The Everpix project was started around April 2011, gathered speed through the summer and was officially announced at TechCrunch Disrupt San Francisco early September 2011.
Following TechCrunch Disrupt, users were allowed to register on the website and get on a waiting list. Everpix was quite early in its development at this time, and allowing anyone to create an account and import photos would not have been useful. Nevertheless, we had to start allowing some external users, if only to test the system, but also to follow up on a “public promise” made during Disrupt. Around mid-October, we started sending batches of invites from the waiting list and stopped around mid-November (since preparing for the Everpix Public Beta).
The primary goals for Everpix Private Alpha were:
- Get a decent size set of testing users and their photos: this is required to simply test the infrastructure and software.
- Demonstrate we can move a lot of photos from people’s computers seamlessly to the cloud: any service can import photos that are already online, so the difficulty is with offline photos.
- Validate the bet that people would be willing to import all their photos to a third-party service: though other services, users have been “trained” to typically curate before importing, an assumption we need to break.
The full analytics as of December 31st are available in the attached document, but from a high-level perspective and during the 3 months period of Everpix Private Alpha, we have the following:
- The exact number is not available anymore but around 5,000 users registered on our waiting list and the vast majority were invited.
- 2,500 users created accounts and of those, 2,080 of them imported photos. This discrepancy could be explained by the fact some users realized they weren’t really interested or were using Windows for which we had no import solution.
- 11,100,000 photos were imported. 67% from Mac computers (i.e offline photos) and the rest from online services.
- 732 users ran our Mac Uploader at least once. We didn't have an accurate way of measuring it, but an indirect estimate from looking at the update server logs shows that 500 to 600 users were running it daily.
- For users enjoying the full experience i.e. Mac users, the average number of photos imported was 10,000.
On the upside, we can say that goals #1, #2 and #3 were achieved - in the scope of a set of users mostly made of early adopters. We can’t guarantee such metrics will translate to mass market though. We also had 3 important features build and “market validated”:
- the Mac Uploader,
- duplicate photos detection independently of their resolution or compression artifacts,
- instant and seamless sharing from Everpix website or iPhone app of a user photo to Facebook, Twitter or email.
On the downside, we discovered some serious issues with our back end infrastructure built on top of Google App Engine:
- It wasn’t as responsive as it should have been and optimizing it was becoming difficult.
- We had some strange and very hard to track down data corruption creeping in.
- We realized it would be very hard to control costs going forward as can’t dive into the system and tune it precisely for our needs.
The difficulties with these 3 important issues were compounded by the fact Google App Engine is essentially a black box, which means a lot less maintenance, but also a lot more limitations to investigate and optimize things. Getting a Premium Account with extra support doesn’t alleviate these issues either.
As you may be aware of, we have announced Everpix Public Beta a few days before new year’s eve 2011. There are a number of improvements over Everpix Alpha but the product is not that much closer to the full vision of Everpix. The main user visible difference is that anyone can create an account. The main under the hood difference is a new back end infrastructure (more later).
This difference alone was the driver to switching to a public beta this early: as for the back end infrastructure, the Mac uploader and the website, we do need a significant number of testers on the iPhone app early on. Unfortunately, restrictions on the iOS development process makes it very complicated and cumbersome to have more than 50 to 100 active testers. The best solution to reach a broad range of testers is therefore to release the app on the App Store, which requires open registrations.
The Everpix iPhone app was submitted to the App Store late December, approved but not released, as it needed to be updated to take advantage of the new back end and also get some extra polish. We plan to release it publicly before the end of this month.
It’s important to realize that with Everpix Public Beta we are not trying to acquire users, break records in terms of photos imported or generate buzz. We are still at the product building phase and just need to get enough users to test our products and validate our vision. We therefore did not try to get any PR coverage when launching.
We did however get a couple of unexpected articles on TechCrunch and MacGeneration (a big French website), which drove thousands of users and registrations to our website. As of today, on top of the 2,500 users whose accounts were imported from Everpix Private Alpha (but not all of them did activate their updated accounts), we gained about 3,000 users at launch of Everpix Public Beta due to the unexpected PR coverage.
These new users resulted in about 8,000,000 new photos imported in 24 hours, which effectively slowed down almost to a halt our new and minimally tested infrastructure from a scaling perspective. We had to disable new user registrations for several days to be able to take a step back and fix things.
As mentioned above, the major change under the hood between Everpix Private Alpha and Everpix Public Beta is a brand new back end infrastructure designed on top of Tornado & MySQL and which is entirely running in Amazon Web Services. This required a large amount of re-engineering, justified by the fact we realized we needed to own the full infrastructure stack to be able to resolve the problems mentioned previously with Everpix Private Alpha - and also from a strategy perspective not be dependent on an external vendor. Google App Engine was great for Everpix Private Alpha and saved us a lot of time, but we had reached the tipping point.
