How we put sampling in SIBUR on new tracks

    And what came of it


    In production, it is important to monitor the quality of products, both from suppliers and those that we give out. To do this, we often carry out sampling - specially trained employees take samplers and, according to the instructions, collect samples, which are then transferred to the laboratory, where they are checked for quality. My name is Katya, I am the owner of the product of one of the teams in SIBUR, and today I will tell you how we improved the life (at least during working hours) of the sampling specialists and other participants in this fascinating process. Under the cutscene - about hypotheses and their testing, about the attitude towards users of your digital product and a little about how everything is arranged with us.


    Here it’s worth starting with the fact that our team is quite young, we have been working since September 2018, and one of our first challenges in the framework of digitalization of processes is production control. De facto, this is a check of everything at the stage between the receipt of raw materials by us and before the final product leaves our production facilities. We decided to eat the elephant in parts and started with sampling. Indeed, in order to put laboratory research of samples on digital rails, someone must first collect and bring these samples. Usually arms and legs.

    The first hypotheses concerned avoiding paper and manual labor. Previously, the process looked like this - a person had to write on a piece of paper what exactly he was preparing to collect in a sampler, to identify himself (read - write his name and time on a piece of paper on a piece of paper), stick this piece of paper on a test tube. Then go to the overpass, take a sample from several cars and return to the control room. In the operator’s room, the second round had to hammer the same data into the act of sampling, with which the sample was sent to the laboratory. And then write a magazine just for yourself, so that in case of something, check on it, who and when took a specific sample. And the chemist registering the sample in the laboratory then transferred the notes from the papers to the special laboratory software (LIMS).

    The problems are obvious. Firstly, this is a long time, plus we observe a duplication of the same operation. Secondly, low accuracy - the time of sampling was written partly by eye, because it is one thing that you wrote the approximate time of sampling on paper, another thing - as long as you get to the carriage and start collecting samples, it will be a slightly different time. For data analytics and process tracking, this is more important than it sounds.

    As you can see, the field for optimizing the process is truly uncultivated.

    We did not have much time, and we needed to do everything quickly, and inside the corporate circuit. Doing something in the cloud at work is a so-so undertaking, because you work with a lot of data, some of which are trade secrets or contain personal data. To create a prototype, we needed only the car number and the name of the product - these data were approved by the security guards, and we started.

    My team now has 2 external developers, 4 internal developers, a designer, a scrum master, and a junior product. Here, by the way, what are the current vacancies in general .

    For a week, we at Django put together an admin panel for the team and a simple mobile application for users. Then they finished and tuned for another week, and then gave it to users, trained them and started testing.


    Everything is simple here. There is a web part that allows you to create a sampling task, and there is a mobile application for employees, where everything is clear, they say, go to that flyover and collect samples from that car. We first pasted QR codes on the samplers, so as not to reinvent the wheel, because we would have to coordinate the more serious tuning of the sampler, but here everything is harmless, pasted a piece of paper and went to work. The employee had only to select a task in the application and scan the tag, after which data was written to the system that he (a specific employee) had taken samples from a car with such and such a number at such a precise time. Figuratively speaking, "Ivan took a sample from car No. 5 at 13.44." Upon returning to the control room, all that was left for him was to print out a ready-made act with the same data and simply put his signature on it.

    The old version of the admin panel

    Creating a task in the new admin panel The

    girls in the laboratory at this stage also became easier - now you can not parse the inscriptions on the paper, but simply scan the code and immediately understand what exactly is in the sampler.

    And then we came across a similar problem already on the side of the laboratory. The girls here also have their own sophisticated software, LIMS (Laboratory Information Management System), in which they had to interrupt everything from the received sampling acts with pens. And at this stage, our prototype did not solve their pains at all.

    Therefore, we decided to do the integration. The situation is ideal when the entire filling that we made to integrate these opposite ends, from sampling to laboratory analysis, will help get rid of the paper completely. The web application will replace paper magazines, the act of selection will be filled out automatically using an electronic signature. Thanks to the prototype, we realized that the concept can be applied, and started developing MVP.

