App campaigns that achieve maximum reach and impact are obviously a huge win for marketers. But they’re often associated with a steep learning curve when it comes to choosing the right KPIs or rejecting the right traffic. We get an insider’s view on this and more from Mikhail Biteryakov, Traffic Manager and Analyst at Joom, a fast-growing e-commerce marketplace headquartered in Russia. Mikhail, who goes by Mike, speaks candidly about his personal journey from a focus on UX to a passion for data-driven marketing. Mike, a Mobile Hero recognized for his app marketing achievements, provides his pick of strategies marketers can take to achieve success and the approaches they can use to fight mobile ad fraud and win. He also talks about the future of marketing and why dynamic creatives are the one to watch.
So we continue our world tour and I’m really excited because for the first time we have a Mobile Hero from Russia, I mean, really from Russia with us from Moscow today – Mikhail Biteryakov, he is Traffic Manager Analyst at Joom, and Mike, I’ll call you Mike, welcome to Mobile Presence today, all the way of course from your office in Moscow.
Hello there, Peggy, it’s an honour to be here.
It’s great to have you here too and for our listeners who don’t know it, I mean, Joom is quite an interesting company, I was researching it – somewhere between a Russian Amazon and a Wish. I mean, really a marketplace on the move. Maybe you want to tell me a little bit about Joom and also of course about yourself, Mike.
Well, Joom is an app marketplace, we are a mobile-first marketplace with a mission to let users purchase anything from anywhere in the world at the best price possible. Now, we’re currently operating in Europe and SIS countries, we’re not that big in the United States but you can still use our application or our website here.
And for me, well, my name is Mikhail, I used to play in a band since high school, used to be a wrestling fan, used to be a gamer but now in the last two years, I am 100% dedicated to my work here in Joom.
So, a band, I’m going to bite here, what kind of band, Mike?
Well, it was like – I cannot even tell you the style that we were playing because we were just a couple of friends who were spending great time, we were playing punk rock or maybe some metal, maybe some doom gaze or something like this, so a lot of strange music but mostly we had fun.
Well, that’s what it’s about, it’s about having fun in your job and whatever else you’re doing and you mentioned UX – UX to marketing, I’d be interested to know what kind of leap that was for you because the way I’m hearing it is, you know, UX product is so important in marketing, it’s an important skill, it’s an important talent – do you feed some of this into what you’re doing in marketing because you know sort of what makes an app special and, you know, make it sing so to speak, to continue with your band metaphor – does that work in marketing for you?
Well, it was a leap indeed, the only thing in common, I believe, both things are digital. I used to prototype and design e-learning course for nine year old children and no surprises, it’s different from design for adults. You’ve got to think different, you’ve got to add additional constraints to what you can do because children are not all that familiar with interfaces so I learn drop first any given problem from several points of view taking into consideration constraints so it helps me a lot due navigation right now and also due to my past I can see different dashboards as well.
It's interesting because that’s it, having to do UX for apps for children, yeah, for digital for children teaches you I guess a lot about onboarding, users of all ages of course because it has to be simple, it has to be frictionless.
Everyone is a child.
Everyone is a child indeed. So, we know about your background and we know a little bit now about what moves you now currently right now in November 2019 because you’ve written a great blog – I like it because I love things that are to the point and you’ve got this blog of shortcuts for strategies to simplify your app marketing efforts and improve results. Now, listener’s we’ll tell you about it later in the show where you can find this and everything else but, for the moment, let’s just understand why you chose that topic. I mean, is it based on some sort of personal experience and you said, “That’s it, these are the challenges, we need shortcuts” – what was it that moved you to write it?
Well, actually, it was mostly be chance because I got into digital advertising without any prior experience and it was just too many information to be a pro from the very beginning so I just had to simplify things to actually get things done so I couldn’t just take into consideration dozens of different metrics and so on, that’s why we chose to look for only one metric. And this, the same idea goes to everything – we were a small company, now we’re not that very big so we cannot afford to have dozens of people doing advertising, we are just six people and through them, through us, ML, machine learning developers so we got to simplify things to be able to do everything that we want.
