Pair & Share: How formal Mentoring pushed REWE Analytics to a new level

Speaker 1 [00:21]

Hello and welcome everyone. Today we have a pleasure to have as a speaker Axel Budendijk, who is an astronomer who turned into a data scientist. After finishing his PhD in 2015, Axel started working in data teams at different companies, continuously learning and building up new skills. In 2022, he joined the analytics department at Reva Group as a senior data scientist. And today he's going to explain to us how formal mentoring pushed Reva Analytics to a new level. Warm welcoming to Axel.

Speaker 2 [00:58]

Yeah, thank you very much. Hi, I'm, yes, you already heard, I'm Axel Budendieck. I'm a senior data scientist at Rewe Group. And yeah, we'll talk about the formal mentoring program, which is called Pair and Share, which we started at our analytics department in summer last year. Before I get into the details, I'll run a quick agenda. And I will first give you some context about Rewe Group, because probably not everyone will know it, and about our analytics department. Then I will tell you what our motivation was to actually start a mentoring program and how we then set it up and got it running in summer and in fall last year. And I will also give you some experiences. I will share with you experiences which we collected during the first iteration then. And in the end, I will give a brief outline how, well, what's coming up next for mentoring at our department. All right, let's get started. So for those of you who are from Germany, you will probably know Rewe, I guess, from our supermarkets and from our discount stores, Penny. But Rewe is actually quite a big corporation. We also have tourism, we have hardware stores, lots of international entities. But I'm working in the analytics department of retail Germany. So there are also some smaller analytics groups and some other entities. But my colleagues and I were mostly concerned with Rewe and Penny in Germany, which makes up about half of the yearly revenue of Rewe Group. And we already said we are a few colleagues here, and I will first give you our mission statement in the analytics department. So our goal is to enable smart and data-driven decision-making in our stakeholders' key business decisions. So that's what we want to do. We want to be there from the start to the beginning. We want to help with our data and data science expertise, not just sit in the end and work on some tickets, but actually talk to people, help them be better. And in order to do that, we are now about 150 colleagues, so quite a lot, and we've grown a lot in the last couple of years. When I started four years ago, we were about 70 or 80, so now we've doubled that, actually. And you already heard, we want to use our data and data science expertise and that's why we have lots and lots of data science working in our department. We also have all the other data roles you can imagine, machine learning engineers, data engineers, we have analysts and so on and so forth. And because we're also working in an agile fashion, we have product owners as well and as we're such a big group we also need some kind of management, so we have team leads, we have head-offs and so on. And I will not go into the details of the actual data science projects because that's not the scope of the talk today but I will give you a few examples so you will have an idea on what kind of things we're working on. So if you've used the Revo or Penny apps in Germany already you might have seen coupons there. They're usually personalized and that's because we have recommender engines. We have teams that build recommender engines. That's one of the many cases in retail you can see. We're also doing forecasting because in retail we have lots of logistics obviously and again because in retail your stores you might have fraud so we're also working on fraud detection so this is just like three very basic projects and we have many many more coming so all the very standard data science things you can imagine all right and because we're doing that well we thought we want to grow a bit and then we figured formal mentoring might be a good thing and I will give you like two directions of why we thought that would help so first I will give you like a general overview of some arguments why I think formal mentoring is actually very important and helpful and then also a few examples of things that we wanted to tackle with the formal mentoring program in our department so I guess you all are kind of interested in mentoring so you might have already realized that it's helpful for example with personal with technical growth because you have someone you can trust and you can talk to and who is more experienced than you do so this might be a help then if you have like a mentor this will probably improve your employees satisfaction because you know you can grow and it also works the other way around because if you're a mentor and you like what you're doing this is also probably a reason to stay at your company at your current position. This then goes on because of course that helps with recruiting if this is something that not every company does so we believe that this will actually help us attract talents and it prepares the mentors for being manager later on because it helps you to reflect and to guide people. So those are the general arguments on the general reasons why you want to do that and then there were a few things as I already said before thought we should tackle in our department. So before we started this program, which was in summer last year, mentorship actually happened only by chance. So if you were a junior starting in a team, you know, you may be lucky and you will have like a more experienced colleague in your team, but you may not. And