This presentation explains how brands can mine social media data, both text and images, in order to find insights about your customers and markets that can provide real business value.
AI Virtual Influencers: The Future of Influencer Marketing
Deep Social Insight
1. Deep Social Insight
The secret of mining text and visual social media
data for valuable business intelligence
Roy Jacques, UK Managing Director, Sysomos
@royjacq @sysomos
2. The universe of data is expanding exponentially.
It will grow
50x
between 2010 and
2020 alone.
3. The holy-trinity of social media
Owned
Earned
Paid
Most brands are doing paid,
earned and owned social
media, but few have a
smart, integrated approach
to all three.
4. Here comes the new boss, same as the old boss
For years we’ve been telling
marketing to tear down silos,
while accidentally building
new ones for social data.
6. Owned social media channels
Brands owned social media
networks have become
complex and unwieldy. We
need joined-up management
tools.
7. Visual social media
Brands have a big blind spot
if they can only find social
media pictures that are
tagged. They need image
recognition technology.
10. social listening social intelligence
snapshot reports predictive analytics
information overload context from social data
brand monitoring industry understanding
complex queries ad-hoc keyword search
influence = popularity influence = relevance
Brands should be able to use social data to build
a better business.
From: To:
11. product
price promotion
place
$
What should we
build, and who
should we build
it for?
What should we
charge, and is it the
right price?
Where do our
customers buy
from us, and
where do they
want to buy?
How do we
convince our
customers to try
and buy what
we offer?
Social data touches on every actionable aspect of strategic
marketing.
13. Social intelligence campaign process
• Research historical
data
• Benchmarks and
KPI’s
• Contextual image
analytics
Planning
• Live social analytics
dashboard – owned
channels
• Realtime social
monitoring
• Image virality tracking
Execution • Overlay
paid/earned/owned
analytics
• Identify correlations
• Any surprises?
Evaluation
Use paid to boost organic content when required, understand how budget has performed
It’s not news to most of you that the amount of data being generated is growing exponentially – it will have grown 50x in this decade alone
We’re moving towards a world where current tools are simply not enough to understand the huge amounts of social data that is being created – marketing teams have limited resources
The sheer volume of social data is overwhelming – there are a variety of sources and it’s extremely difficult to know where to start and how to get the most value from it
One of the key challenges marketers face is the proliferation of data.
Improving the performance of marketing means being able to ingest, process and analyze incredible amounts of data, and that problem is only going to get bigger. In fact, within 10 years the total digital data available in the world will increase by 50 times.
Today, Sysomos processes and analyzes nearly 4 terabytes of data each and every day, which totals over 2 petabytes annually, and we expect that to multiply several times over in the next two to five years alone.
Only data science can solve against that, and we need common formulas that can be applied to data attributes over time.
Paid – Earned – Owned
Lots of businesses are good at different parts of this – maybe they monitor for brand mentions, or they track their engagement levels, or how their ads are performing, but it’s rare to see all of this stuff really joined up and integrated intelligently.
For years we, in the social media space, have been saying that it’s time to tear down the old marketing silos so that we can work more effectively in the digital age, but all along we’ve been quietly building up our own new silos.
There’s a lot of valuable data sitting in these different silos, and it’s going to be much easier to mine it for actionable insight if we can connect it all up. So let’s talk a little about what these new data-silos are:
Textual analytics - this covers discussions in places like Twitter, Facebook, Blogs, Forums etc. A lot of businesses currently only track a set of keywords (e.g. brand and competitors) so this means they miss a lot - it makes it hard to retrospectively look at things that you haven’t set up searches for. What you really need is full access to everything that was said on social media - the full firehose - so you can dig through it like a search engine and analyse it in a meaningful way.
Owned social channels - increasingly businesses have sophisticated international networks of owned social media channels, across different platforms, business lines, markets/geographies, etc - and even though you manage all of those channels, it’s kind of hard to get a holistic view of how your content is performing across that complex network, and what’s working well for your competitors. And to further complicate matters we’ve got paid/promoted content sitting alongside organic content and we need to understand how it all fits together.
Images - visual social media is huge, but we have a an equally huge blind spot. Most brands can only identify images that are correctly tagged - that’s just not good enough, because people don’t tag their photos. We need technology which can analyse all of those social images and not just tell us when our brand is featured in images, but also provide meaningful context.
Even if this person had tagged Starbucks with this image, they still wouldn’t have shared all of the additional context – and that’s the real value that image recognition can add, it extracts all of this additional rich information that would simply have been overlooked before
Then we have our web analytics, and data from our other marketing functions such as media relations, direct mail, etc. And of course all of this has to be linked in someway to the bottom line, so CRM/sales data needs to fit into the puzzle too.
This really all boils down to the shift from social listening to social intelligence.
Being able to analyse the data in this way means that we can gain valuable insight which touches on all areas of marketing: (product, price, place, promotion)
Right now we’re in a world where we have all of this data available, but it’s sitting in different silos with different tools and different stakeholders, so we’re not able to use the data to its full potential - it’s hard to see the connections between all of these different activities and outcomes. Two things need to happen:
1) End tool fatigue - we need platforms that unify everything – lack of efficiencies – everything in one place
2) Get better at working with APIs and integrating different data sources - even when we provide a one-stop-shop, you still need to tie it all into your own data sources (web analytics, CRM, etc). The marketing department should be working with developers to build apps and dashboards that do this
Today, most of our data analysis - especially in social analytics - has been descriptive, meaning it helps us look in hindsight at historical activity and report on it. The essence of social data at first only really helped us capture activity streams and try to make meaning of it.
We are moving toward a more associative environment, layering on useful insights against that reporting to correlate past activity with possible information that can help drive business decisions. That’s analytics at work.
To date, of course, that insight has been largely human-driven, requiring the brain power of a person to draw conclusions, derive the actionable information from a set of conclusions, and decide how to act on it.
We need to insist that social data not only anticipates trends but can, for example, identify them in disparate, emerging data, overlay data sets, identify velocity points in data and pre-emptively configure actions and engagements in response. The smarter our systems and more diverse their data inputs, the closer we get to helping marketers deliver relentless customer relevance.
Research/planning phase
Historical social web research to understand what’s worked for others and how audiences have responded to similar campaigns – not just your own, but competitors or comparable organisations in parallel industries
Set benchmarks/KPIS based on this research
Image analytics – get context cues and a deeper understanding of how people use products, talk about your brand visually
Execution phase
Live data from your owned channel analytics to understand how audience is responding
Realtime social media monitoring to track how wider social media, beyond your own channels, is sharing your content and responding to your messaging
Image virality monitoring – spot which images (UGC and owned) are getting distributed
Evaluation
Overlay all of the campaign analytics from your different channels to spot any patterns/correlations between them and anything that surprised you
Paid – understand when to boost content with paid support, and how