Latest / KAJ Business & Money LIVE / Data-Driven Decisions: Why Most Companies Get It Wrong | Kirill Shmidt
Transcript
- 0:00Speaker 1: H God.
- 0:34Speaker 2: Welcome to cage A Business and Money Live. Today we
- 0:37Speaker 2: are joined by Kiril Smith. He's a data and AI strategist.
- 0:42Speaker 2: He has worked across at tech, fintech, medtech, SaaS and
- 0:46Speaker 2: IoT and we learned from him why companies with plenty
- 0:49Speaker 2: of data still struggle. Also learn more about more about
- 0:54Speaker 2: why most AI projects failed to deliver miserable value, which
- 0:58Speaker 2: metrics actually matter, and how to build a culture where
- 1:02Speaker 2: insights drive decisions instead of intuition or politics.
- 1:07Speaker 3: Welcome to the show.
- 1:08Speaker 4: Kiril, Hello everyone, thank you for inviting me for the show.
- 1:14Speaker 4: Glad to be here.
- 1:16Speaker 2: You're welcome, Welcome to the show, kill Welcome to India
- 1:19Speaker 2: in this online form, and I'm sure not just in India,
- 1:22Speaker 2: but a lot of people across the globe will benefit
- 1:25Speaker 2: from what we are going to talk right now. We'll
- 1:28Speaker 2: talk about why once AI one's data isn't helping, what
- 1:32Speaker 2: are the reasons for us that, and how to make
- 1:35Speaker 2: that course correction and much more. But before that, let's
- 1:38Speaker 2: talk a bit about your journey, your background, so that
- 1:41Speaker 2: the audience knows who they are listening to.
- 1:45Speaker 4: Yeah, so I'm in data area for more than twelve years,
- 1:50Speaker 4: so I started as a financial analyst in the company
- 1:54Speaker 4: and then kind of gradually developed into data and lists,
- 1:57Speaker 4: started to using pythons, started to using different data stack.
- 2:03Speaker 4: Roughly sent seven years ago, I started to be ahead
- 2:05Speaker 4: of data created bi in medical center and then moved
- 2:10Speaker 4: to sazas business changed country. So I now I'm living
- 2:15Speaker 4: in Track in Czech Republic UH and working for American company. UH.
- 2:21Speaker 4: We are doing aot We are processing data from water,
- 2:27Speaker 4: water meters and detective links. So kind of white white
- 2:35Speaker 4: white understanding of different industries and I have worked in
- 2:39Speaker 4: different areas of the business in terms of making decisions.
- 2:46Speaker 2: Absolutely absolutely call so to understand it better. When you
- 2:51Speaker 2: say that you know data, there's no problem with the data.
- 2:56Speaker 2: It's the way people look at it or derived make decisions,
- 3:01Speaker 2: you know, beyond there, beyond what the data help us understand.
- 3:06Speaker 2: From that point of view, what do you actually mean?
- 3:09Speaker 2: Where do you when did you first notice about this
- 3:13Speaker 2: part that you know? It's something else that that is
- 3:16Speaker 2: actually you know, making the decisions and not that data
- 3:20Speaker 2: is the real problem.
- 3:23Speaker 4: Yeah, because it's how what is the real dynamic inside
- 3:27Speaker 4: the company. In many cases, then you start to work
- 3:33Speaker 4: with data, you see that stakeholders are continuously asking for
- 3:36Speaker 4: new things, and it feels like we do analysis for
- 3:40Speaker 4: sake of analysis, and in many cases a way to
- 3:44Speaker 4: cover problem of indecisiveness and inability to formulate what should
- 3:50Speaker 4: we govern and how should we govern? And then if
- 3:54Speaker 4: you do not have answers on these questions, then you
- 3:57Speaker 4: kind of try to just create work by asking analysts like, Okay,
- 4:03Speaker 4: I don't know what to do, Let's look at some data.
- 4:06Speaker 4: And that's how you create endless work and data wants
- 4:09Speaker 4: help you. So that's kind of typical, typical problem in
- 4:14Speaker 4: middle management and in sea level management across many industries.
- 4:20Speaker 2: Absolutely so, when they are spending so much of money
- 4:25Speaker 2: to get the right kind of data that they can,
- 4:30Speaker 2: you know, used to make better decisions.
