Apologizes for last week’s “brief” briefing. I sent a longer one later in the week to my Marketing BS list. You can read it here (I think it was a good one!).

Yesterday’s essay was for subscribers only about Youyang Gu and “early vs Late funnel metrics”. Recommended.

Onto this week’s (on time) briefing:

Edward, Startup Curated


  • Amazon Health: The Amazon/Berkshire/JPMorgan JV attempting to disrupt health care may be dead, but Amazon is pushing along on its own. The company started with a pilot program in Seattle, but has now applied for a license in 21 more states. It is currently limited to Amazon employees, but that is Amazon’s model: start with being the “first best customer”, refine the product, get it to scale, and then offer it as a service to others. Watch out hospitals, Amazon is coming for those health care dollars…
  • Marshmallow Test: I loved this experiment. I even had my kids do it. But now it has been shown pretty clearly not to hold up to replication.
  • Walmart: The retailer has hired a banker from Goldman Sachs to oversee the “Bank of Walmart”. Customer acquisition is the hardest part. Once you have the customer, you might as well sell them everything. And banking is a big, profitable category.
  • TripAdvisor: Everyone is building a subscription business these days. The idea for TripAdvisor is to sell subscriptions that give travelers discounts on a subset of hotels. There are already niche travel sites that do this (including one that offers unlimited free hotels and small group travel travel for a very high annual price). I explored this a lot when I was at Expedia. It is very difficult to build a large business on this model. Trip has one advantage: they are owned by a private equity firm that also owns a large number of travel assets. This will allow Trip’s offering to be strong off the bat. But it is still questionable if there is product demand from consumers.


  • Google anti-cookies: A lot more has been written in the last week about Google’s elimination of cookies (and similar tracking technology). This WSJ piece is a nice summary. Effectively Google is using machine learning to assign every browser to a “cohort” of ~1000 similar users. What cohort you are a part of is tagged locally on your browser (not in the cloud), which is theory allows anyone to target by cohort. But since the cohorts are so small, only Google will really have an understanding of how to effectively use them. My point last week stands: This is “more private” but it is also (coincidentally?) really good for Google and anti-competitive (or at least “hyper-competitive”) to other advertisers.
  • Media Habits: While consumers are heterogeneous, traditionally media habits from one group to another have generally been highly correlated. In 2015 16-34 year olds and 55+ year olds were 58% “similar” in the media they were consuming. That similarity rate has plummeted to 21% in 2019 and down to only 8% in 2020. This is the biggest media change/divergence that no one is really talking about.
  • Addressable TV: Roku has acquired Nielsen’s addressable TV tracking business. Just as digital targeting based on users (instead of content) is being limited by Google and the platforms, programatic television advertising buys based on users is coming for traditional television.
  • Advertising Bans: It is illegal for Australian Doctors to promote getting a COVID vaccine due to “The Therapeautic Goods Act 1989” which “prevents medical professionals from advertising or promoting treatments including prescription medications”. (Anti-vaxers are not limited on arguments against vaccination). Related: Last March the Advertising Standards Authority in the UK banned facemask companies from making claims that masks prevent the spread of the Coronavirus. Rules have consequences.
  • Funnel clogging: Marketers often work to fix “leaks” in specific parts of the funnel. If we can increase the click-to-lead conversion rate then the hope is that the extra leads will flow through the funnel and eventually turn into revenue. Unfortunately many of the incremental leads we manage to push through are lower quality than the average and convert at a much lower rate. Angelino Viceisza has done some research on people watching Shark Tank. As viewership of the show increases, more people reach out to the Small Business Administration Centers to explore entrepreneurship (i.e., leads increase). But unfortunately show popularity has NO association with the actual creation of new companies or patents (i.e., sales do not)
  • Feta Cheese: Unlike most social networks, TikTok does not primarily limit content to being shared within a network. Instead the app attempts to show the best stuff to everyone (it is more like Netflix than Instagram). The result is that when something is “good” it gets pushed to almost everyone. This leads to run-away virality at a much larger scale than traditional social networks. The latest example was a video about how to make a Feta Cheese Baked Pasta in a single pan. And now the country is facing a feta cheese shortage.
  • Personalization: McDonalds bought a technology that allowed the company to “suggest upsells” based on customer characteristics and order patterns (think Amazon’s book recommendations, but for burgers and fries). The company expected the tool to provide a 1% lift to sales. It under-delivered and the QSR is now de-vesting the product. If McDonalds can’t make this work, why do you think your company will?
  • Long Term Impact: Airbnb CEO: “What the pandemic showed is that we can take marketing down to zero and still have 95% of the same traffic as the year before. We’re not going to forget that lesson”. This is dangerous thinking. Apart from performance marketing, marketing impact is very long term. Coke can cut their marketing spend to zero on a given week and see no measurable drop in sales - but that doesn’t mean there is no impact. (I am still long Airbnb, but this attitude gives me pause…)

Marketing to Employees

Edward, Startup Curated


  • Competitive Advantage: When marijuana is illegal, production and distribution is done by criminals (by definition). When pot becomes legal the existing players do not expand, but instead they tend to leave the industry entirely and move on to other illegal activities. Pot dealers’ competitive advantage is NOT pot dealing - it is black marketeering (pre-published paper summarized on Twitter).
  • Communication: The argument AGAINST better communication. More communication means less of something else - in this example, “building”. Which is why at one point Amazon attempted to “eliminate” as much communication as possible.


