Marketing BS Briefing: Certainly a Slam Dunk Edition
Mobile marketing, link building, anti-personalization, dating, children's stories, Ethics, Attractiveness and more!
Welcome to new subscribers who found me from Tyler Cowen’s excerpt of my review of Talent. I generally flip back and forth between essays and “briefings”. This briefing includes an interesting reader follow-up comment on that Talent review (last section below). On to the briefing:
Marketing
Mobile Marketing Winter: Apple’s changes have provided a significant tailwind for mobile advertising. Both targeting and tracking have become a lot more difficult. Last week Eric Seufert published a primer on what you need to know to survive through these changes. Recommended.
The State of eCommerce Advertising: Related to the above, Fospha (a marketing measurement company) has a detailed report on their generalized findings on this year in digital advertising, including incrementally by channel, impact of upper funnel marketing, and more. Take all of these “general” findings with a grain of salt, but they are useful to sense check what you are seeing with your own data (i.e., your retargeting should have a much lower incrementally factor than your prospecting). You will need to click-through to get the report, but it does not require an email address submission.
CTA: Which Calls-to-Action have the highest click-through-rate on Facebook? Thesis shares their internal testing data. Results vary by industry, but the overall winner is “Shop Now” in both CTA and cost per purchase. Also insightful: Giving Facebook multiple CTA to test makes a big difference (single choice CTR 0.66%, 3-option CTR 0.96%). Very clear takeaways.
Gift Cards: WSJ argues that companies WANT you to use your gift cards, because most people will spend more than the amount on the card. Maybe some individual marketers at some companies feel this way, but generally if a customer wants to give you money in exchange for nothing, that is pretty good economics. See my essay on Starbucks, Loyalty Programs and Breakage.
Link Building: Siege Media has an essay on their approach to link building. They have found creating high quality content is more effective than creating content specifically for the purposed of getting links (comes down to the effort in creating “shareable assets” is better put towards creating more quality content). I particularly liked their blog checklist:
The content section linked in the top navigation.
15+ links were present on the content hub and content hub categories.
3+ related links to relevant posts post-content.
5+ links were added from old content to every new post.
Passive link generators prioritized in internal linking.
A high-quality hub and blog post design
Creative Predictions: Marketers were asked to predict the performance of various creatives. They consistently failed. Your instincts are worse than you think. Always test.
Personalization: The myth that it is a good idea never seems to die. Samuel Brealey has a good essay reminding all of us to just stop doing this. Personalization is needlessly complicated, and most of the time the hard part is getting ONE good message. Once you have that you should usually push it to everyone - instead of trying to come up with a second (not as good) message to send to some subset of your customer base (let alone dozens of different messages to different segments)
NFTs: A few readers have been asking me to write about non-fungible tokens. I clearly have not. I guess I just don’t have anything particularly insightful to say. I think there is something to bringing ownership and scarcity to digital assets, not unlike the art world. But I am not particularly expert in the art world either. In any case The Drum writes about the NFT strategies of L’Oreal and LVMH, and my first thought is: “Some executive asked the marketing team, ‘what is our NFT strategy’, so a bunch of people started ‘doing something’ so they would have an answer”. It reeks of companies building islands in SecondLife or running ad campaigns to get Facebook likes. “Just do something and we will figure it out later”. This quote from the article sums things up perfectly: “There’s no measure of success on this activity – it’s simply about getting its customers used to engaging with the brand in new web3 environments”.
Art Market Prices: Related to my essay series on “Legibility” (Downside. Upside.), I was recently recommended “Talking Prices”, a behind the scene look at how prices are set in the art market world, and how (or how much) the “art” part is separated from the business part. Is next on my reading list. Maybe there are learnings here for NFTs as well?
Analytics and Data
Efficient Markets: The EMH is pretty solid, but a recent paper has found that more than 100% of market returns happen overnight (i.e., if you buy stocks at the end of the day, and sell them at the start of every trading day you will do much better than if you do the reverse). The FT digs in a little deeper, and it seems like the effect is real, and at least partially a function of global markets buying overnight. Hard to pull off at any real scale in practice, and I assume hedge funds are now all over this and the effect will be arbitraged away in the future.
Very Likely: What do people actually mean when they say things like “probable”, “very likely”, or “rare”. From HRB.
Dating: Seth Stephens-Davidowitz has been doing a great job promoting his new book, “Don’t Trust Your Gut,” (at least in the media I consume). This excerpt in Wired talks about some of the findings related to dating, specifically that the traits we say we care about (and act on), are not predictive of happiness once one is in a relationship. His conclusion is things like attractiveness, intelligence and commonality do not matter in relationship happiness, but I have a different interpretation. Once a player is in the NBA, height is not a predictor of how many points they will score. That doesn’t mean height does not matter (it matters a lot!), it means that in order to get into the NBA, short players need to be a lot better at other things in order to make up for their limited stature. I think the same thing is going on in relationships: People don’t setting down with someone because of any one trait, or even any group of traits that the researcher was able to make legible. They settle down with the full package that they believe is the “right fit” for their needs (and someone who presumably feels mutual). It doesn’t mean beauty, or intelligence, or common interests don’t matter - it just means that to the extent their match does not have those things, there is something else that makes up for it - like the height of NBA players. Selection effects rule the world.