    The prototype of the previous version of the mobile application

    MVP of the new mobile application

    Fingers and gloves

    Here it is also necessary to take into account the fact that work in production is not +20 and a light breeze ruffling the fields of a straw hat, but at times -40 and an outright windbreaker on which to remove gloves to tap on the touch screen of an explosion-proof smartphone, you do not want to. No way. Even at the risk of filling out paper acts and wasting time. But the fingers are with you.

    Therefore, we changed the work process for the guys a little - firstly, we sewed a number of actions on the hardware side buttons of the smartphone, which are perfectly pressed with gloves, and secondly, we pumped the gloves themselves: our colleagues who are engaged in providing personnel with personal protective equipment found us gloves that meet all the necessary standards, while also with the ability to work with touch screens.

    Here a little bit in the video about them.

    More feedback came about the labels themselves on the samplers. The thing is that the samplers are different - plastic, glass, curved, in general, in the range. It is inconvenient to stick a QR code on curved ones, the paper bends and may not be scanned as well as you would like. Plus, under scotch tape, it is also scanned worse, and if you wrap Scotch tape from the heart, it is not scanned at all.

    We replaced all of this with NFC tags. It’s much more convenient, but we haven’t made it really convenient yet - we want to switch to flexible NFC tags, but so far we have come to an agreement about explosion protection, so our tags are big, but they are explosion-proof. But we will work it out with colleagues from industrial safety, so everything is still ahead.

    More about tags

    LIMS as a system itself provides for printing barcodes for such needs, but they have one significant minus - they are disposable. That is, I pasted it onto the sampler, finished it with work, and I had to tear it, throw it away, and then stick a new one. Firstly, it’s not that all this is environmentally friendly (there is much more paper left than it seems at first glance). Secondly, for a long time. Our tags are reusable, rewritable. When a sampler is sent to the laboratory, just scan it. Then the sampler is carefully cleaned and comes back to collect the next samples. The employee at the factory again scans it and writes to the label already new data.

    This approach also proved to be quite successful, and we thoroughly tested it and tried to work out all the difficult places. As a result, we are now at the stage of developing MVP in the industrial sector with full integration into corporate systems and accounts. It helps here that at one time a lot of things were transferred to microservices, therefore, there were no problems in terms of working with accounting records. Unlike the same LIMS, nobody did anything for her. Here we had certain roughnesses in order to integrate it normally with our development environment, but we mastered them and in the summer we will launch everything into battle.

    Checks and training

    But what kind of case was born from a rather mundane problem - once there was an assumption that sometimes checking samples shows results that are different from the norm, because samples are trite poorly taken. The hypotheses of what was happening were as follows.

    1. Samples are simply incorrectly taken due to non-compliance by field staff with the process.
    2. Many newcomers come to production, not everyone can explain them in detail, hence the not quite correct sampling fences.

    At the start, we criticized the first option, but just in case, we also began to check.

    Here I will note one important thing. We are actively teaching the company to rebuild the way of thinking towards a culture of digital product development. Previously, the model of thinking was such that there is a vendor, he only needs to write a clear statement of work with solutions once, give it back, and let him do everything. That is, it turned out that people de facto started immediately from potential ready-made solutions that should have been included in TK as a given, rather than starting from existing problems that you would like to solve.

    And now we are shifting the focus from this “generator of ideas” to the formulation of clear problems.

    Therefore, upon hearing a description of these problems, we began to come up with ways to test these hypotheses.

    It is easiest to check the quality of the samplers using video surveillance. It is clear that to test the next hypothesis, it’s not so easy to take and equip the whole overpass with explosion-proof cameras, the knee-on calculation immediately gave us many millions of rubles, and we refused it. It was decided to go to our guys from industry 4.0, who are now piloting the use of the only explosion-proof wifi camera in the Russian Federation. According to the description, it should resemble something the size of an electric kettle, but in fact this contraption is no larger than a marker for boards.

