Some days ago, I talked with Chris Johnston, CEO and co-founder of Adoreboard.com, a UK business intelligence company who are creating waves in Europe and the US (they’ve just picked up Best Tech start-up 2014 award from the influential Silicon Valley Technology Forum). The conversation was very exciting for the great platform that he and his team are creating, so the idea of an interview came to the table, so Chris and I worked together to prepare one technical interview to their Chief Technology Officer Dr. Fergal Monaghan, to take a closer look at the core of the technology and ultimately what is creating the clear differentiation they’re achieving in the big data and analytics space.
[caption id=”attachment_175166046" align=”aligncenter” width=”263"]
Dr Fergal Monaghan CTO of Adoreboard.com[/caption]
Thanks a lot Fergal for this interview. Please, could you explain who you are and your background? Sure. I’m from Galway on the west coast of Ireland where I completed my Ph.D. and post-doc at NUI Galway’s DERI institute, the world’s largest semantic web research institute which recently co-founded the Ireland-wide INSIGHT big data research centre. As part of my research I was lucky enough to collaborate on attachment as a visiting researcher with other internationally leading research teams at Stanford University, University of Innsbruck and Seoul National University.
During this time, I gained a specialisation around the combination of Machine Learning and Semantic technologies. I also realised that I prefer to build innovative new technology rather than write papers about it. So in 2010, I moved into applied research with SAP where I led a research team building new analytics technology for SAP’s Business Intelligence products. Here I learned about scalability, best practices like agile, and productisation: how to build scalable products rapidly rather than prototypes.
So let’s start by understanding what Adoreboard does Fergal?
Adoreboard is a revolutionary platform that allows you to understand how the world really feels about your brand. With Adoreboard, you can define your brand’s impact as a single metric linked to a set of real time emotions and plan marketing activity based on accurate and actionable insights.
Tell me how does Adoreboard’s technology work?
The answer is: Big Data Analytics. Adoreboard’s technology ensures marketers are always aware of how people perceive their brand through Big Data analysis.
Whilst Big Data has certainly become a buzzword, it is most useful when taken as a problem space. In particular, the problems with dealing with Big Data are best summarised by the “three ‘V’s”:
- Volume
- Velocity
- Variety
Firstly, we deal with the volume of brand intelligence data with a compute space architecture. Data to be analysed is put into the compute space where a pool of idle worker machines pull them for processing. When the volume increases and the compute space begins to fill up, more workers are simply added to scale linearly as needed. A fancy way to say we’ve a scalable architecture.
Secondly, we deal with velocity by having our workers process data as a real-time stream so that as each piece of content is written online, we process and present the analysis to our users as it becomes available. You can sit back and watch individual mentions of your brand literally arrive on screen as they’re written. Finally, we deal with the variety of brand intelligence data by combined analysis of structured and unstructured data from heterogeneous sources to provide a 360 degree perception of your brand. The “Adorescore” summarises human perception, regardless of whether that perception was expressed via likes, shares or comments.
Let’s talk about Concept disambiguation
Our technology gives you a clear, accurate view of your brand’s perception through concept disambiguation. Let me give you a quick example, when you search Wikipedia for “SITA”, you are brought straight to a page describing the aviation information technology company (who are also one of our customers). This is Wikipedia’s equivalent of Google’s “I’m feeling lucky” search option, but Wikipedia also provide that little disambiguation link where you can explore and choose from the other concepts matching that name.
For example, you may decide that you wanted Sita the waste management company, or even Sita the wife of Rama in the ancient Hindu epic Ramayana. Similarly, we provide concept disambiguation to our users using a technique called Explicit Semantic Analysis. To monitor the performance of a brand or concept, simply type in its name and choose from a list of disambiguated concepts.
Our Adorebots then tailor their filters for relevant content relating only to the concepts you care about, and exclude noisy content about waste management or ancient literature which could pollute business-critical brand intelligence.
Now, focus on Common-sense reasoning
Our technology provides you with hidden insights you would otherwise miss via common-sense reasoning. We model and reason over fuzzy emotions as well as hard facts surrounding common sense concepts. For example, a hard-facts-only knowledge base will tell you that a “birthday cake” is a type of food, and has specific ingredients. However, that is an incomplete view of common sense human understanding of a birthday cake, so we also model that a birthday cake *commonly* has pleasant emotions and a pleasant occasion attached to it.
Common-sense reasoning over this combined model allows us to detect the implicit expression of emotion even in short pieces of text where no explicit emotion is expressed e.g. “Just had my birthday cake at Windrush Cuisine”. In this case, we can give the brand concept of “Windrush Cuisine” a bump in the pleasant and satisfaction dimension. This approach goes a long way to bridging the gap between human common sense and computer understanding today. By the way, Windrush makes a mean rasta brownie!
What about Affective modelling?
