The Cold Start Problem 5/7
Hockey stick growth is often accompanied by another hockey stick – of the costs of running a business (whatever the growth drivers are, but often – hosting and bandwidth).
Growth teams are looking at the key metrics (the usual – acquisition, engagement, etc.) and identifying user segments with the high preferred activity and low (and possibly a bunch of segments in between). If a company can reliably acquire high-value clients and let its competitors chase the low-value targets, this creates a nice competitive advantage. Low-value clients may cost the business more than they will ever bring. As a consequence, not everything people do on a platform should be facilitated even further: some activities, whatever their engagement level is, don’t contribute to the productive (monetizable) outcomes.
Being able to identify members of a dense network and understand what brings them together (can be as simple as working for the same company as is evidenced by their email addresses) helps make a bigger sale or keep the client company longer.
“Escape velocity” - the focus is on maintaining a fast growth rate and amplifying the successful network effects.
18/ The Trio of Forces
Scaling a network is very hard: in addition to growing (which isn’t straightforward, either), the company has to fight the opposing forces, including government regulations, competition and market saturation [MK: it always feels like the ceiling is 3 months away].
The network effect is actually a mix of the three well-known ones: Acquisition, Engagement and Monetisation (the author uses the term “Economic effect”).
Acquisition in the networked products requires viral distribution / customer sign-up so that there’s limited reliance on paid advertising in a way that a paid sign-up results in some “free” ones, too. This is done via referrals (it’s straightforward and rewarding to invite other users) and improving conversion rates for the key sign-up metrics. All this reduces the CAC (customer acquisition cost).
Engagement means higher stickiness and an increased use of the key features (linked to the value the network brings to its users). Engagement improves with the quality of content consumption driven by the emergence of celebrities and opinion leads on the platform. Effectively, high engagement rate improves the retention curve, which is the #1 concern for all products with users.
Monetisation (economic effect) is the ability of the product to improve its economics via increasing customer monetisation or optimising costs. Offering new features (and charging for them accordingly) is also an option. In app stores the more apps there are – the higher the monetisation of each user is (this is not true for the sellers, though).
Growth rate is the indication of the ability to consistently grow the mix of these three forces over time. That’s where the good old subscriber base modelling comes in handy. [MK: the author calls is the “Growth Accounting Function”, meaning the two formulas below, but yours truly was doing it in Excel 20 years ago, so this is not new news.]
Gain/loss in active users = New + Reactivated – Churned
Actives (t) = Actives (t-1) + gain/loss. t can stand for a month, a week, etc., whatever the internal reporting period for the product team is.
Revenue (t) = Actives (t) * ARPU (t).
All these parameters need to be looked at separately and it’s quite common to have separate teams looking at each of them individually.
19/ The Engagement Effect
The author talks at length about the A/B tests on cohorts. I guess, it’s fair to skip this.
The sad truth of pretty much all application is that people don’t stick. [MK: my personal experience 13 years ago was when me and my co-founder built a service for making a simple home budget for people lacking financial discipline. The service was so effective that most people churned within 30 days after having solved their immediate problems.]
70% of app users (Google Play) don’t open a newly downloaded app the next day, and 96% of them churn out in the first 3 months. This is depressing, but if this statistics applies to you, too, at least you’re not alone.
A16Z suggests the following benchmarks: 60% retention after day 1 (RR1), 30% after day 7 (RR7) and 15% RR30. Only the networked products can improve on these numbers, because they are supposed to become stickier over time. In rare cases users can massively reactivate.
[MK: if you remember Marketing 101, there are 3 growth drivers: Users, Uses and Usage. This will be important here.]
Engagement relies on the emergence of new use cases for the product (“Uses”). In the case of Slack, this means new channels are being created for convenience reasons.
Increasing “Usage” is far from straightforward, but still needs to be done on a product level, focussing on improving certain engagement metrics (not necessarily ARPU). Relevant users should be targeted with incentives or messaging so that they explore new use cases and increase their usage. Frequency, lifetime value, certain use cases can all be optimised for.
Moving people from a low-usage to a higher-usage segment requires a certain (usually a segment-specific) lever. The trick is finding such levers. [MK: at the very least, discovering, increasing, and maximising value are three different levers and apply to different segments.] Analysis of the behaviours of the power users is helpful, but can be misleading for modelling the usage by the lower-use segments. That’s why A/B tests are important.
Getting users to explore new use cases requires education (how-to materials), social proof (“people like you are doing X”) and incentives (giving something of value – usually in the form of subscription - for completing certain actions). Product companies run lots of such experiments.
An engagement loop is the process of making a product more valuable over time: users derive value from others in a network. In a social / communications / entertainment products this looks like this: a creator publishes their content, their followers engage with it via likes, shares and comments, and this is the payoff. For marketplaces this is the process of putting a product in front of a wide audience. Each of these loops consists of a number of steps, and improving these steps benefits all downstream actions. A sparse network doesn’t lend itself to the emergence of such engagement loops; the lack of a meaningful number of such loops contributes to user churn. Increasing engagement builds trust in the networked product. Growth requires accelerating these loops by making it easier to perform a desired action (upload content, bring a marketplace listing in front of a large audience, buy with one click, etc.). The existential question becomes “how can we build a dense enough network to enhance the loop?”.
User reactivation is every product manager’s dream as it allows tapping into an audience of people who already have made the first step of signing up / downloading an app at some point. Traditional products don’t have a big toolbox of things they can rely on: spammy emails, discounts and pushes for some made-up reasons – all of them have a very low response rate and a high irritation rate. Networked products have a distinctive advantage of being able to reactivate a user by having another user ping them by sending a message, tagging, or sharing a piece of content with them.
