LITSO: Medium – The Stream

In the first edition of LITSO – Life In The Shoes Of – posts, I’ll talk about the Stream functionality that was introduced by Medium early 2015. In a previous post, I’ve described the context in which Medium operates.

What Medium cares about

User engagement metrics

• MAU / DAU
• Number of articles read per user
• Total Time Reading
• Connections on Medium
• Discussions
• Churn
• Mean time between visits

Monetization metrics

• Monetized pageviews
• Subscription and partnership based revenue
• LTV of user
• Cost to serve page
• Acquisition costs

Top of funnel metrics

• NPS
• Organic vs inorganic traffic acquisition / sourcing

Detailed metrics on user understanding and content quality

• Spam and fraud control
• Clickbait articles control
• Paywall hit rate for non-paying user
• Interest matching
• Content discovery ease
• User patterns like distribution of article read frequency vs article read duration, category etc.

This is not an exhaustive list. Generic SaaS business metrics can be found here. This explains the unit economics of a generic system.

Unit economics. Taken from www.forentrepreneurs.com/saas-metrics

The Stream

The stream is an infinitely long sequence of posts sampled from the universe of posts that are of interest to the user with a certain ordering, usually algorithmically decided. It’s a natural way of controlling the content that is surfaced to the user, employed with great success by the likes of Twitter and Facebook.

Value proposition

To the user, Medium can surface an infinite number of relevant articles, ensuring that the user never has to take another action to get additional articles. Personalization of the order of articles according to the user’s preference enhances user engagement.

The stream format works to Medium’s advantage as it gives additional control over content shown, that can be optimized for multiple objectives – user engagement, content discovery, connections, and monetization.

The current experience also enables recommend and bookmark actions to be taken within the stream – for when users make a quick visit to the site or choose to read the complete article later. Also, the short posts are shown in their entirety while the longer ones are summarized in the stream providing an easy reading experience.

Key metrics for the stream

Given the objectives of the stream, the metrics that would matter are –

• Number of articles seen per user – indicates enhanced discoverability
• Number of articles read per user – user satisfaction with the content surfaced
• TTR – aggregate and per user
• NPS – indicator of long term value add to users
• Signups and churn – as guardrail metrics, ensure no regression
• Revenue – for ensuring sustainability
• Distribution of article read frequency vs article read duration – to understand user characteristics and preference
• Recommend and bookmark actions taken from within the stream

The most interesting part about this feature is that the feature specific metrics and the platform metrics are nearly identical. This is an indicator of a feature that aligns directly with the goals of the business and is a must-have and not just an optimization that’s a nice-to-have.

Note that the per user TTR might dip if the articles that are surfaced to users are shorter in nature. However, we would still want the aggregate TTR for the entire network to go up. A second level of understanding might be necessary if the TTR per user drops.

Discussions, recommendations and post-article actions are secondary. LTV is not a metric of concern as there could be a lot of low value users who join the system decreasing the average LTV but contributing positively to the revenues, considering cost to serve.

Great! What do we do next?

There are two obvious opportunities that the stream opens up –

• Content posts – similar / similar-sponsored articles
• Connection suggestions – with publications, brands and people

Medium is growing and quickly. At this stage of the game, I would rather invest in building a strong foundation of user experience and value than focus on monetization to ride the wave. Strengthening the connections and the networks between people and publications – that is what makes the service sticky.

As I had mentioned in the previous post, Medium is a strong content farm with a key asset being its large network of users, giving it an edge over competing offerings. The stream is a natural place to surface suggestions to the user.This would lead to better user acquisition, better user satisfaction and reduce churn because of increased revisit value as updates from a larger number of people would lead to more content of interest.

An example experience could look like this –

It all makes sense. How could this fail?

The stream is, intuitively, an obvious value add to the user while bettering the business KPI. However, there are quite a few nuances that could be unaccounted for in the execution of the feature that might lead to the gating criteria to be unmet.

