Ranking our 5 Biggest Pain Points in UA & how we Overcame Them (With use Cases!) - BoomBit.

[Music] [Music] [Music] up next I would like to invite uh Camille and Galo from boombit uh who will be talking about uh ranking our five biggest pain points in UA and how we overcame them uh hi welcome um small uh disclaimer it seems like our phones are not working so some of the things

might be misaligned uh but welcome I'm Camille head of business intelligence in bomit and with me uh there is Gonzalo the uh VP of marketing and today we uh have a talk for you about five uh pain points uh we think are the most important uh and I would like to show you how we overcame them uh um 30 seconds about boombit uh we are mostly developer we started as a developer but uh recently we also started offering other services we are uh a publisher we are offering uh services like the dashboard I'm personally proud of the dashboard I created and I will show you show you today uh we also selling the quality assurance so testing and uh making playables and about num

uh 29 million us spent last year 54 million revenue and 1.5 uh billion installs in total uh and we also investor investor yes so yep maybe we just want to share you three PR moments of the past couple of years uh one was actually just a couple of weeks ago we were able to put darts club that is one of our uh simulator casual games in the number one uh top charts in the UK uh it's a casual game then we can show you another example of a success case in a midcore game that we have in our portfolio so it's unal and basically you reach 165k plus uh I Revenue in one day which mean that and this is net so it means that gross we we surpass at 200k and on hyper casuals we have as well a lot of EXT extensive extensive portfolio with Hyper casuals and we're able to put sh rays in in the top charts of the US iOS and GP so as you can see we have a lot of experience across different types

of genres different types of uh games uh and that's why we are here today we want to share with you our five biggest pain points over the past couple of years and how we solve them especially camel of course uh so the biggest pain the first pain point it's pretty simple I think you all will be able to relate it's the fact that we have you know if you start counting too many apps uh with GEOS times operative systems times networks can very uh quickly ex scale so just looking at a very hypothetical scenario that is actually you know very uh minimal uh if we assume 20 apps across two platforms across 50 countries and let's be honest some of us have more than 20 apps some of us are more then two platforms and some of us are definitely more uh than uh more present than in 50 countries and if we just take into a consideration five networks and you know at some point some Publishers are even using 10 plus Networks

it can really quickly go to up to 10,000 combinations and these combinations mean campaigns right and at some point even in one single Network you can be running different types of campaigns so pretty quickly this number can escalate so that's why we in user acquisition uh went to camil and to bi and we ask for for support here because it's impossible for manually to to overcome and to work on all this 10,000 combination Plus without uh support from bi so uh about the solution we wanted something that will be very user friendly we wanted to really improve the quality of life for our UA managers so they can spend time on actually optimizing campaigns instead of iterating over every dashboards and spending like hours on clicking uh so we created boombit you can see the uh coincidence with the name of our company uh this is how it looks this is actual data actual uh G from from the dashboard what you can see here uh is the um list

of campaigns or list of adess list of countries list of sites uh along with the performance uh from many sources right uh you have clear indicator of the uh of the quality of cohorts so green red uh yellow uh scale uh you can easily change status name beat and budget in the network directly from this dashboard um you can obviously see data from MMP from scats uh some organic estimations uh the dashboard is really suited for us for our needs uh and basically was created in in in three months an initial version is based on react uh it was created on SES uh it's really easy to to maintain uh as some of the features uh lightning speed of course we customize uh we can customize targets per up uh we are qu in this is important every

Network API every hour so this is something that I don't find possible with any tool of the market usually they limit to like four queries per day or or six sometimes and even if they are doing that they don't query everything so uh with our own implementation we can query every Dimension uh an API uh can offer uh and provide features like one time one click bat budget status changes and it was like game changer for us it's really like the the the Centerstone uh of our user acquisition team so going very rapidly to to the second paino is of course iOS issues right everything that came after ATT and how we are um you know tackling this so going really quickly because we only have 14 minutes left uh now with iOS we know how the flow goes right on the day Zero we start a campaign already with a lot of limitations because some networks they don't don't even uh share breakdown per