- 4:33Speaker 3: How is it that the company does not have a system.
- 4:39Speaker 2: To ensure that that data is being utilized correctly? Or
- 4:44Speaker 2: is it that everyone knows about it and everybody wants
- 4:47Speaker 2: to misuse or misinterpret data as for their whims and fancies.
- 4:54Speaker 3: How should we understand which type of organizations.
- 4:57Speaker 2: Are this or are there also organizations who are who
- 5:00Speaker 2: have that right kind of use of data where they
- 5:03Speaker 2: have got the right kind of checks and balances.
- 5:07Speaker 4: Yeah, definitely, there are companies on different level in terms
- 5:11Speaker 4: of this kind of data literacy and the ability to
- 5:14Speaker 4: use data and decision making. But the root causes culture
- 5:18Speaker 4: inside the company. How company actually do people in the
- 5:22Speaker 4: company have formulated business processes and do they have a
- 5:27Speaker 4: business model in their minds? And I'm talking about model
- 5:31Speaker 4: in more mathemat mathematical way rather than like just how
- 5:35Speaker 4: we make money. It's more like operational description, what's happening,
- 5:40Speaker 4: what's going on, how we serve the clients, what are
- 5:42Speaker 4: the steps? And if company have developed that among the
- 5:48Speaker 4: managers and sea level, then they can then the data
- 5:52Speaker 4: is actually something which is aiding value to them. So
- 5:55Speaker 4: you need to start from creating order out of your
- 5:59Speaker 4: chaos side the company, and only then data could start helping.
- 6:04Speaker 4: So and it looks like company to be mature enough
- 6:09Speaker 4: utilize data properly and such specific cultures inside management structure
- 6:15Speaker 4: and to be prone to the standard understand what's going
- 6:19Speaker 4: on rather than just playing qualitic games inside the inside
- 6:24Speaker 4: the organization.
- 6:27Speaker 3: Okay, okay, let me understand it this way for my audience.
- 6:33Speaker 2: Why do these companies, many of them with plenty of data,
- 6:37Speaker 2: still make.
- 6:38Speaker 3: Poor business decisions.
- 6:40Speaker 2: Is it that their decision making is poor or is
- 6:44Speaker 2: it always the intention is poor? And why there is
- 6:49Speaker 2: if the decision making is poor, where is it poor?
- 6:52Speaker 2: Is it because their analysis of the data is not
- 6:56Speaker 2: good or is it they themselves are poor decision makers.
- 7:03Speaker 4: It's I would say, in majority of cases, it's about
- 7:07Speaker 4: managers themselves so and about kind of how they approach
- 7:13Speaker 4: their business problems the way how I see it. So
- 7:23Speaker 4: to be to become effective in the in decision making,
- 7:27Speaker 4: you need to recognize that it is a process in itself,
- 7:30Speaker 4: so that you need to think about how you decide
- 7:34Speaker 4: where to invest money or maybe what kind of product
- 7:37Speaker 4: you want to create. For that you need to have
- 7:42Speaker 4: as a manager mental model like to decide what I'm
- 7:47Speaker 4: what I do, I need to collect this kind of
- 7:50Speaker 4: information and then if I collect it, I will do
- 7:54Speaker 4: my decision. And it feels like not everybody have this
- 7:58Speaker 4: mental model inside there inside their mind. They have kind
- 8:04Speaker 4: of fragments of something and in this case, data is
- 8:10Speaker 4: not helping. So if you're ask them why, I feel
- 8:15Speaker 4: like in the majority of cases it's more about managers
- 8:18Speaker 4: themselves rather than their analysts. Because if you have a
- 8:23Speaker 4: manager with the clear vision what they want to achieve.
- 8:27Speaker 4: Usually then they can clearly state what they need from analysts,
- 8:32Speaker 4: what they need from data, and in this case, data
- 8:36Speaker 4: could just deliver on clear requirements. I see that it's
- 8:40Speaker 4: much harder in typical companies, much harder to achieve alignment
- 8:45Speaker 4: and understanding what should be done rather than just execute
- 8:49Speaker 4: on consensus inside management.
- 8:54Speaker 2: Right, what are the motives that you know drive manipulations
- 9:00Speaker 2: of the data.
- 9:01Speaker 3: In that sense if I'm using the word.
