When you Said 'Attribution', did you Mean 'Contribution'? - Maor Sadra & Hadar Telem, INCRMNTAL.

[Music] [Music] [Music] the next session that's coming up I'd like to um give a quick introduction to Mar and Hadar from incremental how do you pronounce that incremental incremental fantastic um well hi everyone thank you David by the way was pretty awesome and uh you did an amazing

introduction also almost for our presentation uh so H yeah hi everyone and thank you for being here uh with us how does it does it work does it work it doesn't really bring it to you or try it again no mhm so um just going to introduce myself in the meanwhile so I'm maor sadra I'm CEO and co-founder of incremental uh I've been in this industry for 20-some years so actually even longer than David I spent 14 years in the mobile space you can say and about a decade on the web days uh for some of you like when I started my career I sold pop unders on Kaza and icq some of you have no idea what I'm talking about and and yeah previously I was also CEO for a company called applift applift was a pretty big CPI Network out of Germany and about 3 and a half years ago I started

incremental incremental is a measurement solution that measures incrementality it doesn't rely on user level data sdks or anything has to do with that uh we've been pretty lucky and I would say similar to David also I foresaw kind of like what's happening was literally written on the wall and yeah started incremental just before Apple announce SDK we also celebrated with David around the same day of an announcement yeah and I'm Hadar TM uh I joined incremental about nine months ago after being a customer of incremental I'm in the industry for over 10 years in mainly analytical roles I was I started in a marketing agency then moved to pla as a business analyst then was kind of became the director of marketing analytics at huge games um and I kind of faced all the privacy and measurement changes from The Advertiser side I managed a department of marketing analytics and we tried to give the best measurement Solutions results and

insights to an entire department who was kind of losing data day over day and kind of with at getting in so I actually was looking for kind of uh solutions that would help me see the bigger picture when measurement is kind of losing it um and that's how I got to incremental where I manage kind of the all the analytical and customer success uh teams now what we're going to talk about today is this is the topic and when you say attribution do you actually mean contribution we're talking about how to do budget allocation in a world where you don't really have user level data anymore to make your life quite easier so a few years ago being a marketer used to be simpler it used to be easier uh we had all the data ready we had everything trackable um we had attributed data that was kind of real time and attribution actually leveled up the the like the playing field so thank you David for also contributing to that

space at that time um also making a decision on your measurement and your kind of your RW your uh metrics was kind of simple you had your target more or less you knew what is your performance and you knew where to scale up where to scale down and what is going on yeah and even though like you as you mentioned David Facebook Google Twitter and so on wor self attributing Network the mmp's role was really important because they D duplicated the data and leveled the playing field okay even though Facebook might have claimed hey I've generated a million installs while you're MMP told you 40,000 you had again the MMP to kind of be able to help you rely and make decisions yes and as a media buyer you could also know exactly what's going on you didn't care who is srn and who isn't now as as you've seen in the previous uh deck Apple did this okay dum dum dum they announced this around 2020 around the same time us lucky people started

our companies uh basically saying hey dude would you like to be tracked how many of you actually click allow okay not a majority and that's the point and basically what what it meant is that it just meant that your data your reporting data your attribution data went from this super nice okay um OCD certified to this okay it went totally off also your organic data I'm sure you felt it completely changed it used to be kind of a small portion of your P of your overall paid Plus organic traffic and it shifted everything goes to organic now what really happened also your performance kind of started uh shaking some weird sources suddenly getting over attributed or suspiciously over attributed can say for

sure or you can um also some sources started deteriorating in performance kind of losing it what happened is Facebook not good anymore well who knows yeah now why is this happening so yes as as you've heard some ad platforms or some mmps continue doing probabilistic attribution okay even though it's not allowed and we're not going to call it fingerprinting okay we're just going to call it probabilistic now some ad platforms of course did the migration okay we support ska Network 2.0 some 3.0 some 4.0 and then back to 3.0 and then back to 4.0 so it's very confusing it's very complicated and of course by the way there's one player that says I get to play by my rules okay I have first party data so I get to use ad Services API I get to say these are all the installs I generated by the way my definition for an install is not every body else's

definition of an install okay if a user clicks install I get to say install while everybody else only if a user opens the app big difference by the way now basically what it meant is something like this the outcome often makes zero sense you can't really make a match you can't D duplicate and again data stops making any sense and also when the data stops making sense basically decision making became really off like what happens you reduce your span what happens to your organic they go up they go down they stay the same what happens how do you make decisions when attribution is kind of losing it and David covered it pretty well but um it's not going to be easier we do see the bigger picture we do understand that over the years user IDs at their many levels have become less and less usable um some of Reinventing

themselves but then doing deprecation again Facebook has really supported those changes over the time and they also added um Advanced Mobile measurement that they reduced they removed view through retribution a lot of those changes happened during the those years and basically now with privacy manifest getting into place this year and Google sendbox it will only take a matter of time until they also revoke user level data and the future is going there so we need to get used to not seeing user level data and learning to really work well without it so do we accept that we get this new lost data as a new budget allocation standard I don't think so yeah hope is not a strategy and basically budget allocation when we talk about budget allocation cons should consider contribution not attribution okay in today's world and measuring contribution is actually fairly simple because it's it's not about users okay it's about

