is JUST of the face, but a to make a third attempt it required a $10/month subscription. I passed.
Mafia Accountants: I generally believe in the “Mickey Mouse Effect,” which states that, all other things being equal, the more attractive a job seems the lower the job will pay - and vice versa. So if there are three marketing managers performing the same activities with the same effort for the same number of hours, the Tobacco Marketer will be paid more than the CPG marketer who will be paid more than the Disney marketers (this is another under-stated reason why women are paid less than men. Men are more likely to work for undesirable companies and products). This study looks at Mafia accountants and finds that organized crime groups are able to attract better than average accountants (defined by comparing the quality of ALL financial statements audited by firms with crime connections with non-connected firms). What it does not say, but I expect is higher, is how much more the crime groups need to pay vs comparable non-criminal firms. h/t Tyler Cowen at Marginal Revolution.
GREs: Verbal, Quantitative and Writing score by graduate major. The chart calls out that Philosophy majors lead in verbal and writing, while still being pretty strong in qunat. But there are other interesting data in the image. Physics majors, who have the highest quant score, also have pretty high verbal and writing scores (similarly economics majors). One might conclude that people who score very high score in quant are just “generally intelligent” and will score pretty high in everything. But that does not hold true for computer science majors who are third in quant, but come dead last in verbal and writing. Facinating. h/t this anonymous twitter account.
Blind Rubix Cube: Sure you could watch the Olympics, but if you want tos ee the peak of human potential, it is worth your time to watch a video of someone solve a Rubix cube blindfolded in under 30 seconds. (Note in this video he does it in 28 seconds, but the world record is 15.5 seconds. So this is like watching someone run the 100m dash isn 18 seconds or so…)
Keep it Simple,