AI/Machine Learning/GPT-3

  • Automation: Kevin Roose at the NYTs has a piece on automation. Reading between the hysteria about “robots are coming to take the jobs”, there are some interesting facts: Sales of automation software are expected to be up +20% this year - so the trend that accelerated during COVID (sales were up +12% last year) seems to be only accelerating further. McKinsey has estimated that there will be 45MM jobs “disrupted” by 2030, which sounds big, but that is only 5MM/year, and over 8MM people change jobs every MONTH - so it is a bit of a rounding error.
  • Tom Cruise: This is a “Deep fake” video of Tom Cruise (worth watching). The author makes the point that this could make things very difficult for Cameo. But, as impressive as the deep fake is, the actor in the video was already a world-class Tom Cruise impersonator. He already looked a lot like Cruise and had mastered the mannerisms. The AI helped take the images to the next level, but it is not as simple and just layering the technology on top of just anyone. Related: Here is the same actor’s Presidential Campaign Ad (Tom Cruise running for president. Recommended)
  • Robot rights: Robots that travel on the sidewalk are being given equal “pedestrian rights”. If corporations can be people, then why not robots?
  • SMBC: God as simulated Neural Network (also recommened)

COVID and the New World Order

  • Zoom Fatigue: com has a short summary of why being on a Zoom call is so exhausting vs an in-person meeting (as well as some proposed solutions). I believe the reasons, but I am not convinced on the solutions…
  • Car Accidents: While people were dying from COVID, the belief was that at least the lockdown would prevent other types of deaths (like the flu, which has practically been eliminated this year). Given a lot less people were driving, one might expect that car accident deaths would be way down. But you would be wrong. Deaths were up 8% over 2019 (rate of deaths +24%, total miles driven -13%). No reason is given, but I expect less cars on the road resulting in more people driving faster. Perhaps drinking and driving increased as well as people avoided using public transportation?
  • Company Retreats: As more companies move to work-from-home, the importance of getting people together face-to-face from time to time may go up. TheDiff argues this is the reason why Apollo (a Private Equity Firm) has purchased The Venetian (a Las Vegas casino) from Las Vegas Sands (a casino group). “[Apollo] probably has a better sense of how much of this kind of corporate travel will happen in the future than a pure casino company would”
  • Long tail risk: Now that we have all lived through a “long tail risk” (the pandemic), what other long tail risks should we start preparing for? A large meteor hit is one (a big enough one would cause global extinctions, but the chances of it happening in any given century are very very small). A far more likely catastrophe is a solar storm or EMT attack. Serious solar storms are estimated to happen about every 150 years, and smaller ones are much more common. In 1989 a storm knocked out the Quebec electrical. A 1921 storm was 10x stronger. And a 1859 “Carrington Event” storm was 10x stronger than that. Storms the scale of the Carrington Event have ~10% chance of happening every decade. These “100x storms” can knock out transformers which even a small number (~9) could knock out the entire US electrical grid for ~2 years. The cost to have back-ups would be ~$300MM, but, just like pandemic preparation, that redundancy does not seem to be happening. We do have procedures to attempt to mitigate the damage from a storm (we have advance warning for ~2/3rds of storms, and even when a storm hits, if we turn off the transformers fast enough, they can get through relatively unharmed), but there is a real chance of massive grid failure in any given decade, and we are likely less ready for it than we were for the pandemic.
  • Impromptu Encounters: Part of what is lost from working from home are the random encounters that happen in an office. Rory Sutherland argues that as good as the office is for those things, a better solution is the pub…
Edward, Startup Curated


  • Leaving Google: Noam Bardin, founder of Waze, explains why he left Google. One of his three main reasons: Employees are generally working to get promoted instead of attempting to help users. This is the end state for many businesses. When a company is small, the incentives are for everyone to succeed together. Once you get to a certain size, your individual impact on the corporation is minimal, and your “reward” and “recognition” when the company does well is also minimal. The employees incentive shifts from helping the company succeed to ensuring their own personal career succeeds. And the cycle is vicious, as the people who do not like that environment leave and more room is made for people who excel and office politics.
  • Learning Marketing: There was a debate two weeks ago (this got cut from last week’s briefing) on the importance of training and education in marketing. The discussion started with an essay by Mark Ritson in Marketing Week where he argued that too many “marketing experts” have no formal marketing training at all. Rand Fishkin jumped on this and argued that formal qualifications should not be a requirement for a marketing job. Ritson responded with, [paraphrase] “You can be a good marketer without formal qualifications, but more marketing training will make you an even better marketer, and we should not demean training as something ‘bad’”. Byron Sharpe, who often disagrees with Ritson jumped into support this time, “gradually these types of marketers [i.e., the ones without training] will be replaced with automation”. This hereby ends the “People Magazine” section of the Briefing.


  • InfoHazards: The idea that some knowledge is worse for you to know is an idea I have come back to a few times in these briefings. Here is a new book on the topic: “Too Much Information: Understanding What You Don’t Want to Know
  • Monoliths: The buzz around the randomly appearing monoliths has gone away, but that has not stopped them from continuing the appear. You can track them here.
  • 19th century bios: Arthur Asseraf has collected a few… (check out the full thread with links to Wikipedia pages)
Edward, Startup Curated

Keep it Simple,


Edward, Startup Curated