This document discusses different revenue models for internet companies, including symmetric vs asymmetric models. It provides examples of companies that use each model and notes that asymmetric models where users and payers are different dominate due to low distribution costs online. The document then outlines several steps companies can take to generate revenue, including getting users and selling their clicks to merchants. It emphasizes the importance of focusing on revenue metrics from the beginning and using A/B testing to establish causal relationships and guide business decisions.
3. Symmetric vs. Asymmetric Business Models
Symmetric models Asymmetric models
are ones where the are ones where the
user of a product user is different from
the person paying the person paying
for it are the same the company for the
product
4. Symmetric vs. Asymmetric Business Models
Amazon.com Facebook
Salesforce.com Google.com
Yahoo Mail (old) Gmail.com
iTunes Store Yahoo.com (mostly)
All virtual goods eBay.com
Hyatt.com Twitter.com
iphone games Like.com
Notice a trend...
5. Symmetric Business Models
Are easiest because your team only has to focus on 1 type of customer
There is no butterfly model of focusing users and customers
No push and pull within internal teams on “selling out the UI to advertisers”
Revenues and up a very accurate proxy for user interest / satisfaction so you
only have to focus on one variable but there is friction to distribution (huge)
Unfortunately for B2C companies usually asymmetric models dominate...
I’d say 80% of the market cap (my very rough back of napkin) of Internet
companies is essentially asymmetric
6. Why Are So Many Success on the Web Asymmetric?
Because the cost of distribution is almost zero: so very easy to make it free for consumers
Source: The super smart Mike Spieser @
http://laserlike.com/2009/04/18/microeconomics/
8. Background Story - History of Like.com
Ojos/Riya Like.com Style Shopping Engine
Face Recognition Visual search engine Visual search + all features
Technology focus but mostly used for and sites needed to build
shopping for soft the best way to shop for
goods soft goods
$0 rev $10MM rev $20MM+ rev
Proprietary & Confidential
9. Don’t Be Too Oblique. Oblique = Miracle
Case Study - Original Riya Model
Step 4 Step 5
80% will be UGC Google
public like Images w/o
Flickr copyright
Step 1 Step 2 Step 3
Face rec They will We will tag
will get users upload lots and they
of photos
Step 4 Step 5
They will CPM monetize
come back
often
10. Shortest Distance Between Two Points...
Case Study Like.com
Step 1 Step 2
Get users to Sell their clicks
search for to merchants
products on a CPC
11. Note to Self: Silicon Valley and Death Valley Are Both in CA
13. Consumer Intent to Do Commercial Transaction
Don’t Be Low Traffic and Low Commercial Intent
Lead Gen Nirvana
High Intent
Billshrink Only Google Search
Like.com Twitter?
Top 20 sites
Death Valley
Facebook
Low Intent
Most Web 2.0
Twitter?
companies
Online Games
Low Traffic High Traffic
Scale in Terms of Number of Users
14. Consumer Intent to Do Commercial Transaction
Corollary: Relevancy Doesn’t Equal Intent
Any ad with Budweiser Girls on a Botox ads on dating sites
Mommy blog
High Intent
Geriatric shoes on a MMORPG Dogfood ads next to photos of
gaming site dogs
Low Intent
Low Relevancy High Relevancy
Relevancy of the Ads Put
15. Revenue: How to Build it. Step 1: Focus on it
You can’t move a number you don’t focus on everyday
16. Start Early - Takes 1 year of tuning to make it work
Daily Average Revenue at Like.com from Launch
2007 2008
First year
17. Do A/B Testing for Causality
Data Information Knowledge Wisdom
One Two points Correlations Causation
point compared
It is 72 3 degrees When there Jet stream moving
degrees today warmer than are clouds south brings
yesterday it tends to be storms to CA
cold
You don’t want a data warehouse you
want a small room of causations
18. Keys to Wisdom Tracking
A/B testing environment that is parallel - not serial
Multiple independent machines (We have 7 sets now)
Define one or two key variables (CTR and CR for us)
Run tests with error bars on the results so you know when to stop
Don’t do releases where any new feature is on - turn on later 20/80
Have scale - as you divide you traffic it will take a while to get results
Hire Ph.Ds (we have 3) as they are better at experiment design
19. Parting Thoughts
Successful entrepreneurs are more
interested in being successful than in
being right
One self serving note: We are hiring for
backend Java developers and front end engineers
munjal@like.com
munjalshah (twitter)