13. ...there are 17 spurious correlations.
@willmcinnes @brandwatch
14. 3 main risks with social data
1. Sample/selection bias
Assuming people on social are representative of the people you're
interested
Assuming the people you're interested in are posting on social
2. Inference problems
Things like sentiment, gender, location, etc. are inferred with less
than 100% accuracy
3. Being creepy
@willmcinnes @brandwatch
19. Be real.
These are all real examples.
All but one are Brandwatch customers.
22. Goal:
Blend social data with weather data to find insights
How:
Got social data for customers talking about consuming their ice cream
product using
Got weather data for the same period
Outcome:
Found there were meaningful increases in people talking about eating
ice cream when the weather was bad.
Used that to inform their future advertising strategy
@willmcinnes @brandwatch
24. Goal:
Jump in to conversations about test drives to signpost potential buyers to local dealers
How:
Queries set up to locate social mentions that mention car model names with ‘test drive’,
dealers names.
Using Rules, Categories and Tags to automatically filter these conversations by Colour,
Model, Brand, Dealer etc.
Then matching CRM details of known customers with social handles to explore the
potential of social CRM at scale (they already have a database of >1m customers on
social).
Outcome:
Increase in car sales from test drives.
26. Goal:
More effective ad spend and return visits to their parks
How:
Identified people who met demographic criteria in each of their theme park
DMA region.
Identified topical areas of interest in those demographic segments, by region
Fed those topics into tailored regional advertising campaigns
Outcome:
Uplift in ticket sales + increase in per ticket revenues
@willmcinnes @brandwatch
28. Goal:
Change and Adapt Brand Perception
How:
Matching offline physical event check-in data with the social conversations
around each of the physical events
Matching social handles to offline identities and then observing and learning
Outcome:
New evidence and insight into which events drive the most brand favourability
change.
@willmcinnes @brandwatch
30. Goal:
Understand which brands and items their existing customers
were talking about publicly
How:
Acquired mentions for the key brands that they sell
Worked with a third party vendor to match social identities to their own
CRM database
Outcome:
Used information to promote those brands and items via the website
and email. ROI ‘made the leaderships’ jaws drop’
@willmcinnes @brandwatch
31. The point is that it’s not just about social anymore
• It’s about the business
• The customers
• The market
• Social is just part of it
Social data is valuable commodity that dept needs and wants
A quick update for you
Brandwatch now has clients in 51 countries.
And just since Dec 2014 users have logged in from 93 different countries!
30% of Fortune 100 use Brandwatch data to drive business decisions –
about campaign performance,
about product feedback,
on supply chain and inventory
Crisis comms and command centers
ARTHUR C CLARKE – Magic quote OR, engagement doubles with dog tweet
Picture of cute dog- mymessgage – I’m not going to stoop that low
Be much clearer.
Think about other industries that went through this.
My personal story – I’m much better at hyping than at this stuff, but this reflects where the world is.
One of our Phd / Doctors
Small data – not millions of customers – the 1,500 who were actually talking about test driving their cars
May, 50,000 tweeted ‘I want to go to [theme park] – so not general tweets, but indicating some kind of desire or intent.
Make it more exciting, go for it.
People remember experiences!! You don’t remember names but they remember experiences.
The reason that photos are so popular is because…
Have a look at IFTTT
latitude and longitude coordinates of the message locations have more than 5 decimal places of precision
Every dept needs it
Shingy // Hype is gone // Most of the gurus are gone – it’s REAL BUSINESS NOW