Future single B2B market or game in technology?

    Hello, Habr! Today we are sharing an interview with the Director of Digital Technologies of the RTS-Tender company, which organizes electronic trading using blockchain, machine learning and big data analysis. Let’s ask about all aspects of the business, and most importantly - about the technical part. Join now. Well, under the cut - the text version.

    Digital Transformation Series

    Technology articles:

    1. Start .
    2. Blockchain in the bank .
    3. We teach the machine to understand human genes .
    4. Machine learning and chocolates .
    5. Loading ...

    A series of interviews with Dmitry Zavalishin on the DZ Online channel :

    1. Alexander Lozhechkin from Microsoft: Do you need developers in the future?
    2. Alexey Kostarev from Robot Vera: How to replace HR-a with a robot?
    3. Fedor Ovchinnikov from Dodo Pizza: How to replace the restaurant director with a robot?
    4. Andrei Golub from ELSE Corp Srl: How do I stop spending a ton of time shopping?
    5.Vladimir Sveshnikov from Robot Vera: How to replace HR-a with a robot? Technical part.
    6. The future single B2B market or game in technology?

    Who is the interview with?

    Dmitry Zavalishin is a Russian programmer, author of the OS Phantom concept, organizer and member of the program committee of the OS Day conference, founder of the DZ Systems group of companies. In 1990-2000, he took an active part in the creation of the Russian Internet segments (Relcom) and Fidonet, in particular, ensured the transparent interaction of these networks. In 2000-2004, he was responsible for the design, development and development of the Yandex company portal, created the Yandex.Guru service (hereinafter - Yandex.Market). Read more on the Wiki .

    Vladimir Grigorenko- was born in 1969 in Brody, Lviv region. In 2013, he completed internships at North Easten University (Boston, USA) and at Aoyama Business School (Tokyo, Japan). From the same year, he began working for the company RTS-tender as a deputy general director. Currently, Vladimir is working as Director of Digital Technologies at RTS-Tender.


    Vladimir, the word blockchain is now well-known and everyone says different things about it. There are positive connotations; there are negative connotations. It is so, a little, raped fleur buzzword, which is used in the case and not in the case. We have already discussed that blockchain is actively used in the technologies of your company. Can you briefly say what business value the company sells on the market and where is blockchain located in this business value?

    Well, firstly, I want to note that I absolutely agree with your position that this fashionable word is used today both in context and not in context, as young people say “on hype”. But we are practitioners, we solve specific practical business problems. I’ll make one amendment that we are talking about the RTS-Tender Company not just about one company that participates, conducts, actually ensures procurement for government orders, we are talking about a group of companies. And the tasks that we solve are the tasks of the group.

    From here, accordingly, that business need was born to close some technological problem in the group for the exchange of key information on counterparties. This, in principle, of course, is due to the regulatory framework that we have in the group of companies, because on the one hand we have a very regulated state order, on the other hand, absolutely commercial tasks, because we are a commercial company. And this, in principle, necessitated the use of this technology in a group.

    That is, your main client is, in fact, government procurement, which, on the one hand, puts out tenders on your site, and on the other hand, these are suppliers who are fighting for these contracts. Is this the main landscape?

    This is the main landscape, yes. This is a global B2B market. No matter how paradoxical this may sound, it’s probably the state order and purchases of state-owned companies - this is probably by far the largest B2B market in Russia, which, in principle, if it is well developed, it probably can turn the economy around, so to speak.

    It is interesting that with this position no one has yet voiced this matter before you in my memory. What is the real structure of the state order, the structure that is built around FZ-44, it is a B2B market. And it’s very interesting, actually. Yes, if you look at the turnover, then it is obviously the largest in the country.

    On the other hand, if we take the traditional market landscapes, then not only the bidding platform itself is present in them, where the supplier and buyer are fighting for the price, but around this there is still some kind of infrastructure, navigation. Here I, for example, as a supplier, would be interested to see who buys the things that I produce. And in the opposite direction, it would probably be interesting for the buyer to expand his circle of ideas about sellers. Is there anything on this subject in your picture of the world?