- 9:03Speaker 4: Correctly, Yeah, well, manipulation maybe it's a strong word. The
- 9:09Speaker 4: thing is that we all have our own agenda inside
- 9:12Speaker 4: the company, so we need to kind of especially we
- 9:14Speaker 4: shilion making process. We need to recognize that there is
- 9:17Speaker 4: an interest of company and then there is an interest
- 9:20Speaker 4: of specific people. So if I have responsibility on some
- 9:26Speaker 4: kind of budget, I am simultaneously try to deliver value
- 9:32Speaker 4: for the company, but also I want to play some
- 9:36Speaker 4: kind of politic games so that people around me would
- 9:41Speaker 4: keep me in power. And we're always playing these two
- 9:45Speaker 4: games all the time and not always staying in power
- 9:50Speaker 4: and being in your position is the same game as
- 9:53Speaker 4: giving value to the company. Because for giving value to
- 9:57Speaker 4: the company, I don't know, you need to accept some risks,
- 10:00Speaker 4: or maybe you need to acquire somebody, or maybe you
- 10:03Speaker 4: need to do unpopular decisions, or maybe you need to
- 10:07Speaker 4: argue against your superior because you're superior. Is I don't know,
- 10:13Speaker 4: just just are going to make some mistake. And for
- 10:20Speaker 4: many people it's rather easy to play politics games rather
- 10:25Speaker 4: to be right in terms of business value. And if
- 10:29Speaker 4: it's if you happen to be in dissipation, you start
- 10:34Speaker 4: to ban your narrative. So for example, I don't know,
- 10:37Speaker 4: you have some bad worter in terms of sales. You
- 10:40Speaker 4: need explanations. So to satisfy your your superior, and maybe
- 10:47Speaker 4: you truly don't know why it's bad, then you're gonna
- 10:50Speaker 4: involve data to explain yourself, to build a good narrative
- 10:54Speaker 4: for your superior. And that's maybe not even a manipulation
- 10:59Speaker 4: per se. It's more like bending the numbers, or maybe
- 11:02Speaker 4: bending the angle at which you see the problem, like
- 11:07Speaker 4: I don't know, shifting the narrative. So it's more like
- 11:11Speaker 4: typically it's more like this rather than blat and falsification,
- 11:16Speaker 4: because slalsification usually ends up with the severe consequences for
- 11:21Speaker 4: anybody who was involved. But more slight shifting of narrative,
- 11:27Speaker 4: maybe kind of showing data in a good angle. It's
- 11:31Speaker 4: more common things. And then at some point if everybody
- 11:35Speaker 4: in the company doing these things, like they are slightly
- 11:39Speaker 4: adjust numbers or maybe slightly change the narrative, you end
- 11:43Speaker 4: up in a completely unmanageable situation because nobody knows what
- 11:49Speaker 4: actually happens.
- 11:53Speaker 3: Right right, So.
- 11:56Speaker 2: We CEO is the boss in that sense, decision maker
- 12:02Speaker 2: and CEOs get a lot of freedom depending on which
- 12:06Speaker 2: company the CEO is, and if it's a listed company
- 12:11Speaker 2: then the board is there. But mostly CEO has that
- 12:14Speaker 2: freehand most of the time.
- 12:16Speaker 3: If I understand.
- 12:18Speaker 2: And that freehand is given so that they can think
- 12:21Speaker 2: of good ideas execute ideas. Should we understand that everything
- 12:27Speaker 2: that happens CEO is aware of that right use of data,
- 12:33Speaker 2: misuse of data, manipulation of data, you know, creating another narrative, politics,
- 12:39Speaker 2: everything else, And how should we understand that?
- 12:44Speaker 4: Yeah? So, I feel like the good thing to do
- 12:48Speaker 4: in companies to try to set your analyst aside so
- 12:52Speaker 4: they could be independent from the hierarchy in the company,
- 12:57Speaker 4: so they could deliver metrics and veilus regardless of somebody's opinion.
- 13:02Speaker 4: This is kind of quite important and that's the way
- 13:04Speaker 4: how the to management CEO could try to see picture
- 13:09Speaker 4: without somebody's narration, kind of true ground, ground truth, and
- 13:18Speaker 4: then all the people inside the company should explain numbers
- 13:24Speaker 4: which were delivered independently. So I feel like this is
- 13:26Speaker 4: the way how CEO should keep pulse on what's going on.