Adoreboard helps you understand why your brand is perceived how it is via affective modelling. We bridge the semantic gap between human-understandable emotion labels and a machine-understandable vector space using a psychologically-motivated affective model. This affective model allows users to easily understand the specific emotions expressed about specific concepts while at the same time allows our technology to easily aggregate and rehash the analysis at will.
This has been shown to outperform the valence model used in traditional sentiment analysis for the past decade. Something marketers will be familiar with as positive, negative and neutral sentiment. Valence models a single dimension which, in the words of a Chief Marketing Officer of a global ad agency we recently spoke to, “has failed to deliver” based on lack of sufficient insight on what action to take. The affective model we employ allows you to drill past your Adorescore into the dominant emotions that influenced it, and further into the brand mentions that expressed those emotions. Furthermore, these emotions and mentions can be broken down by demographic, channel, geo-location and comparison to competitors.
What’s the biggest pain point Adoreboard solves?
The single biggest pain point we solve is the high cost, staleness and blinkered nature of customer surveys and other insight tools. Marketers spend large budgets on regular customer surveys which are expensive, have a long lead time from initiation to report, must predetermine what questions to ask and which are only responded to by a certain subset of the customer population, which can introduce bias in the analysis. However, customer surveys can be a powerful and irreplaceable tool for gaining detailed insights from customers. So rather than compete, our technology compliments and removes the pain of customer surveys by indicating when a customer survey may need to be triggered like a sudden drop in Adorescore, for example.
Our technology also allows exploratory analysis: finding out what questions to ask rather than just getting answers to known questions. If customers are expressing an interesting perception about a particular concept online, it may be time to ask them some detailed questions about this particular concept. This allows marketers to drastically cut their customer survey budget, both by reducing the number of surveys and by gaining a higher ROI by triggering surveys on the right concepts at the right time.
Another massive pain point we’ve discovered through our customer engagement is also the lowest hanging fruit technology-wise for us to solve: the high cost and staleness of integration. Marketers today are using multiple tools to gather a 360 degree view of the perception of their brands online. They then expend effort manually integrating all the useful data from multiple sources into a single report (often via a slide deck) which they then present to decision makers for consideration. This integration effort obviously entails losses in money and more importantly time. One customer we spoke to spends one week out of every month compiling such a report, which is then several days old by the time it is presented. Our technology removes this pain by automating seamless integration. We integrate in both directions via API to multiple structured data vendors such as Google Analytics, Facebook Insights and more niche tools like Marketo.
While we visually integrate this private data for the user side by side with our own analysis, we keep this private user data private, and do not share or integrate it in any shape or form, even at the data aggregation level. The Adorescore is based purely on public perception data, while the numbers from the tools you know and love sit hand in hand within your own private live report. This way we radically innovate our user’s analysis of their brands’ public perception while seamlessly supporting their existing processes and familiar tools. Now marketers can spend that extra week in the month on high-value activities that require their expert eye, and when they do need to present, report or make decisions they can do so at a moment’s notice with up-to-the-second analysis on a 360 degree view.
So how do you bring this innovation to market?
The agile approach we take allows us bring the highest value innovations to market quickly. It also allows us to scope out a roadmap of what innovation we will bring next at a high “story” level. Stories describe the desired needs of a user which in turn determine feature sets. This process allows customers to provide feedback on our roadmap and help us reprioritise and change direction if necessary.
How do you make big data analytics accessible to marketers?
Many well-intentioned data analytics providers are focused on milking their technology for that extra 1% of accuracy. You can imagine the picture: data scientists feverishly working away in a rather sterile lab setting. However, what we are hearing from customers is that available solutions are missing the point: customers don’t care about that extra 1% accuracy if they can’t readily make decisions and take action from the data presented. Today you can access some really quite powerful and impressive analytics, but they are usually presented in a rather opaque dashboard full of charts, statistics and filters that aren’t doing much for the user.
Our experience is that decision makers like CEOs or CFOs need to be able to look at the data and within 5 seconds make decisions. Getting experts to explain what the data means is counter productive and entails both time and money.
Adoreboard’s manifesto for making data accessible to marketers and CEOs is to ensure our dashboard *is* the slide deck for translating decisions to actions. The key battleground where we compete with larger analytics companies here is the user’s experience with our frontend. For that reason, I personally believe our Design Lead is the most important team member at Adoreboard: even more important than me!
Our experience is that decision makers are receiving insights on how their brands are performing and are now saying “So what? What do I do now? How do I improve my return on investment?” Our technology focuses on providing the user a clear path to ‘Decision to Action’ so they can use the insights we deliver to make a business decision. We don’t just make data impactful for marketers, we also make decision-making easier by showing them what is working for competitors, what worked in other similar campaigns and not just making recommendations but making it straightforward for them to take action on those recommendations such as by taking out a sponsored ad on Twitter or initiating a new Marketo campaign, for example.
So, if you want to know how to improve your brand’s performance and revamp your customer engagement, just talk with Chris, visit the site or use their Social Media profiles to reach them: Adoreboard’s Facebook Page @adoreboard Adoreboard at LinkedIn