Some interpretations of regressed metrics are –

 

• Number of articles seen per user – ordering of the content is awry
• Number of articles read per user – interest mismatch
• Distribution of article read frequency vs article read duration – a skew to the shorter read duration indicates a combination
• Aggregate TTR – to be seen in combination with number of articles read per user. Could mean user preference for shorter articles or interest matching issues
• NPS, churn – a side effect of the user experience design, or the above mentioned factors or both. A cohort study will reveal more insights.
• Signups – lack of increase indicates user’s perceived worth of personalization of article recommendations
• Revenue – due to fewer monetized pageviews per user or user churn.

Based on the regressed metrics and the associated top-level hypothesis, a data-driven approach can be employed to understand the reasons for the success or the failure of the stream.

This feature has been going strong on the Medium website for a year now, we should see some monetization built into it soon.

The curious case of Medium

Medium is a social idea-sharing network, where the content is primarily in long written form. The positioning of medium.com is interesting in that it places strong emphasis on the quality of content. The editors did a great job of it in its seed phase where they had contributions from the likes of leading businessmen like Elon Musk and authors like Walter Isaacson before opening it up to the public at large.

It has a global Alexa rank of 389 with 34% of the visitors coming from the US and 17% from India, followed by Brazil and UK at ~3% each, implying a mostly English user and content base.

According to crunchbase, Medium has received $82M in funding in two rounds, from 20 investors, led by Greylock Partners and A16Z.

The platform has a rather unique value prop that requires alternate business models.

To skip to the part where I describe the importance of the Stream that Medium introduced last year, click here.

The stakeholders and Medium’s value proposition

Medium’s about page has this to say about itself –

Medium is a different kind of place to read and write on the internet. A place where the measure of success isn’t views, but viewpoints. Where the quality of the idea matters, not the author’s qualifications. A place where conversation pushes ideas forward and words still matter.

The foremost are the content creators who come in two flavours – individual writers and Publications. What Medium refers to as Publications are cohesively themed articles written by a group of people.

Discovery: Medium drives discovery of the high value content for the writers. It has elements borrowed from aggregator sites like Reddit in the form of upvoting and the Facebook/Twitter elements like the stream, related posts, metadata tags, recommendations, and author based clustering.

Partnerships: Specifically for Publications, there are partnerships in place to drive brand content marketing ¹. However, there isn’t any requirement to be restricted to Publications.

High Value Content: Medium’s offering to the website visitor. Special care has been taken in building out experiences for both the contributors as well as the readers to ensure that there are as few distractions as possible and that the focus is on content creation. This is exemplified in the simple uniform theme and the editor.

Business model

Publishers can monetize their content using one or more of –

• Paywalls and  subscriptions
• Brand content sponsorship
• Ads and native ads

Ben Thompson wrote a brilliant article on the state of the publishing industry. To quote –

First and foremost that means publishers need to answer the most fundamental question required of any enterprise: are they a niche or scale business?

• Niche businesses make money by maximizing revenue per user on a (relatively) small user base
• Scale businesses make money by maximizing the number of users they reach

The truth is most publications are trying to do a little bit of everything: gain more revenue per user here, reach more users over there. However, unless you’re the New York Times (and even then it’s questionable), trying to do everything is a recipe for failing at everything; these two strategies require different revenue models, different journalistic focuses, and even different presentation styles

Medium is a niche publisher reaching into scale territory. This is plenty obvious from them opening up authoring articles to the public and the inline editor to enable quick posts.

Given Medium’s focus on high value content, traditional ads are out of the question. This leaves native ads as the only mechanism to monetize at scale. Interestingly, this ties in very neatly with brand content sponsorship on the website (re:form etc.) and the website philosophy.

Within this expansive publishing surface, Medium Publications serve as the niche focal points, capable of having subscription based monetization and forging a unique character to the offering.

The native ad mechanism itself is self-sustaining in this context because of the aligned incentives of the publisher and the advertiser to deliver quality content. Every writer can promote their post and can effectively be an advertiser too, while earning through other writers’ content links.

The social network structure and discovery is what sets Medium apart from other sites like Blogger and WordPress in functionality. While Twitter and Facebook Notes have the network advantage with a larger user base, these websites cannot afford an experience to rival Medium’s which puts content front and center.

However, achieving scale with longform articles is a challenge in itself because of the amount of time required for reading and writing. This is another reason why the social emphasis is key, to make Medium a destination site – a site that users search for and go to directly, and not solely through linkage.