    We took this baby and came to the overpass, telling the employees as much as possible what we were giving here, how long and for what. It was important to immediately make it clear that this is actually for testing the experiment and temporary.

    For a couple of weeks, people worked as usual, no violations were detected, and we decided to test the second hypothesis.

    For quick and detailed training, we chose the format of the video instructions, suspecting that an adequate video tutorial, which will take you a few minutes to view, will be much more clearly shown by anything and everything than the 15-sheet job description. Moreover, they already had such an instruction.

    No sooner said than done. I went to Tobolsk, watched how they took samples, and it turned out that the mechanics of sampling there have been the same for the past 20 years. Yes, this is a fairly routine process that can be brought to automatism with frequent repetition, but this does not mean that it cannot be automated and simplified. But initially the idea with the video instruction by the staff was rejected, saying, why shoot these videos if we have been doing the same thing for 20 years.

    We agreed with our PR, equipped the right guy for shooting in the video, gave him a great shiny wrench and recorded the sampling process under ideal conditions. This exemplary version came out. Then I also voiced the video for clarity.

    We gathered employees from eight shifts, gave them a cinematic screening and asked how it was. It turned out that as when watching the first “Avengers” for the third time: cool, beautiful, but nothing new. Like, we do this all the time.

    Then we asked the guys directly what they did not like about the process and what was inconvenient. And here the dam already broke - after such an impromptu design session with production workers, we brought a rather large-scale backlog to the management, aimed at changing the operational processes. Because you should first make a number of changes to the processes themselves, and then make a digital product, which in the new conditions will be correctly perceived.

    Well, seriously, if a person has a large inconvenient sampler without a pen, you need to carry it with both hands, and you say: “You’ll scan the cell phone there, Vanya” - this is somehow not very inspiring.

    People for whom you make a product should understand that you hear them, and not just get ready to roll out some fashionable thing that they don’t need right now.

    About processes and effects

    If you are making a digital product and you have a crooked process - you don’t need to introduce the product yet, you need to fix this process in the first place. The concern of our direction now is to tune such processes, we continue to collect backlogs not only for a digital product, but also for global operational improvements, which sometimes even happen to be implemented earlier than the product itself, in the framework of design sessions. And this in itself gives a great effect.

    It is also important that part of the team is located directly at the enterprise. We have guys from different departments who decided to build a career in digital terms and help us with the introduction of products and the study of processes. Such operational changes are prompted by them.

    And it’s easier for employees, they understand that we are not just going to sit here, but actually discuss how to cancel unnecessary pieces of paper, or make 1 paper out of 16 necessary papers for the process (and then also cancel it), how to make an EDS and optimize work with government agencies, and more.

    And if we talk about the process itself, we found one more thing.

    Sampling takes an average of about 3 hours. And in this process there are people who act as the coordinator, and all these three hours their phone is torn and they constantly report statuses - where to send the car, how to distribute orders among laboratories, and the like. And this is on the side of the laboratory.

    And on the production side sits the same person with the same hot phone. And we decided that it would be nice to make them a visual dashboard that will help to see the status of the process, from requests for sampling to the delivery of results in the laboratory, with the necessary notifications and more. Then we think to connect this with the order of transport and optimize the activities of the laboratories themselves - to distribute the work among employees.

    As a result, for one sampling together from digital and operational changes, we will be able to save about 2 hours of human labor and an hour of downtime, in comparison with how we worked before us. And this is only for one selection, there can be several of them per day.

    From the effects - now about a quarter of sampling is done like this. It turned out that we are freeing about 11 personnel to engage in more useful work. A reduction in wagon-hours (and the composition of hours) opens up the scope for monetization.

    Of course, not everyone fully understands what the digital team has forgotten here and why it is engaged in operational improvements, people still have this wrong impression when you think that they were developed, made an application for you in a day, and solved all the problems. But the operating staff, of course, is happy with this approach, albeit with a little skepticism.

    But it’s important to remember that there are no magic boxes. This is all work, research, hypothesis and verification.

    Also popular now: