Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Data for Hong Kong startups
1. The big deal about Big
Data
(and small data)
for Hong Kong startups
Guy Freeman
guy@dataguru.hk
2. Who am I?
A statistician working as a Post-Doctoral Fellow
at the University of Hong Kong as a modeller of
influenza, amongst other projects
In my spare time I am interested in how data
science / analytics / machine learning / artificial
intelligence can change/save the world
Want to help startups use their data better
3. What is Big Data?
Big data:
● So much data that you can’t easily or
quickly:
● store
● search
● analyse
● visualise
It’s a hot topic right now...
...but don’t forget Small Data!
4. Data is a valuable resource
Your startup is collecting data on usage rates,
purchases, clicks, interactions... or at least it should
be. How do you use the data to make business
decisions?
Data = information ^ n, where n is the number of
“views”
How many views do YOU have?
5. Different “views” of same data
The same data can be used for optimising:
● App/site aesthetics/presentation
● Fraud detection
● Marketing
● Pricing
This can be done through you reacting to data,
or automatic algorithms doing it for you!
6. Data-based startups
With so much data whose value still hasn’t
been realised, some startups focus purely on
that conversion.
In Hong Kong, already have:
● Demystdata: Better credit scoring for
financial industry
● Multichannel: Just launched “AI for
marketing”
7. Lean Analytics
I’m sure you’ve heard of the “Lean Startup”:
“a combination of business-hypothesis-driven
experimentation, iterative product releases, and validated
learning”
You can use all the data you collect, and
external data, to achieve this through “lean
analytics”.
8. Overview of data-driven business
1. Collect data
2. Work out reasons for current state of data
3. Change business in light of reasons
Much harder than it looks ;)
9. Easiest way to implement data-
driven business
Experiment! e.g. A/B testing
Then put away your ego
and act on what the data tells you
DON’T USE VANITY METRICS.
And don’t be delusional.
10. Harder ways to implement data-
driven business
Many (most?) times you can’t experiment, but
you still want to understand what is making
your affecting your metrics.
There are sophisticated statistical / machine
learning algorithms, but I’ll leave description of
them for next time ;) But take-home message:
CORRELATION DOES NOT EQUAL
CAUSATION
11. Conclusions...
Not using your data to run your business is
equivalent to driving while blindfolded
Automate as much of your business as
possible, reacting to data in realtime
Second-best strategy is reacting to appropriate
data manually, perhaps using visual summaries