Category Archives: Analytics

3 Steps to adopting AI

I’ve being asked for my view on AI a lot this year, more so than last. It would appear the industry has caught up with the ‘hype’ being a reality.

I’m not getting the slightly twitchy ’Is it going to steal my job?’ anymore though, now the questions are; Is this something I need to be bothered about?, Can I afford it?, Where do I start?

All good questions. If you’re a brand or business looking to get ahead, simply keep up, or maybe even catch up, you can’t afford not to be thinking about this and getting a beta into place. If you don’t your competition will, and then you will be sat watching them eat your share of market or voice, or both. Either way, I can’t imagine that’s something you’re keen to see happen.

Roughly a third of the brands I work across either have a version of an AI (artificial Intelligence) ready IA (Intelligent Assistant) or have jumped straight into an AI trial or beta. Every single one of them has seen positive results. Every single one of them is now developing a roadmap with us to put in place milestones to be better, faster and more informed on a real time basis.

The shape of these solutions vary from bettering service response levels to informing fashion design and everything in-between. But the steps to get there are the same, and here they are;

Work out what the problem is you want to solve

Ok, obvious right? But actually I highlight this because I recommend you don’t ‘do a chatbot’ because your competitor did.

Is there a challenge that advertising or marketing isn’t fixing for you right now? Do you have a human centered design idea that you can’t quite get to grips with? Do you have micro communities you don’t understand or can’t reach in meaningful ways?

All of these are problems AI can help you with, quickly and effectively. So consider where you might want to turbo charge a solution and put a brief together around that.  Be clear about your brief as well, if you’re vague about what you want to achieve it’s tricky to train an AI to think comprehensively, it in turn will be vague.

Review and understand ALL of your relevant data

AI is only ever as good as the data you feed it; the more data you have, the more connections can be compiled and the faster it will evaluate and learn. It’s not magic, it’s algorithm on speed. 

Define the goals you want to achieve in order to reach the objective in your brief, or work with an AI data partner to do this (most good agencies should have someone who can help you get started and then find the right partner for you, the answer isn’t always ‘Watson’ btw). You will likely have a mass of data you understand and a bank of data you’ve never really thought about, once you have it all in one place you need to work out where the gaps are and fill them in.

This up front bit seems tedious, that’s because it is. But don’t cut corners as you’ll only pay for it further down the line. The better the data set, the more robust your AI solution will be and the quicker you will see results.

Choose your AI partner

What you want your AI to do will depend on what supplier or partner you choose. There are many solutions already available at both scale up and enterprise level to choose from. They offer everything from; language skills, analytics, tech stacks that speed up services, listening, finding ‘moments of serendipity’ through to predictive analytics and forecasting.

A read of IBM Watson and AWS are good places to start if you want to dig more into what’s on offer, but also check out the likes of DigitalGenius and DeepMind for something smaller or a bit more creative.

Of course you may be looking to create something truly bespoke in which case you may have to hire a bunch of experts to create your algorithm from scratch, or seek a start up willing to work with you and co-create. There are an abundance of really cool start ups just about to break on to the scene so this is a truly valid and cost effective approach, don’t rule it out.

That’s it. From here, you should be in safe hands. You know what you want, you have the data in play to get it and a partner who knows what to do with the data to get what you want.

My parting piece of advice is to remember that AI / ML (Machine Learning) are solutions that learn and develop, think of it as a child going from kindergarten to PHD level but in weeks rather than years. There may be a few mistakes along the way but be patient and think big, because with direction and correction the results are nothing short of impressive.

And it’s not just customer service stuff either…

marchesa-thumbnail (1)

Marchesa and Watson’s cognitive dress – read more here

P.S. Here’s a mini wiki;

IA; Generally speaking an Intelligent Assistant is a pre-structured agent used to deliver automated responses but does not include self correction or ‘learning’, therefore is not always classed as AI. It’s often the step before AI and used to validate the quality of data. That said some do include NLP (Natural Language Processing) and are connected to the IoT (Internet of Things) so the line is often blurred.

NLP; Natural language Processing is a computer science that uses AI and handles human speech between computers and humans.

AI; Artificial Intelligence is an intelligent or cognitive behavior exhibited by machines, sometimes also referred to as problem solving or learning.

ML; Machine Learning is a sub-field of AI that includes programming computers to deepen the learning process.

P.P.S. If you find AI interesting generally you might want to check out my other blog, co-written with @kayperbeats – it’s a bit more off the wall but it’s insightful none-the-less.

Advertisements
Tagged , , , , , , ,

music is morphing

The audioscape is shifting. In 2014 digital music revenue matched the physical and though the report isn’t out yet I suspect the balance tipped over towards digital in 2015. Whilst downloads still just about make up the bulk of that revenue, streaming services show continued growth to satisfy the personalised on-demand needs of the ever impatient consumer.

