Our idea was really to have these two expertise within our team. Today, even if that in a sense is more and more the trend, it’s not so often that you have this kind of combination, because we are now 150 consultants, and all of these consultants are doing strategic consultancy side and also, the data science side.

It’s important for us to really ‑‑ because we’ll see later ‑‑ to create the bridge between the business stakes or the stakes for the government, and the use of data. If it’s really a thing, if you are able to team up both sides, you are able to deliver more value because you are not lost in translation between those.

What we do is, we work for many Fortune 500 companies leveraging their data to answer key strategy questions. Questions about their investments, questions about how to manage a customer database, about pricing, positioning.

We love working with global companies on how to adapt the methodologies from one brand to another, from one country to another. You see that we are partnering with especially Google and Facebook.

For Google and Facebook, we are one of the three or four suppliers for them for some modeling about efficiency of digital, as compared to other type of investments. We are also partnering with cities and with governments in France, mainly, because we have started a company in France.

Where we are following that initiative, and where we see that this question of teaming up and having around the table a different type of profile? It’s really a complex situation. We are not talking about that data challenge, we are talking about change management, and changing the way of thinking.

The first thing we see is that if you only take that kind of change through the technical and the technology lens, you miss something to actually make it happen.

The way we work, and what we think is the right recipe to deliver value, is always to start with a purpose or start with the business questions, start with something that is really independent from all the data questions. Then see if there is a data ecosystem, and data that is existing is sufficient to answer that question.

Be able to leverage both small and big data, as long as this data makes sense. Create also the way to give meaning to that data. Meaning that when you do data science, it’s not about taking all the data available, putting that in an algorithm, shake it, and have a solution.

It’s about defining the right KPI, so that you’re sure that you are mastering what you do, and you’re sure also that you’re able to tell the story around that. There’s nothing worse than an algorithm where you’re not able to explain the mechanics.

Then, the last point is about methodology, technology, etc. Once you have well‑defined the first two steps, actually it’s easy to implement it, meaning also that we live...the big difference between 10 years ago and now is everything is open source now.

You have to be agile enough to use one technology or the other, and to change from one day to the next. To reuse what is already existing to go one step further.

That’s big change, meaning that the big names of IT, such as IBM, or even Salesforce for software, etc. they have this way of thinking about delivering all‑inclusive solution.

Anywhere you pay high fee for a license, and you have to put everything in it. For us, our point of view is that that does not work anymore, because that’s something where a lot of failures are linked to the fact that they asked company to fit their software or their own way.

Actually you have to do it the other way around, to find a way to drive that from your perspective, with your process, the existing solution, and to make technology fit to your culture. That’s something we see a lot with companies that are brick and mortar companies and that try to digitalize their activity.

If they seem...it’s impossible for a brick and mortar company to say from one day to the next, "We are a digital company." They have to find the way digital is fitting their organization, their way of thinking, and brings more value.

That’s for us the commission to go further than just bottomless, or meaning just descriptive, and doing dashboards, etc.

What we want to do is really to be predictive and prescriptive, meaning that we want to be able to have an impact on decision making and really inform, give more insights to people making decisions. The way to do it is always being able to tell a story.

Recently I’ve heard a speech from CMO of MasterCard, and he was talking about being creative. [laughs] He was talking about media, but that’s something...he was telling about the end of the age of storytelling, we’re now about story making. That could be a good objective for our work of data scientists.

It’s something interesting to tell the story of the past, what we have seen etc. It’s even better to have tools and models to help people make better decisions. We have to be humble. An algorithm of model will never take the decision by itself. That’s something we need to inform and reinforce people with true business expertise.

That’s it for the picture and the philosophy of the company. As for your challenges and with the first contacts we had at the conference, it was really interesting.

What we see is that, in terms of technology and science, you have a really solid foundation here. To make it happen is more a matter of finding the right process, the right way of testing, but testing while building something robust and long lasting.

What we have seen, especially with clients that we’re working for since the beginning ‑‑ in 10 years ‑‑ it’s less and less a question of telling people that it’s important to use data. It’s more and more a question of how to implement it, so that it’s useful and used, actually. It’s a matter of change management.

Maybe to start the discussion, what we had in mind when we read your statement, etc. is that it’s a question of skills.

We see it, because opening offices in...We have started in Paris, but we have opened offices in London, New York, Hong Kong, and now, Dubai. We want to have in each office a majority of local consultants, and we see that we don’t have the same level and the same way of teaching science in all these countries.

That’s really important to create the conditions to have this kind of balanced profile that are experts in data, but that are also ready to apply it to the real world. That requires two main qualities, lot of curiosity and to be open minded, but also being able to step back, and to also be aware of the limits of the methodologies.