- 13:32Speaker 4: And after a while, I feel like everybody knows who
- 13:36Speaker 4: is try to be honest in their reporting and who
- 13:39Speaker 4: is trying to bend the narrative, especially if you have
- 13:42Speaker 4: this kind of independent reporting. But kind of in management,
- 13:51Speaker 4: I feel like people need to be a bit suspicious
- 13:54Speaker 4: and or maybe always keep in mind that there is
- 13:56Speaker 4: always second interest of any employee that they're going to
- 14:00Speaker 4: want to defend their position, So if something that is happening,
- 14:04Speaker 4: they definitely going to report it in a better way
- 14:07Speaker 4: than it is. So you need to have kind of
- 14:12Speaker 4: understanding that there's a piece of salt in each report
- 14:15Speaker 4: and that you need to and you should try to
- 14:18Speaker 4: obtain independent verification for any number or maybe any idea
- 14:23Speaker 4: which was reported by your.
- 14:25Speaker 3: Subordinated Absolutely, let's understand it.
- 14:31Speaker 2: The model that you are suggesting, is it practical data
- 14:35Speaker 2: is available? In that sense, then the data that particular,
- 14:39Speaker 2: raw data may get leaked outside your competitors, may you
- 14:43Speaker 2: know benefit from that. Regulatory things can get kicked off
- 14:49Speaker 2: depending on the data what.
- 14:50Speaker 3: The data says.
- 14:51Speaker 2: Secondly, you know a lot of organ departments within the
- 14:55Speaker 2: organization will perhaps get exposed because of what data says.
- 14:59Speaker 2: Perhaps the de garments will need to be close.
- 15:02Speaker 3: In fact, maybe maybe the.
- 15:03Speaker 2: CEO might come out as the weakest link in that data.
- 15:08Speaker 3: Is that the right way to do it?
- 15:10Speaker 2: If you were to become a CEO, what's the right
- 15:12Speaker 2: way to establish that whole data thing so that it
- 15:16Speaker 2: is used properly also as to you know, to plug
- 15:19Speaker 2: the loopholes within the system, but to actually create a
- 15:23Speaker 2: culture of right usage of data that is available.
- 15:26Speaker 3: But whit you are paying so much?
- 15:29Speaker 4: Yeah, so I'm not talking about that CEO or should
- 15:32Speaker 4: have access to raw data. So because there's quite a
- 15:35Speaker 4: lot of transformation and the process to make sure that
- 15:39Speaker 4: it's work correctly. But I mean like that, if you
- 15:43Speaker 4: if you're analysts and people who kind of produce data
- 15:47Speaker 4: and produce report, if they are the direct subordinates of
- 15:50Speaker 4: the same people who are reporting, you have higher chances
- 15:53Speaker 4: of these manipulation things because if you have marketing analysts
- 15:57Speaker 4: who is creating reports for chief marketing officer, and this
- 16:02Speaker 4: church chief marketing. Chief marketing officer then goes to CEO.
- 16:07Speaker 4: That gives quite a lot of power over analysts to
- 16:13Speaker 4: write correct meritive Yeah you know. So in this way,
- 16:17Speaker 4: if you have if your data functions, your data preparation layer,
- 16:22Speaker 4: your kind of validation and reporting is under some specialized
- 16:28Speaker 4: person like chief data officer or maybe head of Data,
- 16:31Speaker 4: you have much less power. All the people in the
- 16:35Speaker 4: organization much less power in terms of bending the narrative
- 16:39Speaker 4: because data is kind of prepared independent way, not I
- 16:42Speaker 4: mean like independent by completely independent contractors, but just as
- 16:49Speaker 4: by people whose main job is to deliver truth. So
- 16:54Speaker 4: like marketer, head of marketing is in charge to deliver sales,
- 17:00Speaker 4: there is no mandate to deliver truth. But for head
- 17:03Speaker 4: of data or maybe head of analytics, truth is the
- 17:05Speaker 4: main value, truth and trust, and in this way they
- 17:10Speaker 4: to perform their duty, they should be independent from power
- 17:14Speaker 4: of other people. I see this in a couple of companies.