then you don't have a mentor and then, well, bad luck for you, right so we wanted to change that definitely then already told you we have a huge department we're 150 colleagues and that also means it's usually hard to keep inter team connections alive because we're so many people we're in a hybrid working environment so like the 150 people they are hardly ever like all in the same room at the same time so it's hard to to keep track of each other and that can actually help and I will tell you in a minute why and then there was like there are lots of learning opportunities in our in our company we have like data camp LinkedIn learning we have like internal trainings which we can use but it's all very fragmented and it's self-organized so I mean if your experience if you know what you like what you want that's pretty good but if again if you like start you're coming fresh from university this might be a a lot to digest and to figure out what you want to do there. So what we then figured is, why not start a formal mentoring program where we can try to bring people together and try to solve these things? So this is what we did. And then, of course, there are many different ways on how you can actually do that. So we were four people there thinking about what we can do and how we can shape that. And we decided we don't want to tell people exactly what they need to do when they mentor or are being mentored, but we only want to set like a framework. So we want to just give some guidance there. We'll tell them, okay, you know, we will bring you together. And then you should work on your technical skills, personal growth, maybe soft skills, career progression, you know, these kinds of things. So, of course, it shouldn't be like learning to play the piano or something, but like something involved in your professional life and so this was like the guidelines and then we would actually bring people together so we would match them and i will tell tell you a bit more about that on the next slide and we would provide guidance and and suggestions on what people would do so we would hand out documents we had like a guidebook with examples with example projects example goals for mentoring also you can have a look at it and then decide for yourself what you would like to do so once we figured that out we actually went and got this thing going pretty quickly so i think the first idea was like in june or so last year and then in august we already started the registration phase so we would ask people or we would have lots of advertisements in our group meetings we would hang up posters write teams messages and ask people like okay if you want to be part of this, then please register. If you want to be a mentee, tell us what you want to grow or what you want to learn. And if you're a mentor, if you want to be a mentor, tell us what you can offer, what you can actually teach, what are your experiences. And then we had, I think, three weeks of registration, and people actually registered and came up. We were very happy about that, by the way. And then we did the matching. So, of course, you could think, like, maybe people would want to find a mentor or a mentee themselves but we decided to do that ourselves so we would actually have a look at all the mentees and mentors that were registered and then already said we were four people organizing this whole thing so there wasn't a single participant which not at least one of us knew so we could bring some kind of social component into it so we had social components saying like okay participant a might not match with participant B, but maybe with number C. So we could think about that. And then we did also cross-team matching. So you would not be paired with a mentor or mentee in your same team, but with a different team. So you could actually have the knowledge being exchanged, and you could get to know more people or new people. And then after the matching, we did some opening ceremonies. So we brought people together in one room, got some drinks, got some food, and then, well, created some kind of informal environment where a mentor and mentee could meet for the first time. So they wouldn't have to set up an appointment, which might be awkward, or maybe you're an introvert. You don't want to set up these things because you're matched with someone so much more experienced. So this opening ceremony actually helped with that. And during that, we also suggested and highly encouraged people to create a mentoring agreement. So before you actually start the mentoring process, that you would sit down with your mentor and think about the goals that you would like to reach. So it could be, for example, a career progression, so like a promotion or race. Or it could be something like, I want to learn a new programming language. Maybe I want to learn Rust. So you would need to figure that out and then write it down. and then you have your goals which you would like to reach and you can work on that over the next couple of months. Because we, I forgot to mention that, the iteration was limited for six months. So it's not like indefinite or very short, but like six months. And then in the end we wanted to see whether this is fine and whether this is working or not. And after this mentoring agreement, people started meeting regularly. So all the mentoring happened there, all the magic, and i will go into more detail in a couple of slides um but because the six month they're actually quite long and we don't just wanna you know let it go for you know without us checking on it so we did uh regular pulse checks we did questionnaires in order to figure out how is it going like is everything fine do people still need like information and we did lots of one-on-one feedback sessions so I asked lots and lots of participants