In a follow up post here, I will talk about the Stream that Medium introduced in detail.

 

---

1. Content marketing already sees Medium as a key player.Third party Source1, Source2 of many. ↩

Conversational Commerce, General Intelligence and Personalized Services

As Chris Messina of Uber put it –

Conversational commerce is about delivering convenience, personalization, and decision support while people are on the go, with only partial attention to spare.

Personal assistants with contextual understanding capable of performing an ever increasing set of tasks are becoming commonplace, with startups and large companies alike competing in this nascent space. The investments by major players like FB, Google, Microsoft and Amazon in this space obviates the importance of building out this core capability, especially for horizontal companies, to stay relevant in this upcoming domain.

There are 3 major influences –

• Value proposition to users
• Business model, players and surfaces
• Technological readiness

Value Prop

The attention span of people is reducing ¹ as people take to busier lives and more multitasking. On the other hand, the monthly disposable income of people across the world is increasing and is sizeable. This creates a very cozy spot for services where convenience and user experience can be optimized for ².

The personal assistant is, by definition, personal. She knows your allergies while ordering food and your preferred blend of coffee. And she can do this in the most natural way possible, with natural language understanding and interfaces like speech and conversation.

This is a powerful paradigm that has obvious use cases for all user segments – a working professional to enhance productivity to a kid looking for improving his flirting skills to a granddad who’s trying to figure out how to order a walking stick without having to go to a store. An example use case is –

 

Taken from Second, a personal assistant at https://mysecond.com/

Some ideas to harvest disposable income, in no order of priority are listed below. Note that there need to be a minimum number of things she can do for her to be viable –

• A hardy talking shell for solo travelers and have a medical emergency with a requirement to provide your medical history to the doc (she recognizes your voice and uses a public profile for others)
• Ping the usual Starbucks in the morning just in time for you to go pick up your coffee without waiting in a queue. Wave your smartphone for payment
• Maintain your to-do list with location awareness and to do your laundry
• Listens in on your conversation like a good butler and comes up with a list of things that she can do for you, with your approval
• Can recite a joke and write a song a thousand miles long, without hanging you out to dry
• Can replace a lawyer for simpler things like filling up forms and appealing parking tickets
• Order flowers, get your insurance, schedule medical checkups, book tickets, get groceries, pay your bills and the list goes on

http://www.kleargear.com/1474.html

While this list focuses heavily on consumers as the end users, there is significant value for the enterprise as well. For example, large chunks of everyday project management can be automated with reminders, predictive warnings for feature slip given burndown trend of the current sprint based on historical individual and team performance and postmortem automation. There is a huge market to be tapped into with this, providing project management solutions at scale to small and large organizations.

Business models

There are three pricing models which are not necessarily mutually exclusive – subscription based, free and pay-per-use.

The subscription based service will come with a guaranteed revenue stream but has a higher quality expectation to meet which might be limited by existing technology, hampering adoption. Scale is everything.

The free service is essential to learn user patterns from larger number of users to improve – both in terms of quality of service and set of actions because of the flywheel effect: more users -> more data -> better products -> more users. I predict that assistants are going to be free for a while before having paid variants ³ in the investment stage along with a beta tag and limited guarantees.

Commission from the feet-on-street service providers for successful conversions providing revenues at scale is the most secure pricing model for all parties. Pay-per-use is a hybrid of these two.

The three things that will make or break an assistant are –

• User reach for businesses
• Service diversity for users
• Interface convenience

All players are on an equal footing when it comes to service diversity and are starting from scratch. Google is starting grocery delivery services and Amazon dispatches plumbers which is a step in the right direction. You will see a lot more of these disparate services popping up before being united under a single banner. To operate at scale, automated onboarding of third party services is key. This lends itself nicely to standardization of the interfaces, either through a consortium or through the emergent dominant player.

The key value to businesses to provide services on the personal assistant platform is the density of users and reach that these platforms provide. The quality of the leads generated is also significantly better for conversions owing to the high degree of personalization and contextual intent understanding.