    Of course, there is, simply commenting on the thesis that this is the first time you hear such a position, I suggest just looking at any, in your terminology, supplier, and in our terminology we call him “participant”. That is, any participant who comes to us or thinks to come is an ordinary commercial supplier, who may work not in the framework of the state order as well. Yes, we have 250 thousand suppliers who, with some frequency, do something in this market.

    If you look at it simply as a commercial company, which, in order to offer something, conducts some kind of activity, including it must provide itself with something: purchase components, services, and so on, and so on , etc. Then, probably, this position becomes clear. And our goal for today is to work with participants, as we call them, in order to draw them into this landscape of ours, into our ecosystem, so that it comes to us and, in principle, all of its business tasks can would be there to decide, yes, when we are talking specifically about the supply, or the creation of needs.

    For what you just said, yes, that is, roughly speaking, you are now saying a simple thing: we have a government purchaser who only purchases, there are suppliers, but they also procure, and if you drag them into the system as buyers, it will form where more closed and much more interesting picture. And this picture, obviously, should be weighed through a very simple KPI. Generally speaking, you should probably have a ratio of purchases by state and non-state sites. Is there such a figure?

    Well, there is probably no such figure today. It probably exists in the minds of shareholders, so we are moving just from the government procurement market to the B2B market, to the commercial market, and this is understandable, because, once again I say, the scary and frightening word “public procurement” does not need to be taken literally. And for the main players in this market, whom we know very well, here. Just the task is simply to ensure that the state order is held so that the state customer publishes its need, and the participant comes and buys something - this is already a completed stage. I think everyone understands this, and this is probably a situation two years ago, if not more.

    Therefore, everyone looks at it as a large commercial sector, which can be of great interest to players in this market. What we can, answering your question, to offer, including to the analytics participant, is far from a secret. There are certain analytical tools that there is a BI system that we give to our participants and our customers to analyze this market, and they can, in principle, analyze this market and see which sector they are particularly interested in. This is, on the one hand, such a traditional instrument. On the other hand, I emphasize that we are still a high-tech business, so today there is such a predictive system that the participant, based on the clustering of his interests, which we fulfill - we group, we begin to offer him something specific, what he might be interested in. Maybe he doesn’t even understand it himself, did not think, does not realize.

    That is, he bought it somewhere, apparently, but not from you?

    Perhaps it doesn’t matter where, on which site. We collect all possible information on its activities and based on this information we begin to build some relevant proposal using artificial intelligence (AI) and machine learning mechanisms.

    Here, maybe, a little deeper. How does this work in terms of a picture of the world? What technologies do you use? What do you analyze? How does this work in terms of analysis?

    Well, as you said, of course we analyze his participation in all purchases. This data is open, there is a single information system (UIS). This is understandable, this is the main data source. Further, we, of course, use some kind of behavioral model: what he did, what kind of purchase he was interested in, he downloaded the documentation, he did not download the documentation, looked through whether he was preparing an application. Based on such an enrichment of data, we begin to refine his proposals. And, of course, we, in order for the entire machine learning technology to work, we analyze its reaction to our proposal and, accordingly, the model itself self-learns and issues more and more relevant offers.

    Indeed, in order to decide, you must also analyze the very essence of the tender documentation, which he downloads. Indeed, it is one thing if he downloads the documentation for the supply of milk, and another thing is for the development of software. These are very different things. That is, generally speaking, you should analyze the texts of the tender documentation and make some assumptions on the basis of these texts.

    Of course, this task is being carried out, we have a wonderful “star” product, which, if I remember correctly, has about 15 thousand unique visits per day, where parsing of this tender documentation is done and based on this parsing, including, the analysis is carried out and precisely those proposals that we make to the participant are specified.

    Are these neural networks too? What technology is behind it?

    Yes, of course, including neural networks, some standardized libraries, and our own developments are behind this. I just want to note right away that we always probably use hybrid technology. This is caused by different points: somewhere, I repeat again, this is the regulator, because we have a Persian there. data and confidential data that we are forced to completely depersonalize and use in calculations and which we cannot store outside the perimeters of our protected circuit and send to cloud solutions (in Azure, use Batch for this, Microsoft), somewhere we can, somewhere- then we can’t. Therefore, in all, in principle, technological products that we use, we follow a hybrid such scheme.

