Joe Griston (Director of People and Talents @Freelancer.com) выступил с темой "20 million users and 10 million projects, how to scale like Freelancer.com" на Conformato Conference 2016. Приобрести видеозаписи конференции можно по ссылке https://goo.gl/V6qIJ0
21. 21
We are changing
lives in the
developing world
by providing
opportunity and
income.
We help small
businesses,
startups and
entrepreneurs
turn that
spark of an idea
into reality.
Design of a digital counting
wristband for $220
37. Global, and growing
37
34 Supported Languages 20 Supported Currencies
Languages Currencies
USD
AUD
CAD
EUR
GBP
HKD
INR
CLP
44 Regional Websites
24x7x365 Multilingual Support
Team
Afrikaans
Bahasa Indonesia
Bahasa Malaysia
বাংলা
Català
Čeština
Dansk
Deutsch
ελληνικά
English
Español
Suomi
Filipino
Français
िहन्द
JMD
IDR
ZAR
MXN
SEK
JPY
MYR
NZD
PHP
PLN
SGD
BRL
Magyar
Italiano
⽇日本語
한국어
Norsk-Bokmål
Nederlands
Polskiа
Português
Română
ру́сский язы́
Shqip
Slovene
Svenska
Kiswahili
Thai
Türkçe
Tiếng Việt
українська мова
中⽂文(简体)
54. ➜ Includes staff with PhDs and advanced degrees
➜ We now have 145 engineering staff across 5 labs
➜ Up to 40k+ users on-site at any time
➜ Databases serve up to 65,000 QPS
➜ Reliability of 99.99% YTD
56. • 80 million events per day
• Over 120k user notifications per
day
• Over 500k user messages per day
57. • Decouple independent components
• Targeted scaling solutions
• Insulate against broader impact
• Accelerated development through
working with APIs
• Mobile applications
• User assistance / support
Service Oriented Architecture / External API
61. #1: DON’T BE LIKE MOST COMPANIES!
MOST COMPANIES ONLY TRACK THREE THINGS
Traffic Users Revenue
Phase 1 Phase 2 Phase 3 Phase 4 Phase 5
Get
Traffic ?
Get
Users ? Profit
FREELANCER PRINCIPLE #1
63. • Transparency: All metrics incl. revenue is accessible for everyone in the company
• Monthly/Quarterly goals need to be in quantifiable metrics
• Every member in the team is accountable for their team’s metrics. Performance is strongly
rewarded based on that.
• SIT THE PRODUCT TEAM TOGETHER (NOT FUNCTIONALLY) !!!
Growth-Driven Teams
67. DO NOT HIRE DONALD!!!
Our website is shit, we have to make it more usable!
OK, how? Do we have data to prove this? What metric
defines usability?
I can’t code. I can’t query the database. Can an engineer
find out how many users we have in Australia?
NOPE. They’re busy.
We need our campaign to go viral
What defines virality? How much revenue will it make?
Where do you think the problem is?
Reads Wired.
Oops, is it 5pm already? I have to go to my hairdressing
appointment!
68. DO HIRE WARREN BUFFETT!
• CompSci, Maths, Economics, Stats or Engineering Degree
• Dashboard & data are their source of truth that they look at
and refer to regularly
• They also constantly query and add their own stats
• They can visualise and fit their core funnel perfectly with
DATA
• Can code (AND NO HTML IS NOT CODING)
• Always formulate his/her own hypothesis (mostly data-
backed) , code and validate it him/herself
• SQL Gurus
• Reads HackerNews
70. WE HAVE TO A/B TEST
• Correlation & causation can now be tested easily
• Always form hypothesis and test it by linking it to
success metrics
• Almost all things have to be A/B tested including
your new features
• A/B testing should be done on multiple metrics – NOT
just one metric
• Mobile testing is a bit harder but we can still do it
FREELANCER PRINCIPLE #1
75. • Generate 50+ experiments for the product
• 50 in 50mins
• Group experiments into AARRRR
• Determine the ICE score for each experiment (with the team in a planning meeting)
• Impact (I) = Estimate of how impactful a feature will be (scale of 1-10 with 10 being the most
impactful)
• Confidence (C) = How confident you are that the experiment will achieve the estimated
impact (scale of 1-10 with 10 being the most confident)
• Effort (E) = The estimated amount of effort to develop and produce the experiment (scale of 1-10
with 10 being the least amount of effort)
• Rank experiments by ICE score to determine the experiments most worthwhile to tackle during the
sprint
• Ensure with the development team that all experiments are doable in one day (not strictly necessary
for all growth sprints). THREE SPRINTS THREE EXPERIMENTS EACH DAY -> 30 EXPERIMENTS
IN TWO WEEKS
PROCESS
78. AT FREELANCER,
EVERY QUARTER, WE HAVE
ONE FULL MONTH OF
GROWTH-FOCUSSED TESTS
AND TWO MONTHS OF
DOING THE RIGHT THINGS
RIGHT
FREELANCER PRINCIPLE #3