to meet for a couple of minutes and then share their feedback whether they would need something else in order to figure out everything is going all right and then just last month we did a closing ceremony again an informal gathering we had some drinks some food again and people would share their experiences so you have some kind of ending to it and then we can start the next iteration Okay, so this is the general setup, and now I'll give you some more details about our first iteration and what happened there. So first of all, we had 46 participants with 24 mentoring pairs, and for those of you who can count and do some basic math, you'll figure out, okay, there were like one or two people who were in like two pairs. You would be a mentor in one topic and then a mentee in a different topic. That actually happened. And mostly we had data scientists and machine learning engineers. But it's not only the technical roles, I mean, there are also analysts and BI developers, but also product owners came to us and said, yeah, we want to be part of that. We were a bit surprised, but that turned out to work just fine. And also we have leads and head-offs in the program, mostly as mentors, so that's also pretty cool and we like that a lot and if you remember the beginning of the talk i told you we had 150 colleagues then you can see with the 46 participants that's like almost a third of our department which took part which is pretty amazing we're very proud of that like when we first started thinking about it we were like okay like 20 participants that would be amazing and now we just doubled that so that's that's really cool and i really hope that this will continue like this so much for from the general numbers and now you might want to know like how did people actually like that so let's talk about these pulse checks I mentioned before so those are questionnaires I send out with Microsoft forms and we did that every six to eight weeks we had three of those and I will explain this graphic in a second we had three of those with a couple of questions and it only took two to three minutes to actually fill that out because we wanted a high turnout and Well, usually like 25 people replying to that And then we would ask questions and you can answer on a zero to five star scale So to make it very simple and I brought just the results of two questions with me here And you can see on the x-axis every bar is one pulse check and chronological order and on the y-axis axis you can see the star scale on average of all the participants and the the first question is the one on the top it was how helpful was the mentoring program for you so far and we're actually very happy because we got on average about four stars which is pretty good four out of five that's quite a good thing and it was very steady over the whole course of the program which we liked a lot and the second question was whether people had reached their goals so far. And you can see we started at, I think, 2.5 stars in the beginning, after like six to eight weeks. And then we had a steady progress. So it actually increased over the course of the program, which is pretty good. But we're not there yet. So the last pulse check was also, I think, six to eight weeks before the end of the program. So people might have reached more of their goals. But there might be some room for improvement there, I guess. All right, so this is the overall feedback. So people seem to generally like the program, which is pretty cool. But that's only half of the story, because in my experience, you have to talk to people to actually figure out what's going on. So I already said I did lots and lots of one-on-one feedback sessions. I would just write Teams messages, ask, like, okay, do you have, like, maybe 10 minutes? I would be interested in your feedback. And then usually people said yes, and we would meet for a coffee and talk about that. And I will bring you a couple of the conclusions from those feedback sessions. So the first one is it's not very easy to define reachable goals. So, I mean, if it's like a promotion, it's usually very clear. You can measure that very easily. If it's like learning a new programming language, like when is that ever done, right? What do I need to do to learn Rust or to say I'm a Rust expert? Very hard. So you would actually have to train there. And we will give more concrete examples for the next iteration so people will actually have a much, well, hopefully in a better time defining their goals. Then something which came as a surprise to me at least was that mentoring doesn't always mean talking, but also coding. So there were actually people starting coding projects and pretty good ones, I might say. So that was very cool. And I will actually bring you two examples on the next slide. And then the people really like the networking outside the bubble, already said, where so many people and we hardly ever meet. I mean, we have to go to the office like two days a week, but still it's hard to keep track of everyone. So people, the participants, they would meet like new colleagues they would have never talked to before, and they like that a lot. And together with that, they would actually discover new technologies from their mentors. So that's something that we actually expected, and fortunately, this actually happened. So there was knowledge flowing in both directions, which is pretty cool. And the last thing we learned was that there's not necessarily always a need to meet and talk, especially at the end of the program. Maybe you already had your promotion, or it only took you three months to become a Rust expert. Then you don't have the need to meet again. And this is also something which we will use for the next iteration. so we will decrease the time frame a