    That is, when you use Microsoft solutions, do you partially depersonalize the data?

    In some tasks, yes, which require large cluster computing, for which we use Batch, we, of course, anonymize the data, rebuild the sentence on the fly, and analyze the student’s behavior right now.

    So is this a real-time process?


    You are in real time. And what are the real-time parameters in general if we discuss the work of participants on the procurement portal. Is it seconds or milliseconds?

    Now these are some minutes or tens of minutes maximum. We went iteratively, that is, they were models that were being rebuilt there and proposals were formed once a day, for example. Then we, accordingly, reduced the period using the computing power of Batch. Now we are building our solution based on Hadoop. including hybrid schemes (as I said), the interval is reduced.

    That is, the cloud you buy mainly in order to get peak performance?


    And if we return to procurement, what is going on there? How sophisticated is this technology platform? What is the volume of general transactions taking place there, and are there any difficulties of a purely technological nature?

    Of course, so that everyone understands that today the electronic trading platform (ETP) is not some kind of monolithic software, it is installed on one computer, which is located somewhere under the table.

    Yes, in a corner somewhere.

    Although we know such examples, of course, in our market. Today it is a very serious highly loaded system. We provide 24/7 performance, we have such a regulation. Has x2, x4 duplication. According to the loads, of course, today we conduct 12-15 thousand trades per day. If about 10 participants go to each auction, then you can imagine what a load. This is the first situation. The second situation is that for the last few years we have had a situation where participants also grow with the market (and the market grows, it grows smarter, it becomes more intelligent, too, the participants, respectively). Therefore, robots are increasingly used in trading sessions.

    From the participants?

    On the part of participants, yes, who, at certain boundaries, participate in tenders. Of course, this causes an additional burden on the ETP. We know, see and monitor by various means - both using our own and using previous vendors. We see this load, we track it.

    That is, on average, you see which of them is a robot, and which is a person?

    Yes, of course, and by the behavioral nature to understand other technological methods to see this load.

    It turns out a funny situation. You are doing some kind of behavior analysis, which, generally speaking, suggests that there is a person. At the same time, a person is replaced on the other hand with a robot and it seems pointless to analyze his behavior. Or meaningful after all?

    Look, we are just analyzing human behavior in terms of work before the trading session. That is, we are interested precisely in order that the participant comes to us and does not leave, in the good sense of the word.

    That is, some phase when the supplier determines whether it is a tender or not, and this is human activity? And then the robot comes and further trades?

    Oh sure. And this is not as it is not legally limited now, therefore these robots are actively used. And this creates an additional burden on the ETP, and we are successfully dealing with it today, but this is the situation that we have in the last few years.

    And in the analysis, when, in fact, a potential supplier chooses tenders that are interesting to him, do robots not apply now? Although, it would seem that this is also, in general, such a place where neural networks are able to answer the question. Well, as a rule, the supplier has a certain class of tenders that are interesting to him. And to get at least the primary filtering would be great, I think.

    You know, a decision in the forehead, it is not always, probably, the right one for the participants. Because, as a rule, the participant already understands, it seems, what kind of purchase he is making (on the one hand). On the other hand, as I said, if we turn the situation around when we make him an offer to participate somewhere, he may not know what is nearby, in some adjacent subject area (adjacent - both subject and geographical) his customers.

    So you do it?

    We do, and I say, we turn the situation around, because, as a rule, we also know the whole practice. In large corporations, this is generally understandable - there is a list of approved suppliers. And, often the buyers who work, they just look very narrowly at the situation and, in principle, do nothing else. This is understandable ...

    There are some established channels.

    Of course, this is an additional load, additional resources, and why change what works so? On the other hand, if it costs nothing for a participant, and we give him very good information about looking a little wider, then why not? And we try to just stir up this market in this way.