The key distinguishing factor (unfair advantage ⁴) here will be the surfaces (channels ⁴) for each of these players to provide inline insights and recommendations. FB and Google each have ~3-5 properties with over a billion users that are relevant to this paradigm. Microsoft is a third with half as many properties and half as many users ⁵. Interestingly, China is in the forefront of the intelligent assistant adoption curve. In China, a large marketplace of 3rd party plugins has taken root with the popular social networks and messaging apps. Xiaoice from Microsoft has over 20million users and is the dominant personal assistant in this space.

However, this scenario is very different in the Americas and Europe. FB and Whatsapp with M, Google’s Now, Apple’s Siri and Microsoft’s Cortana are the key players with a ton of Startups picking away at a niche.

In these markets, the user aggregation channels are owned and operated by the personal assistant providers making the company with higher quality surfaces have the key advantage, tilting the battle in favor of Google and FB.

Amazon’s Echo is hugely important in this paradigm for the retailer to stay relevant as it is an always-on natural user interface, but the services need to come first – plumbers, household services, groceries ⁶. The Kindle device wouldn’t have sold without it’s superior inventory. Buying a device has an additional effort-cost to it which is currently unjustified to a user.

Interface convenience is currently largely technologically limited. Google, FB and Microsoft are on an equal footing in terms of investment and research, though Microsoft might have an edge here through Xiaoice. We’ll talk more about this in the next section.

Tech readiness

As Chris Dixon put it in his article –

Each product era can be divided into two phases: 1) the gestation phase, when the new platform is first introduced but is expensive, incomplete, and/or difficult to use, 2) the growth phase, when a new product comes along that solves those problems, kicking off a period of exponential growth.

The Apple II was released in 1977 (and the Altair in 1975), but it was the release of the IBM PC in 1981 that kicked off the PC growth phase.

The internet’s gestation phase took place in the 80s and early 90s when it was mostly a text-based tool used by academia and government. The release of the Mosaic web browser in 1993 started the growth phase, which has continued ever since.

Quite a few breakthroughs have been brewing for 3-4 years now – VR/AR, AI and IoT. Computers are getting better than humans at identifying objects in images, colorize black and white images, and as Xiaoice has shown, conversing with humans giving Ava a run for her money.

Recently, FB has published a research paper ⁷ on how machines can complete sentences using common and proper nouns given context in a paragraph of text.

Big Deal

You wouldn’t outsource wedding planning, would you?

Historically, the ineptitude of companies to keep up with moving efficient frontiers across multiple dimensions has been the single biggest reason for demise (Sears vs Macys/Wal-Mart vs Amazon). This results in obsolescence – of either the company’s business models or value propositions.

Strategically, Amazon is at risk because the marketplace is moving elsewhere – to new surfaces and aggregation points – and Amazon has to fight to stay relevant in this game.

To quote Ben Thompson from Stratechery ⁸ –

So what will bring Amazon down? I’d imagine it won’t be dissimilar to Sears: a dominant strategy over time often disintegrates into the mushy middle, where you’re beat on all of the vectors you used to dominate. For Sears, Nordstrom and Macy’s took the up-market, and Wal-Mart and Target the cost-conscious, and the Internet finished them off.

The upcoming surfaces looking for channels – Slack, Telegram, Snapchat, WeChat – need to tie up with Microsoft and Amazon who are looking for surfaces if they want a piece of this pie.

..and everyone wants a piece of this pie.

 

---

1. Here’s a sensationalized article that speaks of the trend of reducing attention spans. ↩
2. According to this article, 81% of the online shoppers research a product that they wish to buy and, on average, visit 3 stores before making a purchase. A digital assistant can look into the reviews for exactly what you look for and get the right blend of dark chocolate for you. ↩
3. Despite this offering which is subscription based and kicking. There are quite a few startups in India and China which do this for free though. ↩
4. I’m a huge fan of the lean startup canvas but I find it doesn’t work too well in this case because of the breadth of the concept. ↩ ↩
5. This is one of the reasons I think Microsoft should partner more deeply with Twitter, but that’s a story for another day. ↩
6. Amazon Dash was a brilliant product along these lines. I strongly believe that Amazon should give free Dash buttons with every subscription of Prime. More on vertical focused strategies for Amazon in a later post. ↩
7. Incidentally, it happens to be the only AI paper that I know of that cites a children’s book in the references. ↩
8. www.stratechery.com – Huge fan ↩