Warning: Liveblogging. Here be inaccuracy, errors and typos
How similar are the Union Flag and the Tricolore? Both flags…
Bike and a ball? Not very. Expcept, they’re both things people buy for their kids.
This is the problem data has right now. Measurable quantities are important, but easy and limited. Characteristics like brand held. But you also need to understand product categories.
You CANNOT handle all of the data out there. If you want to do things in real time, you have to choose the data you work with. Item to item? If one person bought a bike and protectors, you can recommend protectors to someone who buys a bike. But if they bought the bike for a performing monkey, and also bought bananas, you’re going to get strange results.
So, you use social demographics, because people who share characteristics, tend to share habits. But not completely – would you recommend AC/DC to a mid-40s above average income guy (uh, probably 🙂 )
All of this standard targeting is limited.
So… social media marketing. All friends! Similar likes! Except… grandma is a friend on Facebook. Very different tastes…
So…consumer action mining – similar customers who are interested in the same things. Segment all you customers into groups, without throwing away information. And algorithm based on the physics of complex systems. Physicists laugh at us – because they deal with way more data than we do.
We do data mining on the actions of your consumers – might be buying, might be surfing pages, might be listening to a track or playing a game. All good. All interesting. People have many dimensions. If we show the consumer something, and he doesn’t click on it – we need to capture that. We need negative events, too.
What you need:
- raw data and anonymised data
- unique user ID, unique event ID
- 3 events per user
He gave us an example of four groups of users – who all looked the same. They had to go deeper to sub-categories of activities before they were able to spot differentiation. He seems to be suggesting a lot of pre-calculation, that allows you to match event tracks in real time to particular people.
Claus Moseholm, goviral
Let’s talk video.
The audience is about engagement. You want them to spend time on your video, to engage with you. It’s not about click through to an e-shopping page, it’s about building emotional ties. Traditionally, we focused on the success of click-through – and the rates of click-through have been dropping almost since the internet began. So, there’s been a shift from destination to distribution as a central plank of thinking. Video both gives you the opportunity to tell a story and gives you better click rates – but that’s on the “play” button.
Media are shifting display and TV budgets into online and viral video. Your moving beyond brand awareness into engagement. CPM is giving way to CPE – cost per engagement or view.
Afterwards? Sure, you can put a click through there, but many people use that space for interactive overlays and other experiences at the end of the video. It’s about persuading them that they want to spend time with us. Rates of people doing follow up actions after a video are going UP. <– interesting.
However, engagement drops with video length – the median seems to be just under a minute, according to a graph he showed. Keep it short.
Ciaran O’Kane of exchangewire.com gave a quick update on what publishers are doing with modern ad models. Most interesting idea: Could publishers become media agencies? If they’re inventory limited, could they start selling those extra ads on, using their own data and the trading platforms?
Wolf Allisat of ComScore is “the antichrist of clicks”. He goes around telling people that clicks are the wrong thing to measure. Click-through rates on online ads. 0.11% click through in any month. People try to make up for this with volume. And we could, for a while, when growth rates were 200%. Clever advertisers are buying loads of PPC ads, because they get all the brand awareness with none of the costs… Publishers should change their models right now.
And who are the people were clicking? 62% of clicks from 3% of the internet audience… Do you really want those people?
Significant branding CAN be achieved, according to comScore research. It drives sales – and the offline lift is higher (!) than the online lift.