Geo so you have to limit the way you build your campaigns uh then you have to wait from 1 to three three plus days um to to receive the data coming from from IOS then on day four let's assume you already have around 80% of your data then you can only start taking real assumptions uh on the data that is tracked uh so it's another like 30% 40% and then only on day four onwards you can start somehow predicting your LTV uh this all of this of course if you don't have a solution so we once again went to camil and we said we need your support here we really need to come up with a way to to calculate all of this uh so how we approach this case uh we wanted to provide all data you can trust in one place and this is how it looks um data from UA Network as I told you pulled every hour uh data reported by MMP the purple button this is

something we we get from uh from MP UPS in our case uh then we have data from Scot network data is of course uh delayed by up to few days with scat 4 can be delayed even more uh and what we all also adding is the data estimated on calculated on organic sorry for that mment um so this is the data uh calculated this way we just exclude all the data attributed to the other networks in R data uh and we make the weight at average of all the kpis um according to the countries according to the distribution of clicks or impression that particular adset or campaign have and this is something that we can access even uh of limited word of scan and then we provide the final estimation and what we actually provide as final estimation depends on what is actually selected if it's iOS and self attributing Network we might use scan RP uh and we might calculate uh CPA based

on the um clicks and conversions because we want the instals um also calculate real time not we don't want to wait for scat uh but if it's like classical SDK networks like a plin integral uh we can also use the data from MMP uh and if number of track installs we call it the install diff um Pro truck is high enough we can trust data from MP so uh actually the the the last few uh column uh calculation changes depends on what is selected yeah so the third pain point is of course sub Source Bing because if you remember this calculation in the beginning I mean I think we can assume what happens if you then multiply by 100 and sometimes even a thousand uh sub sources right so this this number completely goes insane um and which means that at the end of the day if you're going manually you have something

like this um we were using this three years ago not anymore also now Ras campaigns are more and more present in our lives but what happens if you don't have a solution right either you end up wasting a lot of time because your UA manages have to go bid by bid subsource by subsource and basically it's almost humanly impossible or you end up having very inefficient campaigns because your UA managers will be focusing on the top 50% 60% of subsources ignore the top 40 uh and it's it's pretty fair so I'm just I want to ask you what do you think out of these two scenarios happen the most well pretty simple both at the same time because it's impossible for some for a UA manager to go deep uh to even if he spends all this time just checking the subsources is always going to end up uh not focusing on all of them it's basically impossible so that's why once again we ask camel to help us out uh so this is feature I really proud of uh so uh as Gonzalez showed there are

this huge CSV files uh I think the one we pull from uplo takes like 30 megabytes or something so even Excel sometimes have issues with opening that and uh if you ask a manager to iterate over a few thousand few tens of thousand rows it's basically impossible uh so we create a solution to uh unify all these um C files and we call that bit files so uh we implemented the API of all the networks supporting uh subsource bidding uh and every day we run a calculation with an internal algorith called algobit we generate one bit file for every uh marketing manager we have in a company so names are changed but this is how it works on our case of course another component can work differently you might have one bit file or or even one per Network one per campaign we just have

one per one acquisition manager it's because it's easy they just go to uh their file they review it uh and uh make changes if necessary but most of the time they don't have to make changes we uh try to make this algorithm pretty stable so they just review what is changing they can uh check the history of the bits uh change it if necessary um and click the button approve and deploy and then those 10 of thousands of bits are uh deployed so no need for manual Work and No Work no need for playing with Excel uh anymore and you have access to all the the same information as in the main dashboard when reviewing the bit file which is what uh we are calling exactly and you know just just going before like just focusing here what I was saying is that then here I think your UA man UA managers can really focus on the top 50 sub sources why because at least you know that the other

top 50 they were they are being changed anyway like the algorithm is acting on them and then the UA managers can just go over the most spending subsources maybe adjust a little bit like they they they will assume that the new bid given by the algo bid is not perfect so I want to increase it to decreases a little bit and that's perfect because you know that other subsources are being uh tackled on and still your UA manager can have an opinion right and can change the bid really quickly as you can see in this example so the fourth pain point of course is arpu and AI estimation um basically you know if you go away from automation go away from boombit uh you just have a pretty clear uh Vision on your campaigns right to start a campaign you understand how your users work you understand how the the evolution from your RPO your LTV evolves and then you start kind of applying that methodology and that Evolution to more or less all your campaigns until you have a big big amount of data well in reality that's not correct right because there's so

