Bookmark the pictures
VisualizeUs was a social network to help creative people collect and connect with the best visual inspiration around the web.
I launched, ran and led VisualizeUs from idea to side project, then to profitable startup and finally to acquisition (acquired in 2016, currently a different project under the same branding).
Born in 2007, three years before Pinterest launched, it evolved from fun project on the side to growing startup that redefined the image bookmarking space to what it is today.
As a reference site for photographers, designers, and other creative visual lovers, the site hit more than +5 million unique users, +37 million page views, +250k signed up users and +9 million pictures collected, with an average time on site of 7 minutes.
My duties as technical founder of the project involved: product design, strategic vision, team leadership, UI/UX design, brand identity and growth, as well as full-stack development.
In terms of team we got to a small tight-knit team of 2 full-time partners and up to 4 external remote contractors.
In terms of tech, we were able with just 3 servers and a great mix of a battle-tested and disruptive stack (PHP, Python, MySQL, Beanstalk, Sphinx) to give top-notch service up to +5 million unique users and serve +37 million page views.
What I did
- Product design
- UI/UX design
- Graphic design
- Brand identity
- Full-stack engineering
- Strategic vision
- Team leadership
+250,000 registered users
Worldwide (51% US). All organic acquisition, mainly via growth hacking strategies.
+9 million pictures bookmarked
Not a single second since the project started without someone somewhere bookmarking a picture. A complex yet reliable system of microservices powering the backend.
+37 million pageviews per month
At the peak of the project. All graciously handled with a small infrastructure using AWS and auto-scaling features.
¼ million lifetime revenue
Completely bootstrapped, freemium model, with zero external funding.
Some product stuff
(in no special order):
- Ability to bulk edit pictures: add tags to several pictures at once, bulk delete, bulk change of privacy.
- Color browsing: Users could browse all images on the site by a 256 palette. Each picture got their color metadata displayed below.
- Copyright recognition and attribution: using 3rd party APIs from Flickr and other services, we detected and applied the right license and copyright to a picture.
- Bookmarklet and browser extensions able to add more than one picture at once: When activated, the bookmarklet would highlight all recognizable pictures (and videos) on the given page (given big enough dimensions) and clicking on one will launch a “Add a bookmark” popup, without leaving the original site.
- Whole set of features to help with tagging content: Autocompletion of tags, last set of tags used to reuse again while submitting a picture, most frequent tags, ability to order tags by frequency or alphabetically, etc.
- Algorithms to improve the curation of pictures displayed in public listings: Using different social metrics (like popularity, original source, user-rating, etc) we were able to display better images in public listings. Also, a NSFW filter was available, which was powered both by an algorithmic approach and a human reviewal team.
- Three different visualization layouts:
Grid, “waterflow” (Masonry-like flow with infinite scroll) and slideshow. For any listing page, including profiles.
- Social features: Ability to follow (or block) users, comment, like pictures, discover new users by affinity to your content.
- Growth features: “Invite your friends” feature, importing from Flickr, Linkedin, Gmail and other services, badge to show your favorite pictures in any 3rd party site, social loops, etc.
- Retention features: User activation via emails with social activity (“Someone liked your picture”). When leaving for an external site, iframe with other related pictures to keep browsing on the site.
- Full-featured RSS feeds: Including user profiles, any combination of tags, colors, etc.
Some geeky stuff
(in no special order):
- Nginx + PHP-FPM + Memcache + Varnish + APC.
- Queue system to process new bookmarks asynchronously via Beanstalkd.
- Proprietary image recognition algorithm to prevent duplicates images as separate entries.
- Recommendation engine based on clustering, taxonomies and social interactions.
- Browsing by colors.
October 2007 — September 2016
The site was acquired in September 2016 by an international investor. Unfortunately, the current domain doesn’t host the project I worked on and created, just the branding remains.
If you want to see a glimpse of what it was, click below to see an archived version gracefully hosted by Archive.org and their Wayback Machine.