bit. Okay, so much for the feedback. Now I already promised you a few examples of how people would actually do the mentoring. So I have three cases here. The first one is the traditional mentoring. This is what I usually think of when I hear the word mentoring. It's people will sit down and talk. They will talk about their daily obstacles, about personal growth, maybe how to get a promotion, like these kinds of things. And the pairs would do that. But there were also these coding projects I talked about, and I have two of these as an example here. There was one pair, which was a data scientist as a mentor and an analyst as a mentee. And the analyst was actually aspiring to be a data scientist or is still aspiring to be a data scientist. So the two of them, they would sit down for hours at a time and then built their own machine learning model. So mostly the mentee would do that. They would identify data, download it, process it, clean it. They would train the model, optimize it, evaluate it. And then in the end, the mentee had his own first machine learning model readily trained, which is pretty cool. I mean, you can do that self-taught. That is possible. I mean, I guess most of you will know how to do that. But it's so much better if you have like a mentor with you who can tell you like, Okay, you need to be careful with that or you need to check this and this so this is actually pretty cool And now we have an analyst who's almost a data scientist So that's also a pretty cool thing and I I guess those two are also very happy with the outcome of their mentoring Let me just take a sip of water All right, and the other example is there was a pair of two data scientists and And the mentee was very interested in generative AI. And the mentor is actually one of our experts in Gen AI. He was writing the Gen AI wave just from the beginning. And so they started building a REC app, a Retrieval Augmented Generation app, which is now going to production. So all our colleagues will be using it, which is just amazing. We started this mentoring program, and then suddenly you have these apps coming up, and people are using it. And this was about identifying competitor stores for our supermarkets. So you would give it one of our stores, like Darmstadt City Center Rewe, and it will tell you, okay, there's an Edeka, there's an Aldi there and there, and those are the key things of these stores. And now, as I said, it's on its way to production, which is really amazing. Okay, so we've learned about what people liked and what kind of feedback they gave, what kind of projects they did. So let me just summarize what kind of changes we experienced in our department during entering. So lots of networking going on. People were matched across the teams and they got to know new people and they really liked that a lot. Then due to this, we had lots of knowledge. So many, many teams working on similar problems, which is just the case in a huge data science department. They were talking, exchanging their solutions, their technologies. That's pretty cool. Then we have lots of personal growth. So yeah, just take the example of the analyst who is now like almost a data scientist. That's pretty cool. But also the other way around, all the mentors, they learn to reflect and to lead and actually to guide people. And this will hopefully prepare them to be like a team lead at some point as well. And those were just the big things. So there's lots and lots of many small which are happening together with the mentoring. For example, there were new exchange formats started, so people were meeting and saying, like, okay, we have so many forecasting experts here and we rarely talk to each other, let's just do a forecasting fix for like every three weeks. And now they do that and they talk to each other, which is pretty good. And we realized when we finished the first iteration just last month that a few of the pairs actually want to stay together. So they want to keep on meeting and connecting and help each other, which I like very much. So it seems like the matching did actually go well. Okay, and you probably figured that already out. I like the outcome of the mentoring program a lot, and we will actually keep on doing this. So we will start the next iteration next month, and we will try to do two iterations per year. And we will keep most of the stuff the same because people seem to genuinely like the way we did it. So we will keep the matching process, we will keep the framework and also the opening and closing ceremonies, but we will also do some adjustments. So we will start providing infrastructure for the projects, which was kind of an issue before, because you cannot always do your Sandbox projects on your Teams cloud projects. So we will do that now. We will decrease from six to four months, so people will have shorter periods to meet. I remember when we started thinking about it, we were discussing, should we do three months, should we do a year? We decided to do six months, and because we're an agile organization, we inspect, we adapt, so we changed it to four months now. And we will try to request more info on what people are actually looking for in a mentor or a mentee, so the matching will work even better. We had like a few couples where maybe the matching did not work that well. I mean, they stayed together and they kept on working, but maybe the social match was not always a given. So we will try to improve there as well. Okay, so to conclude, I actually really liked running this mentoring program. I had a lot of fun organizing it. And it's really good to see how people like it as well, how they grow. And if you don't have anything like this in your organization already, I can highly recommend doing that. And with that, I can conclude. Thank you.

