This was the first time I attended How To Web and I was very impressed by the quality of their Talks. As a tech person, with a lot of things to do, I didn’t have time to participate to all of the Talks / Keynotes / Stages.
Alex Hunter © How to Web 2013
On the first day I arrived at about 13:45 and caught half of the Talk on “Learnings from working with remote R&D teams” by Ivan Brezak Brkan (Netokracija) – moderator, Marius Jumolea (Visual.ly), Michael Levit (Spigot), Peter Brodsky (Soundcloud). I watched Marius Jurmolea’s talk about the importance of allocating more time to tasks. “If a task will take 2 hours, allocate 4 hours to it. 2 Hours to find the best way to do it and 2 hours to do it.” If we think about it, it’s a little hard for us to do that, especially because in a startup there are more tasks to be done by fewer team members in a very short period of time, so it’s expensive in short-term view but actually very valuable in the long term. We can plan the best way to do our tasks and it will have, probably, less problems if we first think about them more, otherwise, we’re likely to find that we just wrote some impressive-looking code that does nothing useful, and be forced to hack it up, or scrap it and start over, that will definetly lose us more time than just asigning some extra time to think about the task ahead. More details and some tips can be found here: http://c2.com/cgi/wiki?ThinkThenCode
The second talk was held by Sitar Teli who is Managing Partner at Connect Ventures (UK). She talked about Core Metrics and What Web and Mobile Companies Should Focus on, explaining that most of us focus on the wrong things like increasing acquisition, which is wrong.
I realize that if I talk in detail about every speech this article will get extremely huge. So I’m going to stick to a very short description, maybe a bigger one for the Big Data Speech 😀
The third Talk was from Alex Hunter – CEO, Rushmore about Getting and keeping customers – a guide to customer experience in the digital age. The main idea is simple. In 2013 everyone is in a rush to build products. But in 2013 the idea is not about creating a software or a product, it’s about creating Emotion. Because Emotion > Reason and usually Emotion leads to Action and Reason only to Conclusions.
And now, about the Talk I loved, held by Philip Kandal which is CTO at Skobbler about Big data on a small budget.
You can use Big Data if you have at least 1000 active users. And the workflow in this is the following:
1. Log -> 2. Process -> 3. Analyze -> 4. Improve -> 5. Eval / Test -> 1. Log
Philip Kandal © How to Web 2013
– Logging: MongoDB, VoltDB and Cassandra
– Processing & Analyze with Handoop & Hbase, Storm
– Optimizing is made with Mahout
– Try to Analyze / Process Async
– Do it on cheap dedicated servers (It’ll be much more cheaper than using Amazon )
– Use open / free software
– Log as much data as feasible so that it’s available later
– For the processing step, move your data to the processing environment, you can also add all relevant data from Facebook so that your data will be enhanced. And enhance it as much as you can.
– Analyze: Cluster the data in meaningful groups and compare it. Compare your data versus reference sets (geography, time of day, past behavior) . Define KPI.
– Improve by learning from your analysis. Do A/B Testing with the new features (at least 1000 users / 1000 users. And after the testing is complete and the feature is good, deploy it).
– Try visualizing your data. Do not put it in spreadsheets if you can use things like heat maps. You’ll understand the data easier.
“As smart as you can be, you’re not smarter than data, because data is objective and you are subjective”.
With the tools provided now, enough will and all the documentation you can Setup in 2 weeks without prior Knowledge. A tip will be to put some time beside for crazy weeks, to check cool stuff, and you’re team will definitely improve in results.
“Do not be afraid to use R or Matlab. Visualize your data! If you can visualize it you can understand it.”
Next was Marco Cecconi from StackOverflow telling us why StackOverflow can server so many pages to so many people in such a fast time. It’s all about caching. Cache all the things, from Network Level Caches (CDN) to Sources Level Cache (httpRuntime.Cache) to Site Level Cache (Redis) to SQL Database Cache (384GB OMG!) – Also there was another one but I can’t remember it 🙁
Simon Stewart © How to Web 2013
Second day I caught Simon Stewart – Software Engineer @ Facebook with How Facebook Builds Facebook on Android. He explained to us how Teams collaborate and work, how they review code with their own tool called Phabricator, that at least 2 devs see every piece of code (at least the person who writes it and the person who approves it). The cool thing about Phabricator is that you can insert gifs as comments. Haha. “SHIP IT!” (*insert image here* http://24.media.tumblr.com/tumblr_lz4u84XJPc1qfu5wgo1_1280.jpg ) And also, he told us what is Buck.
I’ll also include my favorite thing said by him (yes I wrote it :D)
“You have a limited number of keystrokes in your life time. Even if you type faster or type with a keystroke per second, there is still a certain number of keystrokes. And we don’t want to lose those keystrokes by writing emails or testing things or fixing bugs. We want them to do something fun: Write Features.” 😀 WRITE ALL THE FEATURES!!! – haha
Last Talk that I’m gonna cover in a few words is Mat Clayton’s Talk, he is Co-founder & CTO at Mixcloud.
The talk started very well: “If you can’t make it, fake it!”
Fake features to see how the users behave and if they want to use that feature or actually buy it. They faked a system that let users buy ads and after that sent them to a 500 system error, or a message like “We haven’t made this feature yet, tell us how you want us to build it.”
The next thing he focused on was on the Team. The Team is everything! Also included a quote: “Never tell people how to do things. Tell them what to do and they will surprise you with their ingenuity.” – George S. Patton
The next part was about the importance of A/B Testing. At MixCloud they do A/B Testing for every feature that will be launched. One of the tips he gave us was to not forget to do A/A testing. Both the groups with the same features. “You’ll be amazed how different the results can be”. Do this so that you can better calibrate your groups.
And last but not least:
“A good plan violently executed now is better than a perfect plan executed next week.” – George S. Patton