    That is, this is your movement from pure tender trade to a more marketplace format.

    Of course, and I will repeat once again that in its pure form ETP as a trading system is not what it was yesterday, it is the day before yesterday. And all the leading market players probably understand this.

    If we are talking about a global marketplace (I really liked this idea), it really has great value. Because from my personal point of view, in fact, the b2b picture in Russia is very unfinished. It clearly has a large number of holes, a large number of companies have not yet been entered into it, and it is still patchwork - somewhere it is lined up, and somewhere there are probably such failures.

    There are many tender sites, and therefore a rather strange situation arises - different companies are tendering at different sites, and, in a good way, the supplier must track all the sites at the same time in order to have the whole picture of the country. And this is such fragmentation that seems counterproductive and exactly perfect in relation to the supplier. It’s clear with customers - it’s convenient for them to work on the same site, especially if she understands its specifics. Is something being done to glue these sites, to make them a single access point? Do you like this site, plus or minus? How do you look at that?

    It’s being done, this is understood by our legislators and our regulator, including 44-FZ, which introduced the changes that will come on July 1, and which speaks of exactly that. Some lists of electronic platforms will be formed, certain criteria will be applied to them. Including how it sounds in the project (as Kommersant wrote about this today) that the Ministry of Finance suggests that if the site has less than 10% of the market, then it is excluded from the list. That is, such a kind of centralization of the market (first) occurs. Second, participants will have access to the sites and their accreditation will be carried out in the UIS.

    That is, one accreditation?

    Yes, that is, all will have one accreditation. I’ll open the curtain for you, I’ll say that the idea of ​​just using the blockchain within the group came exactly from the same place, because we perfectly understood that a participant, even under the 44 Federal Law, in order to participate, should monitor, for example, 5 sites; accredited at 5 sites - spend money, time, and so on, and so on.

    Still, these are the stop factors that scare the participant away from this big global marketplace. And we, just understanding these stop factors, wanted to solve these problems, so that at least within the group our participant would not have such cordons, and once he came to us (the group has a certification center) and accredited, he would not repeat this procedure and could, having logged in, go in, and no longer understanding how it goes inside - where is the fin. products where tender products are - for him it should all be completely transparent.

    And how is this technologically implemented?

    We talked about this, it is implemented on the blockchain. We built a blockchain within the group, and each technological product has its own node. It is clear that there are public sections that are available to all participants. And this is a closed consortium, it turns out. Respectively. And then we, segmenting the information received at the time of accreditation, using the mechanism of GOST encryption (this is our technological know-how - I do not know any analogues on the Russian market so far).

    That is, the blockchain structure itself uses exactly GOST encryption?


    Is this block binding in a chain?

    Yes. Accordingly, in this way, we KYC information, transmit to everyone within the group. Let's just say that to both business users and technology products. Including the node stands next to our data-lake, which simply, accordingly, sends this data to the data-lake, which are then used for the recommender system as well. That is, we have solved such a technological problem - we removed all restrictions for the participant so that he would seamlessly switch from one product to another and not see this border at all.

    That is, from the point of view of the system, participants are independent legal entities that belong to different areas of the legislation, roughly speaking, they meet different requirements in terms of state licenses. And at the same time they are covered by a single system, and one participant can work with them all?


    And what additional services are generated through such a picture? The main service is understandable - this is the trade, in fact, on the site.

    Yes, but the business process is so structured that today, in the monetization model, the participant needs to make collateral, he may now have free cash, and, conversely, may not. And we can offer tender loans, we can offer a bank guarantee. Moreover, the task today is that we should not force the participant to perform the same actions several times: again fill out the application, send somewhere a separate business process ... We all must do this in advance.

    That is, roughly speaking, as soon as he went into your blockchain authorization of yours, the whole further picture will not be transparent to him?

    For him, it will be transparent, but for us we already understand this participant.

    You have a certain general picture of him: where he went and how much he asked for security ...

    As much as we can, not what he asked, even if he did not ask ....

    Do you estimate his credit history on transactions?