That’s cool, I'm hearing a lot about machine learning, about marketing automation, how that makes, you know, that makes even teams with small numbers, you know, quite big in strength. Is that a challenge for you at Joom, is it a thing that size and scale and that’s a change or what is it that you see as being the biggest challenge for marketers?
The challenge changes every half a year. When I first joined Joom, the biggest challenge was indeed growth, how can we spend more still buying the right tools, how can we do the right thing with lower budget? So, how can we scale expand how we can choose the right partners, how we can fight fraud. Later, we understood that there are different problems and how can we optimise our efforts, how can we make a machine make routine decisions, how can again, how can stop fraud getting our traffic and right now I think the biggest struggle for us are targeting and brand advertising but here we still do not have working solutions.
Retargeting, I’m hearing a lot about it, dynamic ads - you’re going to tell us about that because you’re a great proponent of that but let’s talk about a couple of the shortcuts. Now, fighting fraud, we’re going to talk about that later, let’s just talk about two other shortcuts. So, if you look at what you’ve written recently and you tell me number one shortcut, other than fighting fraud, okay, what would it be?
Well, I would say the bottom line of my blog is do not do something just because everyone does so – do not pay for installs if that’s what you need. Do not pay for clicks if you do not benefit directly from clicks. Think what would be beneficial for your product, it can be even spare model but this is a decision you have to make for your own product. We tried trunk approach in 2018 when we saw everyone is into influencer marketing and spending hard which I did not, we couldn’t make it according to our KPIs and we couldn’t scale resource so we dropped it and focused on firstly that can deliver us our desired volumes, no metrics. So, the bottom line is pay for value, not for metrics.
I like that one, actually, that could be a motto on a t-shirt actually, Mike, because a lot of people did go for influencer marketing, in fairness it works in some places, it works in some verticals, I’m not going to diss influencer marketing but you learned that that wasn’t the way to do it and that wasn’t the KPI you wanted to measure. I mean, how do you measure influencer, we can talk about that in a show. So, you focused on what? Did you focus on specific events, specific transactions because of course you have to pick maybe not the North Star metric but you have to be very clear about your KPIs and they have to align with your goals, so what did you end up looking at at Joom?
It’s quite easy because we are e-commerce marketplace and we need revenue coming from our users. So, we’re focusing on revenue with our affiliates by working on a model called Revenue Share, when we pay them just some percentage of the revenue generated by their users within the first week of their lifetime in Joom. So, this means we made every corporation that we have win-win so if our partners spending hard, he bring us more volumes and that means we pay them more and more and KPI’s are matched then everyone is happy.
That makes sense so then you’re events would be really about understanding who are your valuable, you know, your partners, your affiliates so they’re bringing in business and it’s a way of measuring and incentivising them as well, I would imagine.
Yep, you’re right, still just when we finished buying users, the work is only in the very beginning because we have a lot of push notifications, email marketing, something inside our applications, some splash banners and so on just to make users buy something, be happy and to spread the word about our application.
Anything special that you’re doing to incentivise – you want people to come back as often as possible, of course, it’s a marketplace, not everyone’s a shopaholic, right, so there has to be something to keep them coming back. Is this the usual sort of showing them offers or is it just having really cool dynamic ads which say you’ve been looking at that item, keep looking at it... I’m just wondering what the different approaches, the different strategies are.
Well, we are not very much of a professional here, we couldn’t find our silver brooch here, but I can say that push notifications are working amazingly good and it was a surprise for me but for our audience, great notifications, we’ve well structured message is doing just fine.
Well, I have a message unfortunately that we have to go to a break right now, I couldn’t hope for a better segue, thank you so much, Mike, but listeners don’t go away because when we get back we’ll be talking about another shortcut, how you can combat ad fraud. Big topic, big reason to come back so don’t go away, we’ll be right back.
And we’re back to Mobile Presence. I’m your host, Peggy Anne Salz with Mobile Groove, we have Mikhail Biteryakov, he is Traffic Manager Analyst at Joom and of course, Mike, and I have to wonder if you like Mike, and I like calling you that, it makes it a little bit easier to tell you the truth, but it makes me feel that maybe, just maybe, you were either in the States and think it’s cool or you just think it’s cool in general. So, have you been to the States or elsewhere?