I see that in the US where either you have people that are too light in statistics, or people that are good in statistics. They think that the only truth is from statistics.

For me, the next level of a good statistician or a good data scientist is to be able to say, "OK, I know the limit of what I do, but that’s why I will go one step further, because I know how to combine that with another type of methodology to have two different clues, that I can use it, or to combine that with the expertise of someone knowing the business, or knowing the specific topic."

Meaning that we’re talking a lot of cross‑fertilizations in the academic programs you have, finding a way of having this balance between understanding soft skills and hard skills, and also being really early in the education, being exposed to real life use case.

What I see is that in France we have good engineering schools, some are for us when we are really good, because all that project during a scholarship are around real use cases, and with the ones really good at math that it’s more difficult to train them because it was only theoretical.

I think that the same...if we talk about after that business world, it’s about cross‑functional themes and creates the bridges between people that are around IT, around data, and around business. Often, one of the key success factors of all the projects we have seen is that on top of our analytical skills we wear the dedicated translators on this program.

As soon as the beginning of these projects, you have to have everybody around the table, and I think that with TCA, you have a unique opportunity of being a kind of think tank about how to use data and benchmarks, best practices, and we see that...We were talking during lunch about our offer around training.

10 years ago, it was about delivering or showing that we can use data to improve decision making, but for two or three years we have seen that it’s more and more important for global companies to have this kind of institutional training to show them how it works, and for global organization, to show what is done in America, to people in Asia, and other country, etc. to basically always enrich that kind of conversation.

I think that as a starting point...I don’t know if you have any question, but as a foundation for skills in the way to set up the ecosystem, that’s two major things, and maybe you want to share some thoughts before we go on.

I can share some thoughts.

We did meet during the OGP, and I did learn a lot from the visit. Especially around your approach — I remember the mood boards, all the innovative ways to get what you...I remember you called it "out of system data," things that’s not already collected.

In a large business, they often have a reporting scheme, but actually embedded in one or more of those domain experts’ knowledge. Without the kind of cross‑pollination attempts or the engagement to the domain at hand by your data scientists, I’m sure that this wouldn’t actually happen in the first place.

If all you speak is statistics, then you can’t get that business intelligence out of those unstructured or semi‑structured silos. We’re very much doing the same here, in Taiwan’s internal government systems, because we have a lot of organizational wisdom, embedded in those unstructured and semi‑structured data formats.

That’s one thing that I would like to maybe delve a little bit deeper, or to talk together on how to get this kind of unstructured data out of the decision‑making system, so that we can build a data pipeline as a foundation to support this kind of decision making.

Before we build the data pipeline, we have to first know what kind of water flows through it. That’s the discovery process. That’s what I’m very interested in.

Also, beginning next year, I believe... It used to be that we have this primary school level statistics basic like part of math education, and starting at maybe the 9th grade data science for people who are more interested in higher sciences and all the way to the senior high school and universities.

That means that people, who are more interested in law or maybe in other soft skills and people skills, they don’t get to learn data science. They just know a little bit of math. That creates a problem down the road, because then, people are talking with different languages, essentially, and you will have to retrain them once you get them fresh out of college.

Students basically used live in two or three different disciplines worlds. Starting next year, we’re reforming our curriculums so that we have data science, not just statistics, all the way from the first grade to the ninth grade, which applies to all the schools instead of just technical schools or polytechnic schools.

Even for people who are more into philosophy, for example, they still have some basic concept of data science and how that fits together with the field that they study. It is true that as more and more humanities is machine aided, you need to at least have literacy of how this informs your corpus data or any kind of data.

By which I mean that textual data are now like proper first-class data, not just numeric data, as was the case 10 years ago. That’s what we have been concretely doing, but of course this all curriculum change, it would take a few years before we see its effect, but we are going to that direction.

I think an issue is about data science professional services that is very common in Europe, they are easy to find. They have lots and lots of clients, luxuries, everything, but in Taiwan, we seldom see this...

Right. We see a lot of internal units doing this part of what they call new media or communication analysis, and they do employ data scientists, but they very rarely contract out to independent firms. It is true, yeah.

What we see is that...you know we work for big groups that are pretty renowned for their data scientists etc. even Google and Facebook. Google has got the best engineers on the product side, and they are not on the...so, we help them more on the marketing side to understand the efficiency of what they do.

Even AXA, the big insurance company, they have a lot of, they have a big team around their data lab, but for some questions, it’s nearly impossible to do it...because what we learn doing it for several industries, several clients, is a really relevant experience to be able to deliver it fast in another company.