AI / GPT-3 / Dall-e-2
Cosmo: The designer of the Cosmopolitan cover created with Dalle-2 goes into details on how it was done on her Instagram. 20 seconds per attempt, but hundreds of attempts plus lots of internal meetings plus 100+ hours learning prompt engineering (how to change camera angles, lens types, lighting conditions, etc). Humans aren’t done yet, if only because AI won’t sit in decision meetings…
Children Stories: Edwin Chen attempts to write children stories with GPT-3 and then illustrate them with Dall-e (to medium success). I have tried the first part of his plan and it works “okay”. I find using GPT-3 for writing is far more effective when you give the tool more to work with, and have it enhance your writing, than when you expect the AI to do everything. On images, Dall-e is up to the task EXCEPT there is currently no way to have continuity with a series of images - so if the main character in your story is Pikachu (Chen’s example), you can make a series of images of the protagonist in different situations, but if you use Dall-e to create a protagonist unique to your story, you can’t have that character continue through your story (at least with the same appearance).
Photographic Styles: Michael Green uses Dall-e to create images in the style of famous photographers. AI does this type of thing really well.
AGI: This post claims OpenAI thinks AGI (i.e., AI with human-equivalent intelligence) is a decade or two away. Given the recent advances this seems like a reasonable prediction, but my guess is that this will be like self-driving cars - we will get 90% of the way there and it will seem amazing, and AI will appear like humans within specific situations, but getting the last 10% will be much harder than we thought it would be.
Followup
Fame vs Money vs Power: My career advice mega-post begins with the idea we all have vices, and understanding yours, and what you are giving up by focusing on one vice vs another, is important when figuring out your career plans. Very related: I just read this old Paul Graham essay where he argues “ambitious people” in different cities are ambitious for different things, and you should aim to surround yourself with people who are ambitious for the same thing you are. I think this is true. But I could be convinced you may want to zig when others zag? Is their value focusing on power in LA when others want fame? Or fame in New York when others are focused on money? Maybe?
Ethics Exams: Also related to legibility, in the last section of this Money Stuff post, Matt Levine writes about the failure of ethics exams to increase ethics. An example of how organizations and government often care more about legibility than impact.
IQ and PE: A reader emailed me after reading my review of Talent with the following comment (wrt the point that intelligence is a very powerful thing to hire for):
Vista may be the most successful PE firm of the 2005-2020 period. Robert Smith the founder is the very rare African American in PE and felt he wasn’t given chances early on for reasons that are obvious. So the entire model is about buying Software companies, making everyone at the company take specialized aptitude test, firing anyone with low scores and replacing them with much cheaper folks who have no pedigree but score highly. Eg they find that for old line monopoly software companies rather than pay a sales guy $350k because they came from Workday, it’s more productive to find someone with exceptional sales skills who sold beef to supermarkets and made $75k. Every Vista CEO I’ve met said when they hired against the tests recommendations based on heuristics and experience they were wrong.
It’s not necessarily High Iq but more executive function w respect to the relevant role. So for sales guys more personality and on your feet thinking focused; for CFOs a certain type of detail orientation and quant potential; for CEOs leadership and decision making
Attractiveness: Follow-up on my essay on intra-elite competition, a new study finds attractiveness is correlated with decreased support for re-distribution. Maybe because unattractive people are still the most discriminated against group that has no natural ally.
Careers
British East India Company: Strange Loop Cannon has an in-depth post on how the East India Company was able to survive and grow for hundreds of years with a very small staff, and a geographic limitations that required many employees to operate without any real oversight from the headquarters. Recommended.
Bragging: Academic paper about four experiments showing the best way to self-promote is to “co-promote” - i.e., promote yourself AND another person at the same time. Self-promotion increases perceptions of competence, but decreases perception of “warmth”. By co-promoting one gets the upside without the downside. For the Machiavellian career boosters among you.
Fun
Breaking Bad: This is a really fun story I found on Reddit. Without giving anything away, in one season there is an important scene about a teddy bear falling from the sky:
They decided to film the scene in reverse. Got a big weather balloon, let it go. Unfortunately a big gust of wind came and broke the safety line and the balloon, with the camera on it, went bye bye.
2 days later they got a call from someone in, I think it's 1 or 2 states over, that they found the camera along with the balloon in their backyard. One of the PA's went and got it. Brought it back and they were all excited to see the footage and see what journey the camera went on.
They hit play and all black. They hear stuff but just complete darkness. Dude walks over to the camera and realizes they left the lens on the camera.
Evaporated Milk:
Keep it simple,
Edward