Well, not yet, I am mostly travelling to Europe, no further but I’ve never been to the States or to Asia, maybe it’s just a plan for 2020.
Okay, so some goals for 2020 and, you know, to recap, you are a Mobile Hero, so that is a bit of recognition in itself but it assumes that, you know, there’s a hero concept here, right? You know, which superhero do you maybe feel similar to? Is there somebody you can identify with? I’m a bit of an Avengers fan, so I’ll slip that in.
Well, it’s not about a superhero but from the university years, I associated myself with The Riddler – The Riddler is a Batman villain and he is a master of puzzles, riddles and complicated scenes. I’m not that much of a mastermind but I enjoy making and solving with things I like digging deep into the data to get back with the insider truth or some unexpected things. So, Riddler is my super villain.
That’s a most original answer I’ve heard, that is very cool, I like Riddler, I’m waiting to see The Joker, I’m based in Europe by the way so I’m going to wait a little while longer but I’m sure it’s going to be a thrill. What about speaking of thrills, challenges, something exciting – I mean, one of the top tips in your blog is how to combat ad fraud and of course there is no solution but you do focus it on some great takeaways but I’m wondering is this based on personal experience, so you have your personal battle with fraud, there you are, Riddler with the data, beating it down and combating it and then winning in the end...
Actually, you are almost right because I, at some point, I felt like the matter is very personal for me. Joom has quite a history with fraud, we changed our attitude several times. In 2017, we knew nothing about fraud, we didn’t know the word. We know that Adjust has some fraud prevention suite and we thought that it was efficient enough but in the very beginning close 2018, we suffered significant losses, paying for users that never existed.
Then, we introduced our anti-fraud KPIs and Scholar now that’s a second level of defence and for a while we got quite fine with monthly rejected from some partners but later we decided that it was not enough to feel fully protected against fraud. So, we stopped working with anybody showing any significant signs of fraud and right now, our traffic is quite trusted and transparent and I believe in 2020 there will be new kinds of fraud we just will not be able to find so it’s a really interesting field for me.
Yes, everyone says it’s a cat and mouse game, right, you beat one fraud, another one pops up, I’m counting a few. You mentioned anti-fraud prevention suite from Adjust and also working with Scholar, so they’re two of your partners – how many partners do you need, it’s like a riddle here – there we are, we’re there again – do how many partners do you need to fight fraud?
Well, this theme worked for us but you might have other needs. What I would recommend to anybody, have a second or third source of information, do not trust data if you follow it just in one place, always cross-validate everything you see. For us, a cascade of anti-fraud solutions worked amazingly well. Every install we receive, we got tested by Adjust. If it does not reject it, it got tested by Scala. If everything is still clear, it is tested manually by me to find some anomalies. It can be fraud of a different nature or maybe some long-term LTV can be poor, so even two anti-fraud defences is not enough but I believe at least two is quite sufficient.
Talking about the types of fraud, I’ve been writing about a few of them and I also wrote an article recently over at Forbes where I was talking about bot fraud because it’s going mobile, so it was online for a long time and we dealt with it and it was part of, I wouldn’t say we accepted it but it was part of what you had to deal with. And now it’s showing up in apps. So, how about bot fraud, is that something that’s high on your radar as you go into 2020, is this going to be on the list of, “I have to watch SDK spoofing and everything else, and I’m going to have to watch bot fraud”.
For us, it is not a rising threat for us, it might be baddest times are still in the past as I want to see. As for technology, bot can be used for good and bad things we saw significant traces of bot traffic in 2018 but for now our analysis on the xxx sheet when we test new partner. But here is the catch, as I told you previously, our analysis cannot be as sophisticated as the bots so they can be still in our traffic. But we pay for value and not metrics so just empty installs are not a big deal for us, and everything else, we will see a drop in retention and in some other internal metrics, so we will just switch off the fraudulent source in a while, we would still pay for it but we will not use it in future. I believe in 2020, we will encounter something completely different and fascinating.