many other uh metrics so many other dynamics that are not being accounted on on predicting your rpu predicting your LTV and that's where once again Cam and the the the bi team uh developed a way to calculate and estimate RP day 30 RP a 70 RP day 7 um as precisely as possible yes so there are many approaches to calculate arpo if you dealing with so many combinations and the um usually you go with a solution that precalculate everything on the back end and then allows you to just play with the data that's already calculated in the dashboard but that didn't really work for us because uh our amazing user acquisition team is always asking questions okay but can we also filter by that like filter by at status or filter by uh site filter by device so we needed to make this estimations

calculated uh in real time when when the dashboard is actually opened uh so this is how it works uh whenever user selects all the filters Dimensions groupings uh we just apply the uh calculation in the real time so the results of uh the calculation on very low level very granular level if you sum it might be completely different than when aggregating the same on the high level but it's intended because uh we want both high level and low level of course if you go uh low level um the data will be less accurate but this way we can always uh check uh check the data uh for any any combination so it works pretty pretty quick uh we actually do that with the help of SQL in big query so um that is connected to our dashboards also to looker and every time user uh open the

dashboard um the query to calculate estimations method selections are made yeah uh this is actually the screens I made yesterday or two days ago so obviously the the triangle Market here is the data that is estimated um but we are not showing that because we are pretty confident about this estimation uh one last thing um about calculating the uh true net through Revenue through Pro roias U sorry for Mis alment but usually what Publishers advertisers do they just divide exclude the 30% store cut and they treat it as their revenue but in all the calculations we also exclude the country tax uh so uh value added tax in GB is 20 in Poland it's 23 uh so if you're not doing that in your calculation you basically over estimating your arpu or

Roi um by up to 20 25% uh but also this is very important because what you are seeing MMP and what you are seeing the networks is still 100 funds not 56 so if you don't have your own dashboard and just look at data in network um it's not very uh reable unfortunately networks are not doing anything to uh provide us true net revenue in the in the dashboards but uh I think we have to adjust and the last point the last Point really quickly because we only have three minutes is the fact of missed opportunities so for me and for my team the the gold the golden rule is to act and not react and this is where I think alerting and all the alerting that we have is really playing a huge huge role for us I'm not going to go over this because we only have two minutes I think it's better for for us to go straight to to the alerts and what we are doing okay so you want to detect any mistakes and a problem and opportunities as soon as possible so

everyone loves slack I guess most of you have some slack instance in the company so we obviously sent a lot of alerts on slack and we developed a solution that allows to create us anql query and customize an alert uh sent on slack and this is how it looks you can see many types of alerts here the first one is hey check the Ari it's Dr 10% then another uh creative uh is saying has to check out Point Roy because both of them are dropped on the right you have the results of the uh CPI tests so the creatives and some results are sent to slack as well uh and on the left on the bottom uh there is some profit reports so this is something that our CEO especially love uh every hour he received the estimation on profit if something is going in the wrong direction um he will notify us for sure um this is how it looks the customizing the reports uh um in the in the yl uh so as you can see there are

some rules that can even customize the hers in the alert and obviously they can be uh created for UA manager for the game developer for Creative team we have it pretty customized uh and there are some additional rules that we use for hyper casuals um on the daily basis um and I think that would be all yeah I think just before we end if there is anything that you saw in this presentation that you find interesting and that you liked there is a big chance as well we are uh you know selling them as a service we have a booth right at the entrance so you can meet us there you can you know talk talk with us about anything that You' seen here and we there is a big big chance that we are as well allowing you guys to to try it out and to and to use it so that's it thank you thank you thank you so much guys uh we have time for one quick question if anyone has a quick question over

please it's going changeing color yeah hi thank you for the presentation so I mean creatives are a very very important part of UA but I didn't really see anything about creative optimization and if you can see different networks and see your best creatives and compare how different creatives are working on different networks is this also part of the tool or you don't have creative uh performance in there you go H creatives right yeah so about creatives you currently have a sophisticated creative reporting in lucer so the markets Market to we are working on putting this in boom bit but it will take additional month but yeah yeah creatives are crazy important and we have we have a lot of the alerting like the the creatives are more focused on the alerting side on slack as you saw there was like a few already there and many many more because as soon as one creative loses value we are alerted which is I think something amazing so the the you manager knows that one creative is not being efficient

just needs to go there remove that creative add a new one and yeah it's so lot VI slack thank you thank you a lot thank you thank you guys