If inactive users have a certain number of active users as their first connection, there’s a high chance of reactivating and even increasing the initial level of engagement. Some contact is better than no contact, and even a summary of other users’ activity sent weekly may cause the user to reactivate partially thanks to the social proof.
20/ The Acquisition Effect
The Paypal Mafia has made “buzz” into a science. They were searching (and finding!) the scalable ways of bringing the product to large audiences. Paypal specifically used eBay as a locomotive [MK: or a “donor”, as we say at FunCorp] for its growth.
Viral growth is essential to young companies that churn users at a high pace, too, and the only way for them to stay afloat is to add new users even faster. Without the viral (free) user growth any paid acquisition is arbitrage (buy a user for $X, monetise for $Y, hoping that Y>X) in the best scenario and a waste of money in all other scenarios.
Networked products can embed their viral growth into the product experience itself. It’s the ability to invite other users for the purpose of sharing something with them: a network or a connection, a piece of content, a specific activity (collaboratively editing a document). Such products can use an underlying network of contacts – a phone address book (arguably the most abused growth driver) or a corporate employee directory.
It’s hard to overestimate the importance of an onboarding process – the more streamlined it is (fewer screens / steps / data inputs), the more effective it is. Needless to say, everything needs to be measured and A/B tested.
The viral factor is the # of users referred by a cohort divided by the size of the cohort (cohorts stack up, so the factor of 0.5 doubles the cohort size: (x + 0.5x + 0.25x + …. = 2x). The product manager’s dream is having the viral factor of 1+ when the user base growth is supposedly self-perpetuated and unconstrained. (Not really, because of market saturation and changing user demographics.)
Retention is the strongest lever of the viral growth. The longer a person stays active on a platform, the more sharing/invitation actions they perform over their lifetime. User psychology plays a more important role in understanding why people stay on a platform than Excel skills. Having a working understanding of what drives each user segment and a toolbox of things to try (based on the past experiments and learnings from them) helps to slowly but gradually improve retention.
The psychological elements specific to a product or product category make the viral loops hard to copy. This is markedly different from traditional (even online) advertising that can be bought by almost anyone.
Acquisition is separate from Engagement, but without Engagement it’s of little value (unless there is an arbitrage opportunity, which usually doesn’t last long). Or a product may degenerate into a pyramid or a Ponzi scheme, collapsing when the inflow of new users dries up.
It’s very helpful to understand why and how user groups within networks invite other user groups. The term is “land and expand”: build new networks and increase the density of existing networks. Networks built through organic growth tend to be more dense and engaged than the ones that are built via paid and “pushy” (think of Google launching its infamous Google+ service) channels. The difference lies in the engagement, which defines the retention and the monetisation.
21/ The Economic Effect
Data network effects – the ability to better understand the value and the costs of a customer as the network grows larger. That’s how it becomes clearer how to employ incentives, subsidies, and promotions and grow revenue by improving monetisation and/or increasing conversion rates. If used right, more data means higher efficiency of decision making.
Launching a new network often requires subsidising the hard side, like making up-front payments to content creators and influencers to participate on the platform. [MK: It’s ironic that subsidies without a clear business case (i.e., a clear understanding when to stop or reduce them) can only exist in the modern economy where there’s no pressure on companies to reach profitability.]
One of the goals of subsidies is increasing the density of the network so that the increased usage numbers (i.e., resource utilisation in a broader sense) make the business case positive. [MK: The underlying assumption is that at the reasonable level of asset / service utilisation the service CAN be profitable not just on a unit economics basis, but also when taking fixed costs into account, too.]
Higher profitability may not be the end in itself: it can be used to further subsidise the hard side or invest into the easy side’s wellbeing. Lower prices mean more engagement, so good product economics creates an option to postpone milking the audience in favour of a longer customer lifetime. [MK: a curse of lead-generation or advertising-funded businesses that don’t deal with payments is the inability to resort to discounts / incentives as a retention tool.]
Almost every large marketplace is based on an underutilised asset (officially - real estate for Airbnb, but in reality the largest earners are networks of managed-for-Airbnb properties). As the network grows larger, idle assets are monetised better due to variety, proximity and availability.
More data means subsidies and incentives can be personalised (meaning that companies will stop overpaying underperforming members of the hard side). Customers can be targeted for upsells, and the data will determine the best way and time to reach them.
Conversion rate often lies at the heart of a networked product: increasing the number of transactions on a platform brings higher commissions volumes, increasing the number of paid users creates higher recurring revenues. For networked products conversion rates should go up as the network grows. Product managers need to be on a lookout for the features that will benefit everyone making it a strong driver to convert or upgrade.
When it comes to marketplaces, both sellers and users create additional information making it easier to make an informed decision about a purchase thus effectively increasing the conversion rate from a browing session to a purchase. Some social platforms monetise by selling status (e.g., Tinder), which the status upgrade is only worth something when there’s a sizeable community to stand out from. This applies to in-app purchases to customise one’s character – also a status flaunting exercise in a way.
The Economic Effect creates a defence against an underfunded upstart as an incumbent can spend more money on retaining customers. [MK: in the travel world booking.com has an ultimate advantage over everyone else because its commissions are the highest, meaning that they can spend more money on the acquisition and retention activities.] Never underestimate profitability and access to capital.
Networked products with a large customer base are more or less immune from price competition because of the high switching costs (and/or time spent on the migration). Additionally, it’s not too hard to copy a product, but it’s close to impossible to copy its network and a dataset on which its ML (machine learning) algorithms are built.
Consumers are often scared of the platforms’ pricing power as they correctly believe that at the end of the day it’s them who are paying all bills. However, on the bright side, higher conversion rates and commissions allow for the emergence of independent creators and sellers who earn all their living off these platforms.
Thus, premium pricing is a powerful tool and a nice goal for a networked product to set.