That’s true, I would say it’s going to be always different in the future. It’s interesting that you find fraud in the traffic and you say okay, we’ll just switch off the source which is different from what some people tell me at events and conferences where I am, they just say, well, you know, 10% or whatever, 2%, they pick a number and they say, “You just have to accept it and deal with it” and here you’d rather reject it up front. What does that tell us about the approach? Does it matter, is it different for different – should it be different for different marketers or are the marketers, who say, well, just accept a certain percentage and live with it, maybe they’re not – I don’t want to say they’re not trying hard enough – but maybe it’s not the proper approach to combat ad fraud. What’s your view?
Well, first of all we maybe I misspelled it earlier but we still have 20% threshold for traffic.
Okay, so you have a threshold as well, okay.
Yes, yes, we have a threshold as well because there are always false positives, there are always some legal things why traffic can be marked as fraud and of course there are, for example, if you stop the traffic for a while, there will be fraud coming by the means of Adjust but it will be because their traditional distribution of installs will not be obeyed. So, it would be normal legitimate installs but still be counting as fraud. So, we have zero percent tolerance to, I can say, to very fraudulent forces because it’s a rare case when a person has 30 to 50, maybe 30 to 70% fraud because it’s either less than 20, either it is more than 80.
Of course, there can be a blend, there can be a mix but this, we fight with transparency. We are now not accepting traffic when we do not know their bundle IDs, do not know their exchanges or something like this. We still need something to be accountable for.
That’s great advice actually because what you’re saying is that demand transparency, marketers needs to demand that, that’s maybe part of the shortcut here, is just have very strict criteria for what you’re willing to deal with. You can still have a threshold because as you pointed out we won’t know for sure, there’s so many different factors but rather than say just have a threshold, it’s also have some key criteria to measure traffic against or partners against. What would they be in a nutshell, do you have a couple you could share? One, you said, is absolutely transparency is important, what’s another?
I can say that probably right now we are out of networks to test and every network we still might be testing does not comply with our KPIs so I can say that practically every DSP or maybe even the network that was one day now called DSP, provides us just normal level of transparency, we still see a bundle like this and still that is quite enough to at least try the source. So we’re not testing for anything more significant here.
Quick question, how many partners is too many? I’m hearing that really sort of like seven to ten is about all a team can handle, maybe less, some people are going and being more ambitious. What’s the number of ad partners for you?
Well, for the last two years, we’ve been testing more than 200 different partners.
Oh my goodness.
And so the maximum amount within one month was 76, if I’m not mistaken. But of course we’re not working with that much now, right now we’re working more I believe with maybe 20-25 different partners but the majority of them are just media buying, buying across Instagram, Twitter, Snapchat and TikTok and different other sources. So, I cannot count them as something unique. They use the same traffic sources as others but as for unique partners, I would say just as you said about 7/10.
You got me on the 76, that’s really something, I think with all that data I can see, Mike, that you chose your superhero villain wisely, definitely into the data and you’re going to tell us more about how you deal with the data and what you see on the horizon as opportunities for all mobile marketers. So, listeners, loads of reasons to come right back, so don’t go away, we will be back.
And we’re back to Mobile Presence. I’m Peggy Anne Salz with Mobile Groove, we have Mikhail Biteryakov, he is Traffic Manager Analyst at Joom, a major marketplace in Russia, talking with us today from his office in Moscow, so thank you, Mike. And it’s great to hear...
It’s my pleasure too because it’s great to hear from you, first of all we went through the shortcuts in your blog, we’ll tell everyone where they can find that and read up on that, but I’m an optimistic person, we start with a little bit of doom and gloom here, you know, the fraud threat, the different types of bots coming, let’s switch gears, let’s talk about opportunities on the horizon, you know, because marketers love to hear what’s next and what they need to do to ride the waves. So, it could be a trend or even a channel – what’s your pick?
Well, we see a great opportunity in dynamic product ads as seen on Facebook. Popular belief is Snapchats are great for retargeting but not that good for user acquisition but we see this as a delusion, that is for e-commerce. We’re also pushing our partners hard to beta-test, alpha test such functionality and already seen some success with some DSP called Liftoff, surely and dynamic ads, so my answer is dynamic ads. There is a future.