Speaker 1 [24:42]

Thank you so much, Axel, for your presentation, and we have plenty of questions. I hope you will manage to cover them. One of the most popular questions is, how did you find more senior employees willing to mentor, and how was balanced the time, participation, mentoring, and some other projects?

Speaker 2 [25:03]

Yes, so we had a good outcome with the mentors who wanted to be a part of the program, but we actually, we indeed had a small shortage of those, like a couple were missing when we did the matching. So what we did was we would just talk to people who did not register, who we knew were probably a good fit for the mentees we had in mind. And then most of them actually said like, yeah, sure, sure, I will do that. that and second question was like the the time management yeah because our management was behind the mentoring all the time from the start that was actually not a big problem so people had the permission to do the mentoring and to take an hour here an hour there so and people they liked it they enjoyed it so they managed to do that

Speaker 1 [25:57]

And how, actually, the matching worked? What were the rules, and did you do it manually? Did you have some algorithms for that?

Speaker 2 [26:03]

Yes, well nowadays we usually do things like this on a Miro board, right?

Speaker 1 [26:04]

Yeah.

Speaker 2 [26:08]

So one of the four people of us we were in a remote call and we prepared it a bit so we had like all the mentees and the mentors and what they were looking for on little post-its and One of us did like a suggestion what we could think would would be a good fit and then we would discuss that because yeah like the criteria were not in the same team and that they would match like with what they were looking for and socially and So we we found a good fit I think for most of the people very quickly and sometimes we had to discuss but I think it's like the social component Which is very important. So you would need to actually know the participants in advance that actually helped a lot

Speaker 1 [26:52]

Did this mentoring relationship continue after the end of the project?

Speaker 2 [26:59]

Yeah, so for a few pairs it actually does so we heard that already like there a couple of pairs where they will continue to meet And for the others, I hope they will register again when we start the next iteration next month

Speaker 1 [27:12]

And do you already have some kind of requirements who can be a mentor or who cannot?

Speaker 2 [27:19]

Oh yeah, right. So in the beginning we said like, so we have like the three normal levels, junior, professional, senior. I think most of the companies have something similar. And we said like juniors and professionals should naturally be the mentees and then seniors and everything above should be mentors. But we were not very strict about that. For example, one example I told you about where there was like a gen AI expert. He's just a professional, just a professional, professional, right? But he's one of our main experts in that field. So he was the mentor there. And so we were very open with that. We would not send anyone away who was saying like, OK, I think I know a lot about this and maybe I'm not a senior, but I want to be a mentor. And then we would consider that. And that usually worked out pretty well.

Speaker 1 [28:06]

Could you share a little bit more about the motivation behind this mentoring program?

Speaker 2 [28:13]

Yes, so I mean I showed you like the slide already, so there were a few things that we wanted to see. So we wanted to connect people, we wanted to give the opportunity for everyone to have a mentor or a mentee, because that's not usually the case. We have different team sizes in our organization that could be that you're like only two or three data scientists in one team. Maybe you have a team of 10, 12, 13 people where it's much easier to find a mentor. And then there was my very personal motivation, I changed teams last summer and I went from a big team where I used to mentor one or two people to a much smaller team where we are only seniors, so I wanted to continue to be a mentor. So I thought maybe starting a mentoring program was a good idea.

Speaker 1 [29:01]

During your talk, you also mentioned that one of the, that mentoring program can help for a mentor to become maybe a team leader. What other benefits could you mention?

Speaker 2 [29:12]

Well, I think if you are talking to a less experienced colleague and they are asking you for career advice or technical advice, you have to reflect how you managed to solve these problems and assignments yourself, like maybe a couple of years ago, and what would have helped you there. And I think this reflection will make you quite a good team lead in the end, at least that's what I hope that if you can still remember what you missed or what you liked from your own leads and what helped you in your career then you should be I guess a pretty good leader.

Speaker 1 [29:50]

And did the mentor-mentee always meet in person on video calls? Was it mixed? Did they decide by themselves?

Speaker 2 [29:58]

And they did that.