Words

Prompted by this post on Artifact.

I’ve been thinking about this since I heard the news and more so since you posted this.

Identifying the challenges was the easy part, the opportunities…not so much.

CHALLENGES

1. Adult content: It appears that a significant portion of MAU at tumblr’s peak were related to adult content. Verizon curtailed while they owned it which had a led to a huge reduction in MAU. Automattic relaxed things but this wasn’t enough to win these users back. Equally, MAU related to this category of content are highly promiscuous and consequently not sticky.
2. Network effects: Being both web and app, with web probably being the more “core” product, a lot of traffic to individual tumblrs likely came from organic search. Even if steady in aggregate, this is opportunistic traffic and again not sticky. Combined with the pseudonymous nature of tumblr this meant a lack of network effects i.e. tumblr didn’t get better with more people using it and it didn’t really matter if you knew the other people who used it.
3. Monetization: From the documents in the reporting it looks like tumblr users were highly resistant to Automattic’s various approaches to monetization: they weren’t interested in ads (unclear how or what was tried here) nor were they willing to pay for features.

OPPORTUNITIES

1. Mobile app first: there’s a reason that Artifact and Threads launched as mobile apps and that Substack is pushing hard to become more app-centric: far more people have phones than have computers and those that have both use their phones more often and for longer than they use their computers. Phones are the center of people’s digital lives, they connect the physical and digital worlds. What would this mean for tumblr? Even in its current form it is one of the few apps that enables people to build and manage a website from their phone; but the app is extremely limited in terms of the building part, it is mainly a for posting to your tumblr (which you built using the web app on a computer), following content from other tumblrs and a notifications hub. The best-in-class example of a mobile app to *build* a website is Universe (who I’ve long admired). This solves some problems, but creates a whole bunch of others, notably it doesn’t in and of itself do anything to directly address network effect or monetization challenges, but it lays some necessary foundations to do so.
2. Distribution: creators come to a platform for the audience. If the platform doesn’t offer demand, then creators will take their supply elsewhere (it’s not like there’s a shortage of platforms). When Medium was thriving they drove distribution via curation both with their own publications and crowdsourced collections. It proved too expensive and today Medium is something of a ghost town without curation and their algorithmic personalization never got good enough to fill the gap. For tumblr, deals with a group of hand picked creators (read: partners) — both existing and new — to guarantee a minimum amount of content to kickstart distribution and personalization, and thereby model to a long tail of creators what type of content gets distribution, would be taking a page out of the playbook of platforms who’ve done this successfully (i.e. YouTube, Instagram, TikTok). Not a guarantee but at least a direction.

The ratio of creators to consumers on any platform is highly skewed, i.e. even a small group of creators, posting compelling content regularly, is enough to get a flywheel going: the short head of creators (part one) brings consumers in the long tail of creators (part two) keeps consumers there. This means the platform has to put their thumb on the scale either via editorial curation or algorithm personalization.

3. Daily habit: this is the hardest part, getting consumers to come back multiple times every day. Most people will never post, or post very rarely, so the levers to consider are around consumption. Commenting and notifications linked to comments is one obvious example. Affinity graphs are another, more implicit and latent one — an example being why Substack puts the stacks you subscribe to on your profile by default, it’s a vector for discovery, and a badge for users to signal their identity/taste. Chaining posts together by way of related posts, ideally algorithmically with an overlay of collaborative filtering (i.e. people who read/watched this post also read/watched these posts) is also a lever for discovery and engagement. But none of these are silver bullets, it would need all of them and others, with continuous optimization to have at least a small chance of building stickiness.

RISK

There are many risks but the biggest one is that only a tiny portion of existing MAU are interested in the new features you’ve built. You can nail everything but if your existing MAU aren’t interested in, then you might as well have just built an entirely new product, without any of the legacy of the old one.

Acquisitions of existing products — especially consumer products — are tricky, because they’re subject to one of the most powerful cognitive biases: the sunk cost fallacy. The acquiring company is focused on trying to maximize the current MAU at the expense of building something that would help drive growth (for fear of alienating existing users), all while being burdened with an existing (and often, outdated) tech stack.

CODA

This was an interesting thought exercise that prompted me to also think about:

1. Meta’s acquisition playbook

: Instagram and WhatsApp (along with YouTube) remain the gold standard for how to manage acquisitions of consumer products: Acquire early. Migrate infrastructure. Let the existing team keep doing what they’ve been doing (for a while). Don’t push to monetize too early, focus on growth (which you know how to do really well). Experiment with (alternative) monetization. Leverage central services. Plug into existing monetization platform. Drive revenue.

2. Admiration for Artifact: Thinking through the opportunities section increased my (already high) level of admiration for what Artifact is doing. Unburdened with a legacy product, they’ve built, shipped, iterated and optimized *fast*. All three of the opportunities (and many others) are core to what Artifact is doing. They used existing content from recognized publishers (read: partners) to ensure there was content available from day one, so that they could bootstrap distribution and personalization. Comments (with notifications) was a lightweight way to drive engagement and stickiness. Posting turned all users into potential creators, with both rudimentary interest and affinity graphs built there was already a foundation to drive distribution of user generated content (posts), rewarding creators and increasing the cadence and diversity (of relevant) content for consumers. Now that’s a flywheel.