Case Study: How Pinterest Scaled
Pinterest is a company that most of you are familiar with now. It grew from having 0 page views to billions page views a month in two years. But how did it get here. This article tries to answer that question by surveying a list of articles written by other people and present the findings here.
The list of articles surveyed:
- Steve Blank’s “Scaling Pinterest — From 0 To 10s Of Billions Of Page Views A Month In Two Years”
- Jon Tan’s “How Pinterest Grew From 3,000 to 73 Million Users”
- Justin Scott from GrowthDevil “How did Pinterest get so big so quickly”
- Sarah Tavel’s “Five Lessons from Scaling Pinterest”

Summary of Blank’s article
Blank’s article primarily speaks about the scaling effort from a technological stack perspective. It is full of amazing details. His talk itself is derived from the lecture from Pinterest’s Yashwanth Nelapati and Marty Weiner. On the high level, Blank recommends:
- Architecture is right when growth can be handled by linearly adding more of the same thing
- Evaluating a tech stack should be done at pushing tech to the limit and observe how they fail. Their failure pattern is a great source of information in choosing tech stacks.
The basic Pinterest product architecture — Blank’s description is by far the simplest and intuitive:
- Pins are image with collection of info (description of what makes it important to the user and link to where they found it)
- Pinterest is a social network — you can follow other people and their boards
- Database: Users have boards and boards have pins; follow and repin relationships; authentication information
Blank characterized Pinterest’s growth till 2012 in four stages:
- The age of finding yourself (03/2010–01/2011) — tech backend is very immature and a lot of unpractical things
- The age of experimentation (02/2011–09/2011) — this is the time when Pinterest doubled every month and half. At this time, Pinterest as a product was broken every night and week. Pinterest tries to alleviate the problem by trying a lot of different technology and they all break — this is the time when a massive re-architecting was in order
- The age of maturity (10/2011–02/2012) — at this time, Pinterest is stable. The backend is simple with focus on MySQL, Redis, Memcache and Solr. The traffic to Pinterest’s site still grew at a steady pace.
- The age of return (03/2012–10/2012) — Pinterest grew 4 times its size at the beginning of the year. Architecture remains the same as it was at point 3. That’s the fundamental characteristic of good architecture.

Summary of Jon Tan’s Article
Tan’s article tries to summarize the scaling of Pinterest from a business standpoint. He came up with the following 4 key dimensions in explaining Pinterest’s growth:
- Word of mouth — when Pinterest was stuck at 3000 users, Silbermann (Pinterest founder) resolved to find power users who love Pinterest as much as he himself. Key tactics included meetups at local boutiques, inviting and transform invitees into product evangelists. This is effective because the online experience is augmented through active offline communities
- Reach the entire communities by targeting community leaders — it ran a campaign to get passionate bloggers to publicize Pinterest. These events were invite only.
- Pinterest Users are proud of their boards — each person’s board is a carefully created content, an investment on the user’s part and each is very authentic in their own way.
- Pinterest as a marketing channel — It was the 3rd largest source of referral traffic. When big brands use Pinterest, it drives more engagement.
The first point was valid but the article wasn’t very in-depth nor insightful. These four dimensions are certainly not enough in describing Pinterest’s growth. NEXT
Summary of Justin Scott’s Article
This article recounted of the founding story of Pinterest — described Silbermann’s passion for collection. It tries to explain Pinterest’s growth through quoting its use of growth-hacks. Such as the following:
- In Pinterest’s early signup process, users had to “apply”/request an invite to join — this helped with building the image of close-knit small community.
- During the onboarding process, Pinterest auto-populates contents based on users’ response in terms of contents that appeal to users.
- Pinterest as a product was easy — it was a one page application where users don’t need to go through too many pages and effortlessly go through its innovative infinite scroll
- Email notification became an important hook for Pinterest in keep engaging the users.
Looking at this article, it had an interesting narrative that tries to summarize Pinterest’s growth. Though informative, many of these tactics utilized by Pinterest has become table stakes. It may have been truly effective at the time but given the ubiquity of each of these tactics. I can hardly call these four things gleaned from that article useful. Again, doubt that it was effective.
NEXT
Summary of Sarah Tavel’s article (Formerly Product at Pinterest)
Tavel is currently a GP at Benchmark and she was the product person at Pinterest. Her perspective should be a lot more informative than the two mentioned above. The article is again a summary of lessons she learned from Pinterest. Though she wasn’t disclosing the cause to Pinterest’s growth, she explains properly the important lessons she learned.
- What’s measured matters. It was vogue for measuring MAUs (Monthly active users). Pinterest’s growth team grew MAUs effectively but the problem was different. The product was a leaky bucket — no one cared about new users’ engagement. They shifted gears to measuring instead the NWAU (new weekly active users) — this helped with overall engagement.
- Org chart matters. In most execution related problems, two reasons are to be blamed — wrong org structure + wrong person in the job. Tavel quoted two examples in explaining wrong org structure would cause inefficiencies in critical functions. This reminds of a Podcast I listened to a while back. Org design is essentially a resource allocation problem but with a caveat — instead of allocating resource, it allocates communication overhead. When more people work together, it adds communication overhead. This is inevitable. A good org design is one where this communication overhead is allocated to the least strategic functions.
- Listen to users but know when to ignore them. Remember you are building for the next 100m users. The loudest users aren’t representative of the population nor the next 100m users. They create 2 biases — they are used to how things are and they request power-user only features (these increase the complexity of the product for everyone but are only used by a small number of users). To address the first bias, company should disrupt themselves. To address the second, the feature is only built if more than 5% of the users use it or it has to be a real game changer for that less than 5% of the users.
- Think of user trust like a bank. Should deposit a lot more than you withdraw. It takes 5 positive experiences to make up 1 negative experience. Positive experience at Pinterest was shaped in the form of a delightful product experience.
- It’s not always up and to the right. And it is okay. Pinterest stuck at 30k users for a while before taking off.

Conclusion
Both Blank’s and Tavel’s article are very insightful but just these four articles weren’t enough to understand how Pinterest grew. When I find more scaling Pinterest related interesting information, I will edit and compile them here.