I know what dynamic ads but let’s explain the value here because dynamic ads are going to, as it says, they’re dynamic, they change, they will match your target segment, it’s that typical commerce example, I’ve been browsing, looking at shoes and I’m going to see those shoes in a lot of different places until I ultimately break down and do buy them, usually, so it does work from a user perspective. But is it more than that, is it more sophisticated now? How are you using it?
Yes, we’re using it the same way, we are not gaming company so we cannot just think about new ways to advertise our games, to think about in a creative sense, more new creatives than your videos and your html5 creatives and so on. But, we have the major asset, we have a lot of goods, we have millions of goods and we are just pushing, choosing the best out of our millions and pushing each into the network, whether it is Facebook, whether it is Liftoff, whether it is Google or anything else.
And we see the performance of every single ad and to decide which one to use, which one to test, so basically the idea is simple and that’s a question for me why no-one is accounting for it for user acquisition.
That’s interesting. So, let’s say someone’s listening in and says, “You know what, you’re right – dynamic ads, I can acquire users with that”, it’s not just driving conversion, it’s actually top of the funnel. What would they need to do to get started?
I would say they need a good research team because it’s not that easy that you have 10,000 goods, you just upload them all – you should pick the best ones, the best click through rating, buy through rating or maybe some other things. We spend more than a year just to find the best set-up for our catalogues and it was full time job for our machine learning developer who is great in his job. So, it’s not that easy but I believe that every company that have some feed, they have invest hard into it.
And this would be like the whole idea would be you don’t know a lot about me when you want to acquire me, so you need to know about my behaviour, my channels, some other things and using this with machine learning, you’re going to say, “Okay, Peggy is coming in via here, let’s show her this particular creative because she’s this type of segment” and that’s sort of like make certain that we don’t start off on ground, you know, the first step you have me pictured for a type of segment, so we’re already engaging from the get go – is that what’s going on here?
Yes, segmentation is another great part of what we’re doing because if we are talking about the Facebook, Facebook is doing all the hard work for us because Facebook knows a lot of everyone. But, if I talked about other sources when we still use other feeds, that is a job that we need to do and we just actually started this year to run this so we are in the middle of our way here.
I might have you back to tell us a little bit more about that and how you did that. In the meantime, on a personal level, you have a choice, I like to end on a personal point – either a mantra for all of your friends, buddies, peers, listening in – what do you do to rock mobile marketing, app marketing in 2020? Or, tell me what you’re going to do yourself to make certain that you stay on top of the data and keep building your talents and have fun in your job?
Well, I would love to quote actually my favourite book, although I believe the phrase is not original, and the quote is, “Think of what do you think you know and how do you think you know it?” If I’m not mistaken, it is called the Fundamental Question of Rationality so I think everyone, not even in marketing, should ask themselves this question at least once a day. That was my mantra for the last four years and it will be for the next decade.
I like that a lot, Mike, because it makes a lot of sense because you have to think, okay, what do I think the data is telling me, how confident can I be in that. Absolutely, a great mantra for 2020 and beyond. And I know that our listeners will want to stay in touch with you, maybe hear about what you’re reading, what you’re doing, what you’re thinking – how would they do it? Is it LinkedIn, Twitter, what works best for you?
Well, LinkedIn will be the best way to initial contact with me, I still cannot use Facebook very often, I just cannot stand user experience here, so I try to not touch it very often. But in LinkedIn, it’s easy to find me and easy to connect so welcome.
You’re on, and we’ll add that in the show notes and of course if you want to read up on his amazing blog of the four shortcuts you need to follow, or any of the other blogs from any of the other Mobile Heroes in the series, you can check out their dedicated page over at heroes.liftoff.io.
And if you want to keep up with me throughout the week or find out more about how you can be a guest or sponsor on Mobile Presence, then you can email me, email@example.com, Mobile Groove is where you can find my portfolio of content marketing and app marketing services.
So, my friends, that’s a wrap of yet another episode of Mobile Presence. Of course, check them all out over at webmasterradio.fm or search for them on iTunes, Stitcher, Spreaker, Spotify and iheartRadio, all of the above, just search Mobile Presence. So until next time – remember - every minute is mobile, so make every minute count. We’ll see you soon.