An interview with Evangeline Leong – founder at Kobe Technologies
Singapore is well known as an amazing success story in Asia. From jungle state to leader in logistics, finance and tech in 50 years’ time is nothing short of a miracle. Despite this, its government enjoys a reputation of not wanting to rest on its laurels, because the world is changing fast and one needs to keep up. Singapore’s Foreign Minister Dr. Vivian Balakrishnan is very clear about the changes Singapore needs to stay competitive. This change is driven forward under the Smart Nation Programme, which he is in charge of as well.
During his speeches Dr Vivian will often mention that one of the most important pillars of this programme is a change towards a start-up driven economy. But for decades Singapore moved in quite the opposite direction. Corporate jobs and working for the government were considered the best, safe and secure options for the young. Now things are changing and in our view one of the prime examples of what that change looks like is a start-up called Kobe Technologies, where we had the pleasure of speaking with Evangeline Leong. Kobe are building the Google of Influencer Marketing. Well, sort of.
Kobe Technologies is located at the JTC Launchpad near One-North, one of the many buildings specifically created to house start-ups. The story of Kobe Technologies started when Evangeline was working at Purple Click. This company is one of the resellers in Asia of Google, Baidu and Yahoo. The company was founded in 2006 and Evangeline, who has now left the company, joined them shortly after they started. By the time she left she was a business director and partner in the firm.
“The whole area of Google Adwords, programmatic and performance marketing was my forte. 3 years ago, I worked on this influencer marketing project with a colleague from another agency. We realised that in performance marketing we can easily predict performance using the available tools and that makes things easy to plan. It is also very scalable. For this campaign when the influencer posts went out, the results were exponential. We had 15 times more sales than we expected. For this particular client, to get the same returns, they would have had to spend an amount on media buying that they would have never been able to afford. Yet there was no science involved it was just a lot of fun. At the same time working with the media tools and all the measurement and tracking, I realised that it also limits our imagination for exponential growth. At that point, I had a discussion with colleagues from the other agency and I realised it is really about influencer meets performance. How can you make influencer marketing scalable with a tool or platform that resembles a Google Adwords tool? That was the question we realised we needed to find a solution for.
“I rarely use the tag line ‘influencer meets performance’. A lot marketers are still not very savvy when it comes to performance marketing and don’t understand it. That is why the company is called Kobe. In Chinese this means word of mouth and in my opinion it is one of the most powerful, yet also under utilised marketing tactic around. Calling it Kobe, it makes the product promise easy to understand for Asian marketers. They understand word of mouth but not necessarily SEO, SEM, etc.”
So how did you get from there to Kobe Technologies and why and when was machine learning and AI taken into consideration?
“That is a very good question. I have a performance marketing background but I am not into tech. I never wanted to start a tech company but I ended up doing it because nothing existed that could measure, classify and put a performance metrics on influencers the same way you can with Google Adwords e.g. What I wanted was to match the most relevant influencer to a brand and take the guesswork out of the process. The selection process for influencers was just too fluffy.
If you want to find e.g. parents with babies, the only way to do this is to go through Instagram or Facebook, in short doing a manual research. Wouldn’t it be nice if there would be a better way that works more like a Google type research? Then I started doing my research and concluded that what I needed to enable that search was something called AI. This would be the only way to capture the variations you find in and between languages. It just has to work for Chinese, English, Singlish and tons of other variations.
Evangeline runs us through a demo of the platform. An interesting Google style search page comes up and it turns out that is not by accident.
“Our platform crawls through all the posts and analyses language as well as the image. In the post you see here, the machine can tell us that it is an indoor scene with a child and a woman in the image. It can also conclude that is indoors. It can also recognise that the language used here is very local because of the appearance in this case of the term ‘bak kwa’.
“After two years of learning and development of algorithms the platform is capable of recognising and scoring content on multiple dimensions. Especially the recognition of local language is important, as this further helps finding the right influencer. Other social listening tools might be able to work with big data sets and e.g. predict who is going to win the election. But that is not what we’re looking for at this moment.
So how do you get full access to this data?
“Influencers sign up with us and give us access to their profile. That allows us to go into every single post. So, in the example above, the post gives us 10 to 15 data points. Since we have 4,000 influencers each with 1,000 of posts, we have a bank of data. People sign up with us to get better chances of getting sponsorships and brand engagements.
“There is also an additional upside for them and the brands. Because we can create better matches between brands and influencers and have a number of standardised measures, we can provide guidance on the pricing of the engagement. Without this, putting a price on an influencers engagement would be a pretty random affair, there is no science behind it if you do it that way. We on the other hand, have developed an algorithm that predicts that relevance of the influencer to the advertiser and their power to engage. With this basic level of science, it is a bit like CPC bidding and it’s much more transparent. So, influencers and brands follow our price. We base our commitment and our pricing for our services for the brand to deliver on campaign KPIs.
“This addresses a point that I found very unacceptable, which is caused by the lack of transparency on how and if the influencers will deliver results for a brand.”
Which channels do you cover?
“Influencers sign up with their Instagram profile, where we crawl their data. When we help brands deliver their content, we do this across all social media channels, such Blogspot, Twitter, Tumbler, Instagram, Facebook. We do this in Singapore, Thailand, Malaysia and Indonesia and we have customers there as well. In addition, we keep running analysis of content and we try to improve our algorithms. I thought in the beginning that you develop a product, put it to work and make money but I was very wrong about that.