    Of course yes. We are also the operators of federal electronic document management. That is, understanding that we have primary information or open information about this client, we also use this in our analytical and scoring models.

    That is, if he also drives the document flow on payment transactions through you, do you still know something else about him?

    Of course.

    An interesting question, but do you have the right to know this?

    Of course, we are only talking about public information.

    Which by definition ...

    Of course. In any case, information that is encrypted or has a certain storage period before it becomes open, it will be closed. Of course. But it just enriches our knowledge about this potential participant. And the task is just to be one step ahead of his needs.

    Can I evaluate the topics in which she works on the basis of operational transactions of the company?

    Of course, according to descriptions, by counterparties, we can just this global b2b market: we understand its interaction with other counterparties. Accordingly, we, too, can analyze and understand what clusters, economic segments are connected between them, and already offer it. Here he is a participant, on the one hand, and there he is an ordinary buyer.

    That is, having some such count of counteragents, you can predict what he is doing, what he may be interested in and answer the question, how much can he be given on credit for the next state contract, if you exaggerate at all?

    Yes, because participants also have different financial literacy, and the situation may be such that it is more profitable for him to take a bank guarantee or a tender loan than to take money out of circulation. He does not think about it, but we know it well, we calculated it well, and in the same way we help him to earn. In general, the global goal is to help the participant earn money on the contrary.

    In general, say a terrible thing: you say that fin.dir. a small company understands its picture worse than the services around it.

    Why not?

    Well, I understand that this, in general, is probably good for business in general, but it also means a very interesting thing. As, for example, in the taxi business, the taxi driver himself, from an obviously central business figure, became, in general, a pure performer: some external systems that analyze large models of behavior, his model, riders, someone else, and from this is built by his behavior. What you just said is, in fact, the same thing, only in relation to small business. It turns out that there is a neural network sitting outside, conditionally, which knows more about this business than he does. And she begins, in fact, to lead him. If you extend this line a little, it turns out that you can remove the CEO, Fin, from this business. dira ....

    I would replace the term “guide” with the term “help”. You understand, this may be for now, but for the current day our task is to help him by lowering his castes and expenses somewhere, and offering him a wider market, because he does not see it.

    That is, in fact, do you know better than him the price of money inside the company and outside the company?


    And you can tell him that it is more profitable to leave a working resource there or get a loan at such interest. And this, in fact, is very interesting, very valuable. Because if we look at the country in general, then this is some global increase in efficiency. Even if it will be 0.5% in the whole landscape, then, in general, it is very valuable.

    Well, the goal of this procurement market is for state-owned and state-owned companies and to create a competitive and efficient market. On the one hand, we are increasing competitiveness, on the other hand, helping just the participants, lowering their costs, helping them earn money.

    If we are talking about secondary services: bank guarantees, lending, then today these services are represented by members of your consortium, that is, are these, conditionally, internal services?


    You did not consider the picture when they also become external?

    At the moment, there are such models in business, with the involvement of third-party financial organizations.

    That is, roughly speaking, to provide a marketplace for secondary services, this picture is also being considered.


    Nevertheless, I want to ask in this place - the standard, typical problem of the marketplace of the service market is the quality of services. If we attract a partner whom we sell in the market as a service, then the first question that we ask ourselves and the second that the buyer then asks us is why he worked poorly or whether he worked well. How to deal with this? If we attract, say, a partner for a bank guarantee. What to do to say "yes, this is a good reliable partner."

    You are now specifically about the partners who ...

    Which represent exactly financial, financial and technical services.

    Today it’s such a manual mode for adding these partners. I’m not ready to say how it will be further.

    Actually, I asked a question with an approach. If you have good models that value the buyer. Is it possible to deploy them towards the supplier? Is it possible to build a model that makes assessments at the big data level. Probably, they can be deployed in the direction of the potential supplier, moreover, can they can, in fact, be deployed in the direction of the buyer? Generally speaking, in the direction of the buyer, to say "we know that buyer, this is the kind of problem with him, but this is good ...". Was such an opportunity considered?

