Weekly Reading Notes #1-22

26 June 2023

Jun 26, 2023

Busy few weeks of travel (Santorini & Athens, it’s a tough life) - but I need to get back into a regular cadence on this. In other words it’s time for some classic delivery management - time is fixed & scope can vary aka going forward this newsletter will go out every Friday but may have more/less content depending how busy I am.

For now, strap in for a bumper edition capturing several weeks of notes.

Product Delivery

Related (perhaps very tenuously) to getting the right things done.

Apple Vision, very interesting write-up in Stratechery on the Apple Vision and the future of AR/VR. As a technology I’ve always been somewhat fascinated by AR/VR - the utility for work seems clear and the transformative potential for gaming is clear. Having dabbled with them a bit I now tend to think that the work use case is stronger and the gaming use case more limited (racing/flight sims but not FPS). The most interesting question though is whether the competitive advantage will accrue to the fidelity of the video (Apple) or the depth of the social interaction it enables (Meta/Snap). In a way this is actually Meta applying the same strategy to AR/VR as they are doing for AI.
Disclosure; long Meta.
The Bet via Paid In Full, a good look at the history of ‘robo-advice’ in financial services. In a post-GFC world there was (understandable!) frustration about financial advisors who systematically provided bad advice because it was better for their business as advisors. Throw that in with the internet maturing + the rise of ETFs as an asset class and you get tech people launching ‘robo advisors’. The basic premise of which was for the middle-class to cut out the shady middlemen advisors and whack their savings in a low-cost self-managing fund with a nice UI. These businesses have done well and now they have billions of assets under management (AUM) - but far from killing off financial advisors there are more than ever before. So what gives? Really there are a few factors:
- Rising tide lifts all boats, 2009 to 2020/211 was the longest bull market in history so all types of advisors saw both growth in existing AUM and significant net-new inflows.
- Misunderstanding the value prop, to someone who knows about finance an advisor has limited value (they’re telling you things you already know) and they extrapolated that to other customers. In reality advice is only part of the mix, the more important aspect is relationship and trust - which even a great UX/UI can’t do.
The upshot of this is that robo-advisors have found their actual competition is DIY investors - this is a tough segment to crack because the customer acquisition cost is high (they’re not interested) and the lifetime value is low (they’re cheap). So now you see robo-advisors being the commoditised complement of DIY platforms (where they help retain customers) and full-service firms (where they help acquire customers who don’t yet need a human advisor). There are lessons here for generative-AI boosters.
I also enjoyed this aside: “The first recorded story of a financial advisor in human history was Joseph, advising Egypt’s Pharaoh through a fourteen year stretch of feast and famine. Joseph was paid an AUM-based fee in the form of a percentage of the farmland.”
What the Board Wants to Hear in a Product Presentation, business strategy is product strategy. Businesses have goals (make more money) and strategies to achieve them (sell more widgets, etc). Everything should flow from this and clearly connect to this. Product strategy is a way to deliver the business strategy and goals - the same goes for marketing, sales, technology, etc etc. So what does this mean?
- Centre your product strategy on how it achieves the business goals. The OKR framework is a nice technique to help with this.
- Cascade this right down to deliverables. I like to write epics with both an expression of user value (As [user], I want to do [action] so that [result]) and business value (How will the epic change/reinforce user behaviour and how will this contribute to the product/business strategy). For example a user registration journey might be nice/fast to make the user happy but the business result we care about is conversion.
- Report progress against the goals. Once your work is aligned to the business goals then it’s easy to both track build progress (we have delivered X features to drive Y result), measure results (have X features delivered Y results) and adjust as needed (more/less of X, some Z instead, etc).
Why Kick's $100M Spending Spree Will Fail (via friend of the newsletter), interesting look at Kick’s strategy to try and disrupt Twitch by paying massive premiums for creators to switch to the platform and jump-starting organic growth. The piece is generally negative - paying for top talent may be mistaking the role of the platform as a quality filter rather than an aggregator. I’m a little more positive - with caveats. Paying for top-tier content can bring subscribers (see Netflix) but it’s important not to overpay, I think the way for Kick to structure those deals is with relatively low upfront payment but relatively high performance payments.
How to Model a Product Metrics Dashboard (Part 2), in product it is inevitable that you will need to build a dashboard for the product itself (especially B2B) and/or to measure performance of the product. Having selected some stats this is a decent primer on how to go to the next level of detail on things like mean/median/mode, intervals and visualisations.
The $2B Dumpster Fire That Was Supposed to Revolutionize Hollywood, Why They Died is a great deep-dive series on failed businesses, this one looks at how Quibi burned through $2b before collapsing. Personally I’d boil this down further to to one core problem - they tried to launch at full scale on Day 1 but they didn’t have a content library (so had to buy a bunch of garbage).
5-Bit Fridays: Running experiments that don't suck, successful sales calls, weekly round-up from How They Grow, two especially good insights:
- How to run good experiments; I especially liked the focus on making these statistically meaningful by having an explcit hypothesis, paying attention to sample size, minimising confounding variables (change one thing at a time), and having control groups (A/B).
- How to do sales calls, sales isn’t my speciality but it’s an important skill to develop. The most salient parts here were digging deep to find the impact of a user problem not just the problem (the pain in pain-point) and mirroring this back to emphasise benefits.
Age of Invention: How the Steam Engine was Invented, the whole piece is worth reading but if you want to skip the history lesson and see why I snuck this into the product delivery section then it’s this:
Far from being an invention that appeared from out of the blue, unlocked by the latest scientific advancements, it started to take shape from decades and centuries of experiments and marginal improvements from a whole host of inventors, active in many different countries. It’s a pattern that I’ve seen again and again and again: if an invention appears to be from out of the blue, chances are that you just haven’t seen the full story. Progress does not come in leaps. It is the product of dozens or even hundreds of accumulated, marginal steps.
How Google Measures and Manages Tech Debt, this is an interesting framework for categorising tech debt and systematically managing it but unfortunately despite looking at 117 variables they didn’t find a predictor (leading indicator) of tech debt. Reminds me of Tolstoy "All happy families are alike; each unhappy family is unhappy in its own way." Personally I would differentiate between intentional and unintentional tech debt. Explicitly taking some decision (extra operational effort, worse performance, more costly infrastructure, etc) because of some payoff (faster time to market, etc) and with a plan to pay back the debt is very different to tech debt slowly accruing in the background without people being aware of it.
I gave ChatGPT complete control of my city break. Here's why I wouldn't do it again, this seems like a pretty shallow article but it actually reveals something quite important about the inherent limitations of current generative-AI models. When you ask Chat-GPT to plan your holiday it isn’t actually planning your holiday. It is probabilistically guessing the set of words (tokens) you would like to see based on the reams of blogs/forum posts/etc (training data) which has been fed into the model. They hallucinate because the models don’t actually know things. They can’t find you undiscovered gems because by definition they aren’t in the training data. Generative-AI is exciting technology but I think the benefits will accrue to players who can feed it the best/most applicable data and combine it with human work (co-pilot).

Off-Topic

Not even tangentially related to product delivery, but still interesting (to me).

Why trying to "shape" AI innovation to protect workers is a bad idea, top of the list because this one really straddles the on/off topic divide. The current wave of generative-AI innovation/hype has driven a wave of AI ‘thought leadership’ on labour market concerns. Broadly though this reflects a few repeating patterns:
- Same old hobby-horse, for the generalist commentator (and politician!) any new innovation is an excuse to talk about how this new thing is actually an urgent need for their preferred politics. Self-driving cars? Self-serve kiosks at fast-food restaurants? Generative-AI? These are all pressing cases for strengthening labour laws / redistribution / unions / universal basic income / etc.
- Hubris & rent-seeking, for the innovators at the leading edge their success is proof of their genius and if they were right about X then they are also right about Y. All the better if Y also happens to protect their competitive position. OpenAI pitching for regulation of AI is basically a canonical example of this.
The thing is throughout history we have been very bad at predicting the impacts of new technology and there’s no reason to think we are getting better. Best to stay out of the way (and if you care about redistribution then just make that argument).
Ugh - capitalism, excellent article and recommended read, as I’ve quipped to friends ‘the best thing about late stage capitalism is it just keeps getting later’.
Farm the Ocean, I’m a sucker for geoengineering and manipulating the oceans are one of the most attractive options simply because it’s relatively cheap. While at the moment it is notionally banned I think this will end up happening simply because it’s cheap/easy enough for a developing countries (or sufficiently rich/motivated individual) to give it a go.
Strauss Vindicated? Hints of Wrongdoing in the Titles of Books About Dishonesty, pretty fun story about how a group of researchers published a whole set of books about dishonesty and it subsequently turned out that their research was fraudulent.
Your Book Review: Public Citizens, another great book review this time looking at the pivotal role Ralph Nader played in creating the culture of litigation and bureaucracy which bogs down government.
The Rest is History: Watergate (Part 1 & Part 2), great listen on Nixon and Watergate. In the end I was surprised how sympathetic I was to Nixon, from the end of the podcast:
When you look at Nixon, you kind of look you see yourself. All his anxieties are the anxieties of kind of masculinity, that there are people having more fun than you, that the girls aren't looking at you. You've got the wrong shoes. You don't really know how to behave. You don't know what difference between Burgundy and Bordeaux. You've worked harder than anybody else, and yet they will always look down on you and sneer at you.
Housing vacancy rates (pdf) via Noahpinion, returning to a familiar theme - the number of houses will approximate the number of households because a house must exist before a household can form. Therefore someone claiming that there are enough houses for current households is missing the point. For the housing market to function efficiently there must also always be some vacant houses - just like the job market. If there were no vacant jobs (or no unemployed people) the market wouldn’t work very well. I don’t have a an answer on the optimal level of vacancy but I think we could pretty convincingly say the UK has nowhere near enough and Spain has too much.