value okay how do we determine the value okay when we for example launch a new channel okay first make an assumption like I assume that if I'm going to be spending money with a new channel any conversions I get will be more than what I previously had kind of makes sense no so secondly test your hypothesis I assume that again if I uh start a new channel and I can compare my attribution to my contribution I will actually understand what is the value I'm getting for example if I started Channel a attribution told me hey I counted 10,000 conversions it could look like this channel B can also tell you you got 10,000 conversion but it may look a bit different now the challenge often with attribution data is it only only only counts in positive okay it doesn't matter if there is value or not it only counts in positive and positive a contribution can basically either report a positive value

or no value okay even though in both cases attribution counted 10,000 conversions so contribution is kind of measuring the value of each Channel over your entire market so your your specific OS and Country whatever you're looking at a market level this is where you should see impact so if we take this example further on these are my paid traffic on their own for one of the channels I tested yeah that's the 10,000 conversions we mentioned earlier and this is my entire market for example iOS US My overall conversions so do I really see any incremental value when I combine the twos well no and that's an example for cannibalization you spend more money but you didn't get a lift to your overall conversion as a result this is no

incremental value now every change a marketer makes gives you an opportunity to do kind of an experiment and the way to do it is fairly simple just have an hypothesis before making a change for example if I increase spend I should get more installs if I decrease spend I should get great um now basically daily optimizations that you make and luckily a lot of you make a lot of changes all the time all the changes you make every single day gives you a very very easy opportunity to measure simply by having H hypothesis before you make a change as simple as that making a change tracking it having an hypothesis comparing before versus after now another important concept for Budget allocation is to understand marginal costs so if you have your data containing attribution data so if you have for example a specific Channel you can say for each spend level you start

at 1K you increase to 3K to 6K and so on you get attributed revenues you get attributed R Us let's say this is your lifetime value row us for each spend level it's not enough to look just at the attribution R it's also important to look at your marginal Ros meaning how much would it cost you to bring the next user or the next revenue on your current spend level and this is something that we want you to take into consideration because when you look at this table what's important to realize that assuming that 100% Ross is what you need anything above 9k you're losing money you're basically spending a dollar to get less than a dollar that's not great business so marginal CPI AOS for us in terms of lingo is the same as incremental CPI Ros is the same as icpi or iros we need to talk faster um for

each span this is kind of a visualization of that so for each spend level you see that even though your R is above 100% you get the full payback on your marginal Ros you're actually kind of losing it at some point or hitting a point of diminishing returns diminishing what diminishing returns so it's a proportionally smaller profit or benefits derived from something as more money or energy is invested in it fully clear yeah no I'm going to use an analogy so you want to you want to have flowers you plant a seed you water it you put 10K you get a sprout you keep spending you put more water you keep more you put even more water and suddenly you have flour now Logic for a lot of you apparently says if I'm going to drown it with water I'm going to get way more flowers but no flowers dead that's basically the point of diminishing returns negative returns

okay understanding marginal cost basically just allows you to understand where or should I put my next advertising dollar that's the simple realization behind that concept so if you have the combination of both Ros and your marginal Ros it is making your decision making much simpler and more coherent in today's world post privacy because if you understand both what basically your total um attribution data tells you as well as how much it would cost you to bring the next dollar and how much you're really contributing to the market at your current spend level will give you a much better answer as to what you should do right now scale up scale down and how should you treat your uh Target versus the actuals when you also have marginal uh kpis now are we saying that attribution mmps are useless in a privacy Centric world no no we're not okay we always think that attribution is really good because it's kind of plus minus real time okay yes

ska network is not really real time but it's close enough okay incrementality on the other hand only works in hindsight unfortunately speaking so when you make for example decisions like which of my creatives works better you don't really need anything beyond attribution when you try to figure out which device works better Samsung Galaxy S I don't know 25 versus 27 I have no idea where they are these days attribution is fine is good enough okay in summary basically it's about contribution and not just attribution you need to understand the full value of your marketing spend um according to the market performance and not just attribution data attribution is pretty awesome as a proxy okay it's real time and really it helps you understand again are you you know are you directionally going in the right direction which Remco by the way likes using Joe for later oh also you need to figure out the point of diminishing

returns and understand that if you overwater your plants they will eventually drown unfortunately now we are out of time um yeah sorry yeah so being a marketer can be simpler again if you just work with the right methods and also the right mentality if you don't have like a sophisticated marketing tool you need to get the mentality of not just looking at attribution but also at contribution to the market and changing your bottom line we have launched white paper a couple of weeks ago alongside with some of the mmps you know just to show that we're not anti mmps the talks about again how marketers should orchestrate between attribution incrementality and also this other thing that's called Mmm marketing mix modeling media mix modeling you can feel free to scan the QR code or just go to our website find it download it thank you very much for your time yeah it's been a pleasure