- 17:17Speaker 4: Then basically your analytical department is a separate entity which
- 17:23Speaker 4: gives analysts to other departments. So in this way, chief
- 17:28Speaker 4: marketing officer or I do know, like chief financial officer
- 17:31Speaker 4: is not the superior to specific analysts and if they
- 17:37Speaker 4: have some kind of conflict over I don't know, problematic numbers,
- 17:42Speaker 4: this problem should be going to be solved between the
- 17:45Speaker 4: head of analytics and head of finance, like on peer
- 17:49Speaker 4: to peer level well rather than a subordinate superior. So
- 17:54Speaker 4: this kind of creates better balance of power inside the company.
- 17:58Speaker 4: I feel like it's if the better way and the other.
- 18:02Speaker 4: In many companies, people are trying to put data functions
- 18:06Speaker 4: just as a technical thing, like under the CEO, because
- 18:10Speaker 4: it's a kind of nerdy nerdy pipelines and writing of
- 18:14Speaker 4: the code. In many cases, I feel like it's not
- 18:17Speaker 4: also not the best structure again because this mandate to
- 18:22Speaker 4: deliver truth is going to be lost. In this way,
- 18:25Speaker 4: data going to be treated just as an item process
- 18:28Speaker 4: and it's not full picture from my perspective.
- 18:34Speaker 2: Right, So as an AI strategy, somebody who knows about data.
- 18:39Speaker 2: If I were to ask you, who is your ideal client,
- 18:42Speaker 2: who would that be?
- 18:47Speaker 4: I see that now with the I stuff. The ideal
- 18:50Speaker 4: client is somebody who wants to automate decision making, and
- 18:56Speaker 4: it's some kind of middle size company because they still
- 19:02Speaker 4: have they didn't build already extensive systems, so they have
- 19:06Speaker 4: problems to solve. I. E II gives you capabilities to
- 19:12Speaker 4: automate things much faster and much easier. So we often
- 19:15Speaker 4: say in our kind of in our community that nowadays
- 19:18Speaker 4: call is cheap or maybe cost nothing because of this,
- 19:23Speaker 4: because of the I. So you can build up much
- 19:25Speaker 4: more complicated data collecting systems and you can create quite
- 19:30Speaker 4: a lot of data products for many people inside the company.
- 19:37Speaker 4: And the biggest impact I would say for the companies
- 19:40Speaker 4: is then you have some use. You have started collected
- 19:43Speaker 4: the data, but you haven't solved all these problems with
- 19:46Speaker 4: koes because in the moment and you organize your business process,
- 19:51Speaker 4: then you set up metrics, you start to collect data properly,
- 19:56Speaker 4: you start your reporting properly. You will find quite a
- 20:00Speaker 4: lot of flow hanging improves how you can improve the company.
- 20:03Speaker 4: So I feel like the best company which could implement
- 20:08Speaker 4: the E and data stack in general are kind of
- 20:12Speaker 4: in this middle section. A companies like Facebook scale I
- 20:18Speaker 4: don't know ten thousand people one hundred thousand people. It's
- 20:23Speaker 4: much harder to change something, and in many cases they
- 20:26Speaker 4: have already their own ways to solve such a problem
- 20:33Speaker 4: and also kind of in many ways, they became so
- 20:36Speaker 4: big because they already saw quite a lot of data
- 20:38Speaker 4: issues and implemented quite a lot of data initiatives.
- 20:45Speaker 2: Absolutely, absolutely, there's so much to learn about all this
- 20:48Speaker 2: from you, and I'm sure a lot of people, a
- 20:51Speaker 2: lot of companies would want to connect with you. What's
- 20:54Speaker 2: the best way for them to do so?
- 20:57Speaker 4: The best way is to try to find me on LinkedIn?
- 21:00Speaker 4: So could you provide a link.
- 21:03Speaker 5: In the description of this of the episode, right, right, yep, okay,
- 21:12Speaker 5: so kind of it's the best way you could write
- 21:16Speaker 5: me questions and I'm really glad to answer them. I'm
- 21:21Speaker 5: ready to be in discussion about data and analytics and
- 21:26Speaker 5: the I in decision making and how to make our
- 21:29Speaker 5: businesses more profitable.
- 21:33Speaker 2: Wonderful, but this set up on this very special edition
- 21:37Speaker 2: of the Kje Business and Money Life.
- 21:39Speaker 3: Thank you so much indeed for joining us.
- 21:43Speaker 4: Thank you m