“Another feature that we want to develop is analysis of live data. There are very few, if at all, tools out there that are able to analyse sentiment of comments and other live events on social media.”
How does it work?
“When you set up a search so that e.g. for whisky, you can exclude terms and/or exact expressions, such as ‘beer’. Once you run your search it returns a number of parameters of the influencers that match your criteria. You have e.g. the outreach, which is the total number of followers but also a relevancy score based on our algorithms. Sometimes an influencer with less reach can be more relevant than one with a massive following.
“The AI is still not 100% accurate. You get at times very bizarre results and our developers need to figure out how to improve it.”
How did you develop the product? Isn’t it difficult to achieve product development in Singapore with all the high costs?
“We tried to work with a designer for the front end and then a developer who would build the tech. But it turned out that it wasn’t what we wanted and none of it was used. That was a lot of money wasted. We tried another setup that didn’t work. We finally settled for a sort of product manager that is able to turn our vision and dreams into a product, working with developers.
“We have been bootstrapped up until now and are making money. Our business model is a mix between agency and tech firm. We ended up somewhere in the middle and that’s actually fine. Some of our clients are actually agencies, like Havas.”
How do you work with brands and what do they want to achieve?
“We always work based on three KPIs:
• Number of influencers that we are going to connect to the campaign;
• The total outreach achieved;
• The engagements that result from the campaign.”
But how do brands measure and define success? Ultimately every brand wants to sell.
“There is indeed still a missing link. We can measure the engagements and outreach. The brand impact and tracking the exact effect on sales is more complicated. It has also to with influencer marketing. You don’t want the influencer to start telling their audience to please go and by X, Y or Z. To measure brand uplift, a specific brand study need to be carried out. There is no AI platform for that yet. But some brands have a very strong sales objective. When they come to us we will be very straightforward with them and make them aware that influencer marketing is not a sales-tool.”
Where do you see this going in the future for Kobe and for influencer marketing in general? Are there any competitors?
“I see Kobe really being a media owner through standardisation of the influencer landscape. In 5 years, Kobe would have inventory in all of Asia. Advertisers can come from everywhere however. We protect our algorithms also with patents. We applied in South East Asia and North Asia. We are in stage out 2 out of 10 and it costs a lot of money but I hope to obtain these patents.
“We distinguish two types of influencer platform competitors. There are social listening tools that are pure tech firm. Then there are platforms where you can bid on services, among which are influencers. We are somewhere in the middle, building inventory in the influencer space. For the AI component, I see this evolving to real time monitoring of content and interactions and ultimately predict what will happen. But that is the dream state.”
How do you know that your algorithms actually work? How do you judge the quality of the results?
“We look at two things. Are our influencers happy and secondly are our clients or advertisers, i.e. the brands, happy with the results. That tells us that we’re on the right track, it is as simple as that. We always want to find the right person for the right brand.”
Why do you consider the #3dollarballer campaign that is mentioned on your site so successful?
“The client could have done a mass media campaign. Instead they wanted to go through a different approach and the challenge was to find the right influencers that could generate momentum but were not coming with the high costs of celebrities. The audience was defined as millennials, students and students from SMU because the SMU was right around the corner. We managed to strike the right balance by finding influencers with enough followers to get attention. At the same time these were not too well known either so that the campaign got a credible start.”
Would you expect Google or Facebook to come forward with the same product?
“They would not be very credible in that space. For Google, this would probably go against their SEO and SEM offering and for Facebook it would be hard to offer a neutral service. So we’re better placed. The real question is whether influencer marketing is a fad and will go away. My opinion is that the term influencer is probably not correct. That might lose its value or becomes an obsolete term. But content and storytelling is not going to go away. Whatever is coming next, content will be at the centre and the Kobe engine will still be relevant.”
Do your customers come in with over inflated expectations because your platform is AI based.
“You’ll be surprised that customers are unaware of AI. Or take this example. I know a restaurant owner that would advertise his chicken as the best, because it is cooked by a robot. But the real question customers have is: does it taste better? I only care about the robot if I get good food. That applies to my clients too. They are more focused on using the tool and seeing results, than on what is it that enables it. It is curious to acknowledge this as a platform development company. But so many marketers and brands are still oblivious to tech. It is equally curious to see how much trust they have in something they don’t fully understand.”
What do you consider a weakness of your current product?
“Marketers are still choosing influencers based on popularity or name who they would like to work with. They are not familiar with this concept of picking influencers by relevance, one that fits their company’s image or personality, or the objective of their marketing campaign.
“However, Kobe’s product is very focused on pairing influencer based on relevancy to a brand and I would say that this is a weakness of our current product. As the industry is still young, it will take time to educate the marketers that pairing influencers by relevancy is better for their companies.”
To close this interview, what is the advice you would like to give to fellow or aspiring entrepreneurs? What have you learned that you think would be valuable to know for them?
“I’d say two things:
‘Goals without routines are wishes; routines without goals are aimless.’
I would say set your eyes on the stars and put your feet on the ground. Set a vision that excites you and has the discipline to work and achieve the vision. The cycle will go on.”
‘We have the answers, all the answers; it’s the questions we do not know.’
If you have a business idea or you’re struggling to get your startup kickstarted – congratulations! You know an awful lot about your idea and your industry – and I know with enough grit, you’ll find the answers. Just make sure you’re asking the right questions and working on the right ones! You will be able to find your way out.”
Editor-in-Chief: Matthieu Vermeulen