The way we think about it is that because everybody wants to have things internally, because they know that no data is a real asset for their companies, the good way of working is to sometimes hire people that are specialists for that to discover innovative things and new things, always having in mind that in the end the usage of it, the refresh, etc. you have also to own it.

It’s not for us. It’s not our job to operate things on the systems of our clients, but we are still used as the first road to discover new things. After that, they have their data analysis team to do the long‑lasting job. That’s always, when you want to invest on something new, finding the right balance to be sure that you master it.

That’s why sometimes, could be on the technology’s side, you have to be sure that you manage and at least you master the architecture, etc. so that you can use it internally or is any change, because if you are blocked with one technology, it’s difficult if that technology is automated by another.

Exactly. Would you say that your choice of using mostly open‑source tools is also to make it easier for you after your initial proof‑of‑concept work or exploratory work to transfer back into the unit or the organization you’re helping with, or it’s unrelated?

No. It’s not related. It’s a choice for us to be sure that we are able to stay at the age of innovation.

You prefer to know your tools, basically?

Yeah, because if we use tools on the market, we have always a limitation. If we want to be sure that we are able to do something one step further, we have to develop our own tools. You have a lot of big companies that are not comfortable using their open‑source tools.

Sometimes, you do something of an open‑source technology really efficient, and they ask you, "No, we...let’s develop that on an IBM solution," and it costs half a million a year, but because our IT and the legal department are only validating that kind of thing.

What is important for us is to be what we call agnostic on technologies, being independent enough to take into account this type of concerns also.

Of course. I remember last time when I was in Paris we talked about the data localization part of GDPR and the privacy part, but this issue is more about the technology part.

You, of course, want to own your tools, and know how it develops, and work with the wider open‑source community, but you also want to be agnostic so that when people really want IBM or really love IBM, you can still operate within their organization limits.

Yeah, and often also because we are arriving in an existing situation. We don’t want to have a project where we say, "OK. Change everything around...Change all your technology, and in two years you, you will have something." We want to be able to deliver something within four to six months.

There is also an issue about the regulator. In the banking industry, there’s a lot of models that are trained on specific software suites that now aren’t that relevant anymore, but we still need to use them. It’s because of the regulator. It’s not because of the company, but the regulator in the banking system is applying this.

Doesn’t that add a lot of overhead to your work?

It does.

Good to know.

I have a question about the manifest that you made during the end of December, the Innovative Economic Development Plan for the information. We saw that as a draft. You made already a statement of what would be between 2017 and 2025.

Here you gave us already a first point, which would be about the education, which completely makes sense, and how it’s fast‑forwarding the implementation...

We don’t say anti‑disciplinary because that will offend some professors, but we do say cross‑disciplinary will be the default education mode going forward.

That’s very interesting. Regarding the implementation of what is done actually in the universities, in the academic, we think that there might be some things that can be done very shortly regarding machine learning and other kind of competition that could be done, held on open data ‑‑ so we already have some open data basic structure.

Also maybe have some sponsor companies, some type of these companies that want to put some of their own data to try to do a kind of hackathon, like a competition, it would be interesting and relevant. I don’t know to which extent you are willing to operate these kind of projects within this year or next year, but it is something that we’d be totally interested in.

Have you worked with the K12 level, or it’s always university‑level people? Or you don’t?

We have a partnership with one of the best engineering school in France, which is called les Mines. Jean‑Baptiste is associated with the professor there about the media issue. We have many of our consultants who give courses in both business schools and the engineering school, so more at a master’s level.

We have both these things, which is more like academic research and also like applied masters.

While we have developed it more as at the master level, also in the US have the BA level, so we are starting with a partnership with NYU and Columbia. It’s about showing how we can apply this kind of approach.

I think there’s a lot of room here. In the Digital Nation Plan, there’s three different levels. One is the K12 level, which I have talked about a little bit. Then there’s the bachelor and master level in which we try to focus on a lot of open‑source and open‑data‑based studies and curriculums.

For example, we have four levels from being an open‑source user all the way to a contributor, to a maintainer, and so on. Then we focus on some interesting packages that we think are very foundational, like maybe the Docker ecosystem, that’s one of the stacks, and then TensorFlow, that’s one of those stacks.

Those vertical open‑source stacks, we would want people to make real‑world projects around those instead of just learning one part of the stack, which doesn’t really do anything.

That’s the problem‑solving part. The other part, the third part, is for people who are already engaging one of those industries which are undergoing digital transformation, that’s the polite term, or being destructed.

We do also want to offer training courses for them. This would not be like pick a greenfield project, but how to digitally transform your existing business. That will be a much shorter term, much more focused, but that also requires a higher level of expertise from the lecturers because they have to actually understand the industry that they’re into.