    Today, from the point of view of those people who form the need, these are buyers. Here, everything is more or less stable here, because we have a state-owned company and a company with state participation. There, you know, there are different ones, but there the regulation is such that, in principle, the supplier is sufficiently protected in this regard. If we say just from the point of view of the supplier and the protection of the buyer, then here we have a technology that has the internal name “module for the execution of contracts”, where we are just this one here, sorry for the term, tail after the conclusion of the contract. Because, in principle, the area of ​​responsibility of the ETP ends at the conclusion of the contract.

    Where everyone begins, you end.

    Yes, and this is only the middle part of the entire business process from the formation of needs to the end of contract execution. And we, of course, are going aside now from the moment of concluding and fulfilling obligations under the contract, we give buyers access to this system and the entire course of contract execution in it reflects these data.

    And how you get them is hard.

    We have a structured contract that has certain stages, obligations ...

    That is, if a conclusion passed through you, then you ask ...

    Yes, this time. Accordingly, the entire primary organization, which confirms the fulfillment of certain obligations under this contract, is just the EDI (electronic document management), which we spoke about, in which we see all the documentation. That is, in principle, we know even more about our supplier / participant, how well he fulfills his contracts.

    On the other hand, let's be real people. Now the scandal with IT Consulting is well-known. We all understand that in reality there is a significant gap between the documentary component of the contract, even the state one, which is rather strictly regulated, and the actual picture. And everyone who works in this market knows that this is such a typical gap. It is not that it rarely happens. It happens if not in 100% of cases, then in 99. Can this be analyzed?

    This is probably just the problem of large contracts. But we understand that if we look at analytics, then in 70% of the market these are contracts up to a million rubles. Everything is simpler there and this information ...

    That is, how is the supply of soda - it was delivered, received money and dispersed.


    That is, such commodity is just a picture. That is, these commodity services are the main interest for you?

    Of course yes. This is life, this is a b2b market. For us, one contract for a trillion rubles is not interesting. From the point of view of filling information with information, it’s also indications of influence ...

    Yes, and it’s quite specific as a rule, it’s very difficult to put into some model and from the point of view of its model to predict. Good. Our time is coming to an end, and once again I will roughly outline the landscape as I realized it. You, probably, in a nutshell confirm or supplement. If we talk about the technological picture, then it consists on the one hand of subsystems based on neural networks (as I understand it, these are Microsoft technologies), which are mainly engaged in analyzing the behavior of players in the market, predicting their behavior and based on these predictions A referral system that attempts to increase the connectivity of players within your network and your partners.

    On the one hand, to increase connectivity, on the other hand, help our member to reduce their costs, help him earn money.

    She also works not only on offers, she also works on services, and from all this you have certain chances, purely financial, for him to ensure a certain level to which he would probably not have reached right away. This is one side of the picture. The second side of the picture is a blockchain-based distributed transaction system that stores the main transaction flow through a system that provides centralized authorization of your counterparties, given the signing of contracts with them and the translation of these contracts into some kind of partnership, which consists of fragmented legal entities.

    No, since you and I touched on the blockchain, we are just distributing information about the counterparty to them about technological products and our people.

    Through the blockchain?

    Yes, that is, the whole transactional history of his participation in the auction ...

    She is not present there?

    No, not present. We just shoot, I repeat, we solved the problem of centralized SSO (Single Sign-On) there, increasing the fault tolerance of the system, giving it information, forwarding it there, and delimited access.

    That is, the blockchain serves as a tool for storing the history of counterparties in the first place, and in this case it is a source for these individual companies on the basis of which they can analyze ...

    ... companies and technological products within the group, yes.

    Some business units that are based on this information ... That is, roughly speaking, the blockchain knows everything about your client and is used internally as a source of information for scoring, decision making and partially closes the workflow so that this player is not forced to each come to a separate partner with a new set of documents and sign separate agreements.



    Oh sure. Our task is to stretch as wide as possible in this regard. Today, our most distant point - this is a certification center - where the counterparty first came when he came to receive an EDS, and we offer him to immediately receive an EDS, accredit so that he does not have to repeat this procedure afterwards.

    Also popular now: