The Cold Start Problem 4/7
11/ Tinder
If the scaling process becomes repeatable – this is the tipping point. Building an atomic network or two or ten is not an indication of scalability - yet.
Tinder has lots of complexities: it needs to be hyperlocal, cater to different age groups and grow faster than the natural churn occurs (people find their long-term match and leave the app).
There are several strategies to reach the tipping point where growth happens “just by itself”: “invite-only” (LinkedIn, Clubhouse), “come for the tool, stay for the network” (Dropbox, Instagram), “paying up for launch” (Groupon), “flintstoning” (Reddit), “always be hustlin’” (Uber).
12/ Invite-Only
The obvious explanation for the “invite-only” approach is the ability to scale the product more smoothly as well as to increase the perceived value of the product by creating artificial scarcity. The not-so-obvious explanation is that this method allows creating an engaged self-selected audience capable of self-replicating due to its density. One’s address book is the initial source of like-minded individuals filling in the dense network.
Invite-only mechanics provide a better “welcome experience” for new users (a direct quote). Every new user is already connected to their inviter. Also, better connected people get invited earlier (because … more connections) and tend to invite other better-connected people. The desired outcome is the curated user base creating a positive and engaging welcome experience. [MK: I’d say, the first 100 hours in the product.]
An unintended consequence of being an early adopter is the ability to claim a desired username on the platform – a status possession.
Invite-only mechanics is not risk- or investment-free as it does limit the user base growth as well as requires a conscious investment into the onboarding of new users who were invited (vs “walk-ins”). User numbers should be watched closely to switch to Plan B (open the doors to everyone) if needed.
Handpicking top users to rate an app (and obviously give 5/5 stars) can create social proof for other users who are undecided.
“Waitlist” is another incarnation of the “invite-only” mechanics; it’s simple: sign up for a service and engage with others (usually in the form of product advertising) to get ahead in the line. Alternatively, filling in a questionnaire may also help to skip the line if a user looks like a good candidate to be relevant and engaged.
13/ Come for the Tool, Stay for the Network (CTSN)
A popular use case: start with a “tool” – a product experience having value in itself and then supplement this individual experience with networked effects – collaboration, sharing, communication, etc. (Instagram, Google Suite) Eventually the network comes as important as the tool.
MK: Instagram’s growth strategy included sharing back links to Facebook, which allowed other startups (e.g., Zynga) to use it as a locomotive longer than reasonably needed. Current locomotives don’t do it for free anymore.
CTSN creates a network out of the installed base and makes the required atomic network size smaller, while allowing users to reach the entire network with their content or advertising. At the same time, if worse comes to worst, the atomic network can consist of just the user enjoying their software.
Not every network is built this way, as there needs to be an underlying tool that creates value by itself. There can’t be single-player social networks or marketplaces. However, if there’s a creation [MK: probably of rich content, i.e., not text] involved, a tool that can produce a piece of content or organise a bunch of them on a user’s device is a potential candidate. Also, interacting with content should mean interacting with people.
The following CTSN clusters emerge (quoting from the book):
Create + Share (Instagram, YouTube, G Suite, LinkedIn)
Organise + Collaborate (Pinterest, Asana, Dropbox)
System of record + Keep up to date with others (OpenTable, GitHub)
Look up + Contribute (Zillow, Glassdoor, Yelp)
Tools should offer some new (or borrowed from other industries) approaches (like Trello using Kanban boards) to quickly create, organise, access, and search for information. [MK: The social part is the ability to contribute to this information in a way of adding new, updating and correcting existing details, and creating links between content pieces that didn’t exist before (and help train the AI, of course).]
Not every feature can become social or become so useful that people would shift their behaviours to stick with the tool. If the network looks artificially placed on top of the tool as an afterthought – it definitely is.
14/ Paying Up for Launch
Networked products will slow their growth down substantially (almost irrecoverably) if they start focusing on profitability during reaching their Tipping Point. This may mean subsidising the product to users (making it cheaper or free).
In the Uber case (with guaranteed payments to drivers) as well as many others it’s very easy to burn through a lot of cash thrown at subsidising the hard side. Whether this money can be earned back or not is a big question. If yes – this is the Tipping Point right there.
Referral programs are another form of subsidies: they come in all shapes and forms, and in the case of Uber (for both drivers and passengers) they looked like “invite a friend and you will both get $X if they perform a desired action”. (This is in addition to the coveted “word of mouth” distribution that’s supposedly free.)
Subsidising the hard side works for attracting and retaining content creators (Netflix, YouTube, TikTok, Twitch). In B2B offering free or freemium versions of the software help recruit future paying users. [MK: there’s a debate whether the costs of servicing free users should be treated as marketing costs or spread out as the general costs of running a business.]
Needless to say, it’s a tempting and at the same time – a very dangerous move for startups to scale via subsidies, and more startups have run out of money due to this than they care to admit. Scaling with money should only be done after a product-market fit has been found and atomic networks start emerging. Selling the dollar for 90 cents should only be done to grow the business hoping to milk it later, but not to keep the lights on.
Not all products handle money flows directly, so it’s not always possible to seamlessly introduce direct financial subsidies into the offering. But when they do – this becomes a powerful instrument into the company’s toolbox.
MK: Companies can issue their own cryptocurrencies to reward creators, but this is bordering socialism, and I have no interest this this mass delusion.
When startups choose to invest time (the most valuable resource) into dealing with large customers hoping to get a cut of a large pie, this is also a risky investment that usually doesn’t pay off (but when it does – it becomes truly newsworthy).
Historically, at least in the initial stages growing a business unprofitably and later making it profitable [MK: many companies never reach this stage, which doesn’t stop them from being valued at $10B+] is more reliable than reinvesting profits (which may or may not be ephemeral). If the company chooses to subsidise consumers, this (in theory) helps strengthen the “low prices” positioning long enough for people to start believing it.
Once the market is won, incentives should be scaled back or eliminated. There may not be a need to pay for the same customer many times in order to win them from a competitor. The dominant belief that winning a market will absolve companies from all their money wasting sins.
15/ Flintstoning
It’s a process of replacing the missing product functionality with manual labour until the time comes or when the feature is considered not needed. This applies broadly to the initial selection of content on the platform (e.g., YouTube founders uploaded content early on). The goal is finding out what the product is missing to form its first atomic network(s), prioritising and implementing the required functionality and then to focus on the growth.
This also applies to creating dummy user / supplier accounts to create an illusion of the balance between a chicken and an egg. Automation + scraping of content is as needed for new products in 2021 as it was in 2000. Automation + manual labour (in addition to the activities one would normally expect a company to perform) help solve the hard side acquisition and engagement problem.
MK: If one looks at the “digital transformation” of the traditional industries, one can see a bunch of employees using pen and paper (and maybe a calculator, but no guarantee of that) as a starting point, with software slowly transforming the process into a more automated one, forcing people to work for Amazon warehouses. Welcome to B2B.
“Cyborg” startups are the ones where a user thinks they interact with software, but certain (if not all) functions are in fact performed by live human beings. Cheap labour can do miracles and for a while be as cost-competitive as automated tools.
“Do it yourself” (build + buy the builder) is an expensive strategy that may work very well if the content is unique and is an integral part of the acquirer’s value proposition. Think of Netflix producing its own shows.
Exiting the “flintstoning” strategy may mean either or both of the following (both applying to the hard side): reducing and eliminating automated content generation, uploading and moderation, or increasing real (not assisted or bot-driven) user participation. Proper (user-driven) product metrics must emerge for the service to be useful, self-sustainable and not depressing to organic content creators who can be driven away by automation.
16/ Always be Hustlin’
Uber: while product people can solve any problems, operations can do it faster.
Hustling means exploiting temporary pockets of opportunities (growth hacks, in a way). It’s a neverending row of activities (however small) that generate publicity and help solve the hard side problems.
MK: I’ve always said that scale helps build a toolbox, and teams should be aware that the tools exist and trained on using these tools. Toolboxes are expensive (since time is more expensive than money) but help build a sustainable competitive advantage.
With B2B hustling comes in the form of tapping one’s professional network and getting the first customers. Hence, the ability to build one’s network or at least knowing where would-be customers spend their time is a valuable skill. Consumerisation of B2B marketing helps with lowering the barriers to try a product.
Paul Graham: don’t wait for users to come to you, do things that don’t scale and start with talking to customers directly. Shyness and laziness have to be overcome.
In B2B consulting for the first customers in order to reach the product-market fit faster (it’s important to retain IP ownership) is a valuable, but a non-scalable way to grow. Still worth a try up until a certain scale is reached.
The initial network may pull the service and the company into a grey area where the legality of actions is not guaranteed or obvious. UGC borders copyright infringement, payments border money laundering and terrorism sponsorship, content sharing borders piracy. Over time two things occur at the same time: services add moderation to reduce the exposure to the grey area (the actual risk appetite is in the hands of the CEO and the Board of Directors), and consumers and suppliers (often – copyright owners, but also legislators) become more welcoming.
Growing first (even being exposed to the grey areas) and slowly legitimising operations seems to be the only workable way to succeed. (It’s this old “Ask for forgiveness, not permission” principle in action.)
Part 5.