Re-engineering the back end also allowed to address numerous design issues with the previous implementation (for instance proper color matching of imported photos and consistent scaling across import paths) and add valuable features like much improved analytics.
This re-engineering does come with an important downside in the fact that it won’t scale for free contrary to Google App Engine which handled that part automatically. We can only do so much here with our limited internal knowledge and proper scaling will require hiring an experienced back end engineer (more below).
Learning from the massive input of data at launch and optimizing the back end accordingly took about 2 weeks and as of today, we are in much better condition. We should be able to handle 1 to 2 new million photos per day, although we really don’t expect to reach such numbers anytime soon as Everpix Public Beta is not being promoted.
Finally, although we do have many more analytics on our users compared to Everpix Private Alpha, it’s not worth looking into them at this point since the rocky launch likely affected them. For instance, users creating an account, being unable to import photos or get a good experience browsing them are unlikely to come back for some time.
This early in the project, the major risks are not how to acquire users, convert them to paying customers or simply getting Everpix known, but rather:
- Can we recruit to build the team we need?
- Can we deliver on Everpix 1.0 and its major pieces: innovative photo collection navigation, unique assisted curation and software for major platforms?
- Can we build a reasonably cost effective infrastructure that can handle for the next 10-12 months a potentially massive number of photos and associated metadata?
Since about early December, we are actively recruiting, by ourselves when possible, but mostly relying on 3 recruiters (which implies a ~20% added fee on the future employee’s salary). We receive resumes almost daily, do phone interviews several times per week and have candidates come on site if they are promising, but so far haven’t been very lucky. Finding the right talent is very hard these days, even if we have the money, and that’s why recruiting is listed as risk #1 above.
Currently open positions are (in no particular order of importance):
- Android developer to build the Android client,
- Windows developer to build the Windows uploader,
- Back end engineer to significantly improve and maintain our back end infrastructure,
- iOS engineer to help building the iOS clients (iPad and iPhone).
Our only successful hire so far is Jason Eberle, a talented web developer we found directly in November. He took care of fixing many issues in our website and implemented several new features. Jason is now just getting started on rebuilding the website from scratch with a robust and future-proof foundation - and of course add Internet Explorer support which we don’t offer today.
As you can see in the attached P&L, we spend 141K from September to December which is above the budgeted 125K from our first convertible note (aka seed round A), which was designed to cover exactly this period. These extra expenses can be explained by:
- Higher than expected lawyer fees,
- Jason being hired mid-November while the original plan was to get employees in 2012.
Other high-level comments regarding our budget:
- Lawyer fees are quite higher than expected but cover a number of things: incorporation, 2 patents, seed rounds A and B and other legal matters.
- Not being an employee of 33cube, Inc., Pierre’s personal compensation was recorded as consultant fees, not personal costs.
- Government fees are mostly the cost of Pierre’s H1B visa.
- The total amount of the 2nd convertible note (aka seed round B) is off by $26.80 due to a international wire transfer fee which was eventually resolved last week.
- Do not read too much into infrastructure costs, especially relative to our number of users, as this is currently unoptimized, since our focus is on feature building, not $/user optimization yet.
Going forward, we expect to use the money from seed round B to cover the upcoming six months i.e. until this summer, which matches the next goals described above. This money would only take us so far because our burn rate should significantly increase due to payroll (assuming we can actually find talent).
Of course, we would like to start having paying customers as soon as possible, but it’s unlikely to happen until Everpix is out of public beta - and we still have to figure out the precise pricing / features strategy. So in practice, we don’t expect any income before this summer.
Finally, in the interest of full disclosure, starting this January, we will be adjusting the salaries of us 3 co-founders. Since we started paying ourselves in September, we are significantly below market rate at [redacted], which is extremely tight at this point in our respective careers and lives, even more so in an expensive city like San Francisco. After wrapping up the work on Everpix Private Alpha and starting on Everpix Public Beta, we feel it’s a good time to do some adjustment on our salaries - without going to full market rate though.
Our next big milestone is this summer and the goals are to:
- Get Everpix out of Public Beta into an initial release representing our vision,
- Having the Android client and Windows uploader (dependent on recruiting as mentioned above),
- Start offering premium subscriptions to our users.
Note that this would mean about a 1 year development cycle, which is reasonable for a project this ambitious.
To summarize this 4 months period since Everpix officially started:
- With regards to expectations, our Everpix Private Alpha did pretty good.
- Everpix Public Beta is out mostly so we can more easily get testers even if the product is incomplete and is therefore not being promoted.
- We have a lot to achieve in the next 6 months to get out of public beta and execute a product that really shows the Everpix vision and that we can start promoting.
- Recruiting is our main risk right now.
- Assuming things align, we would be shooting for a Series A this summer.