We’re still formulating the detailed projects, but that’s the three main cross‑disciplinary angles of the Digital Nation Plan.

We have seen that for the last part, so in businesses, this question of changed management. When you launch a new project, sometimes you have a momentum. It’s new. You bring a lot of value because you discover new things and new big opportunities.

After 3, 5, and even 10 years, first of all, sometimes we have the same team following the projects during 10 years, but it means that on the business side or a client’s it’s the third or fourth different team following the project because they changed duties, etc. meaning that if they had another idea or everybody wants to create something new.

10 years in this kind of company, it’s a long‑lasting project. It means also that in terms of you it’s not new anymore. The opportunities are much more granular, and it’s much more a teaming up and a way of piloting it, so it’s less obvious. That’s where we think that we have two main states to make this last even longer. It’s about training.

Basically, benchmarks even from different industries are really rare. Sometimes, you talking about a luxury to a generic, so to an insurance or to an automotive company that is not luxury, it helps to think differently and to also be creative in the business where you can create.

That kind of thing, we do more and more, or even organizing workshops with different industry around the table.

I’ve seen that. We developed a program, so we have started two or three weeks ago with our biggest accounts, as a free addition to the big contracts because we wanted to develop the culture. Now we have an offer by itself around this to pick...

That’s great. So it’s one or two of your facilitators and then a workshop of how many, 20, 30 people?

Yeah. It could be 10‑20 people.

10‑20 people?

Yeah and more or less technical, depending. We have things where we are teaching the basic of data science and other ones where we are talking about use case.

That’s interesting to animate that, to show use case and then to animate workshops around what are the best use cases, how to frame it, and how to make it happen on the short term here. That’s the first thing.

The second thing is about what we deliver. At the beginning, we have decided to be a consulting company because we wanted to follow what we recommend to be sure that it’s applied etc. but more and more we see that on top of that and not replacing that. Delivering tools is really important, decision‑making tools.

We talk a lot about data visualization and more and more not just study data visualization but being able to do some smart decisional opportunities, or being able to decompose. You see that, so you have a good government, but is it evenly distributed between your different entities, and you can explain that.

We see also that in terms of ownership. For business people, it’s really important. Once again, there is something missing on the market. On the one end, you have very generic solutions that are really smart, such as Tableau, but it’s a bit too generic and not so beautiful to be used by non‑specialists.

On the other hand, you don’t have the things that are truly domain-specific.

Instead you often have a lot of ad-hoc dashboards.

Yeah. Finding a way of leveraging technologies after such as D3 and mix that to good back office of strategic data, that’s something that could be game‑changing, be it on the government’s old data or data of businesses.

Regarding what we’ve seen within the digital disruption in the company, big groups, and there are many different cohorts of people, many different generation with all their knowledge and their sensibility.

In order to fast forward this kind of change, it is important to implement new kinds of training for the youngsters, but also to have specific training for those who are already in charge to ease this process, because they cannot just do this.

In what we are doing now, for big companies, Fortune 500, what we see is it’s very easy to train the technical people, the operational, because they’re very open to this. It’s much more difficult for the mid‑top managers to move to these kind of new things.

That’s where the training that are less technical and more use case‑oriented are really affected. There could be some kind of more specific training.

That’s what we thought maybe during the lunch with the people from TCA, so the computer section seems to aggregate, and many different manufacturers who have actual, real matters when it comes to developing the products, and then displaying them all around the world be it HTC, or Visa, or Acer.

To actually have them on what would be the decision‑making, based on the actual tools, and how it’s going to be implemented within the next 10 years. I think there are some things. There is some room there to show them our way of doing things with some other companies.

Of the digital disruption training for people who already are in the industry, in addition to some key metrics, like saying 5,000 people a year, they actually need to be in one of those ‑‑ I think we target 100 companies per year ‑‑ to identify about 100 companies a year who could really benefit for this kind of training.

In the Digital Nation Plan, we are working on solidifying the implementation. Is it per industry sector? Is it better to have, as you said, some aggregation within a larger sector? Or whether it is actually better to have some cross‑sectoral training workshops, as you mentioned?

This, I think, we haven’t yet finalized and that we will welcome advisers. Actually, we should follow‑up over email, because I do have another meeting now. This is something that I am personally very interested in, because for a lot of existing industry here in Taiwan, for the past few years, they did not have that much motivation to adapt to digital transformation.

One part because, of course, a decade ago mainland China was not yet this all‑digital marketplace. Now, mainland China is being digitalized rapidly itself. Also, now that we have this new south‑bound strategy, where we need to link to the other Asian hubs, which are all aiming to become data‑first economies now.