It is becoming more apparent however, that rather than posing a threat to traditional radio broadcast, it is in fact providing pre-cognitive insight to help programmers find the next hit, or know when to stop playing a track to death, thankfully.

Having long been an advocate of services such as Spotify, Amazon Prime and YouTube I’ve pondered several times where the data connections between airplay, streaming and record sales will join up.

A short while back I spoke with Spotify (the world’s biggest music streaming service) about how their platform can inform what’s next, allowing them to be ahead of the curve on everything from up and coming artists to how to name their playlists, the value is clear; it’s a completely accurate analysis of listener choice.

Streaming is a mainstream activity. Over two thirds of internet users accessed a licensed digital service in 2015 and the strength of the industry today is seen in the total flexibility it provides, allowing artists to reach a much wider audience in a way they want to be reached.

This has seen a shift from music models based on ownership to those based on access, which coupled with consumers streaming more and more on smartphone and tablets (up 114% in 2015 according to Wells Fargo) means subscriptions will continue to shape the music portfolio available.

So what’s next? I reckon we have three things to look forward to:

First up, music will become more intuitive. The Echo Nest acquired by Spotify provides an intelligence platform that mixes human skill, clever algorithms and social curation, meaning you can quickly get personal. This thinking will spread.

Secondly, enhancing how we perform by influencing the frequency of our brainwaves will continue to improve. We all know that faster music makes us run faster and slower music focusses breathing for yoga. This thinking is already built into how Spotify’s algorithm can work, for example their partnership with Nike which matches music to your tempo.

Thirdly, a merger of these two approaches to create a constant seamless service that will use prediction to enhance our brainwaves through binaural beats so we all become super intelligent thanks to music.

OK, maybe that last one is a few years away… but it will happen.

 

-3952

Tagged , , , , , ,

A view on the programmatic semantics of binary trading predication

What the..? You may ask.

Well, I’ll tell you… you see most of my meetings this week have been about the uptake of programmatic, commonly I’m being asked; Is it robust? Is it robots? Are the robots robust? How do I plug it in? Do the robots plug it in? Are the robust robots plugged in?

Breath in.

So, having explained this a lot, I find it’s easier to start with what it is not:

It is not: Real Time Bidding (RTB)

It is not: A new type of media

It is not: A new format, a new device, a new tactic, a new insight or a new inventory. 

It is, quite simply put; AN AUTOMATED PROCESS.

Programmatic Trading simplifies the buying and selling process by digitally connecting the buyer and the seller of the ad space. This brings automation to the process adding operational and pricing efficiencies which take the mundane and repetitive tasks away from humans.

It is important to note that this doesn’t mean that creative is any less important, studies show that creative is still responsible for 70% of the effectiveness, the placement and timing making up the other 30%.

Marketing is, and will always be, about getting the right piece of content to the right person at the right time. Programmatic quite simply means we can be quicker, more effective and therefore scale in a more structured and relevant way. 

I love this example from Nike and Google, it’s a great demonstration of what can be achieved with clever design and RTB, and just recently Unilever have explored the use of video in their Romeo Reboot campaign.

So in summary, you still need a wicked idea, a clever plan and some digital genius behind it, but if you embrace the fact that you can’t be in total control of the real time exchange and you’re prepared to sit back and enjoy the ride, then some really cool stuff can happen. 

And contrary to the title of this post, it’s not that tricky…

image found on adweek.com - thank you

Tagged , , , , , ,

Analyse this

I would imagine you’re all familiar with Google Analytics and Omniture, and seen reports for PPC, SEO and Display, no doubt clearly outlining the return for every penny spent including; entry sources, bounce rates, CTR, page views and so on.

I’ll also wager you’ve asked your planners and buyers what the return for Social Marketing is? Yes? You’ll have asked: What’s the value of a Facebook ‘like’ or a ‘retweet’? We’ve all stared at pretty diagrams that show us the reach and potential eyeballs hit but, so far, it’s been an algorithm we haven’t quite mastered with the confidence to go back to the board and solidly say that the money spent has returned an incremental profit of ‘X’ through social.

Well this level of measurement has just taken two more big steps forward.

Firstly, Google recently announced that they are adding social media reports to their analytics suite which will show the social value through measuring; visits and visits via social referral, the conversions this led to, plus assisted social conversions and last interaction social conversions.

Secondly, Adobe has just unveiled its social analytics tool: ‘Adobe Social’. Apparently a more comprehensive version of Adobe Social Analytics, according to their Product Director Matt Langie. The new software still provides the basic listening tools already familiar to users but in addition now allows management of creating and publishing content and ads. It also follows similar tracking to GA so you can report from seed to purchase or drop off.

I wonder what this means to the likes of Radian and Sysomos, will these two giants take over?

 

Tagged , , , ,