It really is imminent for all the existing industries to embrace this kind of data‑based decision‑making. I’m personally very interested in seeing this happen.

Probably we are able to give you our feedback, and our experience about how to manage it, and when, for what kind of topic it’s good to have different industries, and where it’s more important to focus on.

Exactly.

We have seen that, and actually, it’s really important to do both sides. To have things about technical, but also, because digital, we are talking about everything about the Internet, but also about data. Both sides, we are not talking about the same issues, or the same questions.

When we talk about new digital tools, etc. we are talking also about more marketing, or interpersonal relationships...

Or even experience design. These very soft kind of science.

Yeah, because what we see is that many industries are shifting from product‑oriented to service or customer‑oriented. I think this kind of conceptual training is also important, and then it can leverage data too.

People who are already on top management, we don’t want to teach them how to run an algorithm. We want to show them that they have a toolbox, and to be good at briefing specialists up there.

Exactly, asking the right question.

That’s all I have here. Anything you would like to ask? [laughs]

Excuse me. We know that you have a branch in Hong Kong, right?

Yes.

For sure that you will see Asia is a market for your company, in what kind of way that you are trying to go in this market? Can you share something that...?

Sure. Currently, we opened the office in Hong Kong two years or two years and a half ago, because we believed that there was a lots of potential in Asia. At least one third of the business of our clients, which are global companies, was in Asia.

That was obvious that we should go there. That was the rationale to close, and to leave in the region where there was lots of potential.

You choose Hong Kong, because the product, the service there, they were focused on the luxuries?

Exactly. We had the conversation during lunch, where we should place the regional headquarters. In the initial scope, there was Shanghai, there was Singapore, and there was Hong Kong. Based on maybe image of the cities, but more mainly, where are located on continents.

Hong Kong stood as the commonalities for most of our clients, luxury, and insurance and banking. That’s why we decided to set up the office here. Then we are currently operating in the full Asia. We have ongoing projects in Japan and Korea, in mainland China, in Taiwan, of course, in Hong Kong, and Singapore.

It’s really all the different locations where we have business. Now, where do we see the potential and the growth, and in which industry? As Audrey mentioned, clearly, there is the digital transformation that happened in China is massive.

On the social media, of course, when more and more traditional industries are moving toward digital, so that there is a huge energy and momentum happening in China around that. Taiwan, you are firsthand experiencing, of course, for your best players.

Clearly, there is lots of potential here. China is more open‑minded for foreign companies to work in China. Some other markets, like Japan and Korea, are much more complicated to enter, because they are much more closed.

I don’t mean on the language, but by culture, and also by the measure of the ecosystem of companies. Clearly, China is the direction where we are very positive. When I say China, of course, it means all the countries that have huge business with China.

Taiwan will be a firsthand country, Hong Kong, obviously. That’s clearly for the location. For the industries, we at the administration, we work in very different industries. We don’t make any limitations.

We believe that all the traditional industries are going to make the transformation. We have many experience to help them for the past 10 years to make that transition from changing the way they take the positions and their relationships towards people.

Helping either to accelerate on the description, or to defend, because there could be two strategies, some brick and mortars, or traditional industries want to protect against the destruction. They want to fight, again, to be Uberized.

Some, on the opposite, have much more strategic thinking, and they want to go towards that, so we can go in the growth directions. Does it make sense?

Yeah, thanks.

That’s great.

I say that the main users of data in the business field traditionally was consumer groups, companies, and knows they are late in terms of innovation.

When we were looking at the beginning, we were pretty specialized in doing this kind of project outside consumer goods, because in terms of the goods, the market was saturated with companies such as Nielsen or other market‑research companies.

We have learned how to develop things in a more rudimentary environment, and to make it happen. It gives us a good competitive advantage, because we know how to work in a rudimentary environment, create strategy that associates from small data, create clean data, understand the new sector, etc.

Now, these kind of companies, such as Coca‑Cola in the US, they ask us how to make a revolution about market research and on data in consumer groups. Meaning that we, for sure, it’s easier for us for B2C sector, because B2B depends really on the structure of the market.

It’s really a grand statement, but for any B2C company, I think we can bring something valuable, even when all is not traditional as media, etc. Probably, there is something to create here in this industrial network.

Thank you. Perfect.

As is our customs here, are you OK with us publishing the transcript of our conversation?

Yeah, sure.

You can find it at track.PDIS.tw maybe a few days afterwards. I will send you a draft, so you can fix my typos...

...before we publish it.

Thank you. It’s always a pleasure. Thank you.