Mastering Multi-touch Attribution: Unlock E-commerce Growth with First-party Data

[00:02] elar thank you f and if you are not familiar with elvar basically our job is to manage all of the tagging and tracking for your shopy store and all of the essentially online consumer behavior that happens from that ad click or that affiliate link um all the way to their final purchase conversion and we certainly like to feel

[00:30] that we do this better than anyone in the industry um we do this primarily through very um very comprehensive server side tracking and we also have client side tracking for certain areas that do not support that are supported by server side um and if you want to learn anything more about elvar we’ll give you a way to do that in the chat I am joined

[01:01] today by John Cairo who is our lead Solutions engineer and he works very closely with our partner ecosystem and one of our key Partners is Source medium and really the guy that’s going to give us the bulk of the information today is Fe Wang and he is the founder and CEO of source medium I’m going to let them both sort of you know

[01:28] do a hello and introduce themselves the way that we will run this is um it’s going to be very conversational with John maybe a little of myself and Fay and um sharing basically how you can use multi-touch attribution for the success of your online store or if you’re um an agency uh for the success of your client

[01:54] um we want you to also participate but what we’re going to do to keep things flowing is have you put your questions in the chat so if you don’t want to forget your question you can put it in the chat at any time and what we’ll do is at the end we’re just going to open it up to an

[02:11] open discussion Q&A um so please hang on and then we’ve got some special offers at the very end for everyone as well so uh without further Ado I am going to um ask you Fay and John to just unmute and go ahead and introduce yourselves go ahead John you wan to okay um thank you so uh thank you uh the the partners

[02:38] at elar for giving us this opportunity to share a little bit of what we know and also you know the learnings that we’ve uh accumulated over time working very deeply with our customers and scaling with them and I think one of the most fun things so far ever since we developed the L ofr integration was to continuously explore the opportunity

[03:03] that is actually within the data set that alavar produces so I’m very excited to actually get into that a little bit just a oneliner on Source medium and then I won’t say anything more you know we help ambitious Omni Channel brands with digital only with Advanced Data infrastructure so they can have high quality metric system reporting but also

[03:28] some of the more advanced use cases like attribution so um you know everything that’s here today have been you know experimented and co-developed and co-designed with some of our mutual customers um so very excited to be able to to uh you know share my knowledge and hopefully uh folks can learn something new awesome f um I’m John KY I’m the

[03:55] solutions engineer at elar and this is really exciting for us because a lot of the data we produce never really gets processed and a lot of times our clients and partners ask us what’s next after we provide them with data and that is exactly where a company like Source medium comes into the mix so this is exciting

[04:17] because they take our data and make it a lot more valuable it’s kind of like secondary processing so we’re going to talk about all that and maybe some misconceptions about lar and where a company like Source medium comes in so super excited to add some clarity and information for you guys thank you John um shall we get

[04:39] started all right um so a quick agenda you know we’re GNA just start with what is attribution period you know um you know the goal really is to help everybody understand for Technical and non-technical users uh you know so hopefully um if you don’t have any prior knowledge about it you just hear about this a lot you know in everyday

[05:03] conversations this will help you to mystify that a little bit and then we’re going to dive into two of the most popular methodologies multi-touch attribution or MTA and mediax modeling or mmm and most of what we’re going to unpack is going to be around MTA U but if there’s interest to to to dig into other methodologies uh you know perhaps we can

[05:27] have future sessions around that uh and then we’re going to talk a little bit about the first party data that uh every brand already produces but may not be necessarily owning and what the actual kind of like impact of that could be in the context of attribution and hopefully then we’ll have some time to answer questions um and then of course John

[05:54] feel free Darren feel free to interrupt me anytime uh you know I’d like for this to be a conversation um so what is attribution you know it’s just a concept or a methodology or a set of methodologies that helps marketers understand the effectiveness of their marketing efforts you know and their customer Journeys and really it’s about understanding where to

[06:20] prioritize your marketing efforts whether that is about optim optimizing a specific Channel or allocating or reallocating budget it and it has to be a holistic way of looking at it so that you know the last touch channels don’t always end up with all the credit right uh because ultimately marketing is a holistic effort um overall um and with the goal of

[06:48] essentially right maximizing your return on your advertising spend overall before you move forward here Fay I want to talk a little bit about l are in the context of attribution so a lot of our clients will say do you guys do attribution and the easy answer to that is we don’t we don’t do attribution at all but we have the raw materials that

[07:13] are needed to do attribution so Darren came up with a great analogy yesterday about a sports about sports and about how a channel May broadcast a sports game which would be aint to what elvar does with what’s going on on your SES but but we don’t start commenting and saying well you know this guy Leon I’m going to use hockey because I’m Canadian

[07:36] we’re gonna we’re not going to say Leon dry Sidle pass to Conor McDavid but so he McDavid scored a lot of goals but really it was dry Sidle who made all the passes so we should really be crediting dry Sidle with a lot we won’t ever do that kind of stuff but we will broadcast the game and allow a company like Source

[07:54] medium to start making those comments and decisions but we are kind of sort of attribution agnostic but again we have the raw materials yeah yeah that’s a really great analogy uh and just for some interesting data points you know we looked at the L our data set that we have on behalf of some of our customers some of you guys are here right now um

[08:20] what we were able to see is that out of all of the purchasing events that we were able to identify over 80% of them uh and some times over 90 for some customers have uh user ID level identification that can allow us to tie back into prior activities so you know ultimately my understanding of lr’s way of doing

[08:44] identity resolution is it’s more descriptive right it’s not probabilistic and that can give you more certainty but even in that case we’re seeing a lot of touch points that leads to the eventual conversion um so there is going to be a lot of opportunities in terms of the kind of value that you can unlock uh with the data set here yeah

[09:08] exactly we hold the information together better than you can do out of the box with your traditional say Shopify integration we will collect and store information about click IDs UTM that would probably get lost as a user goes through their purchase Journey if that Journey takes any more than a couple days so I’ll stop there but there’s a

[09:29] bunch of stuff that we’re doing to make this data work really [Music] well so you know attribution technology is Advanced but also a lot of it is still the same from like decades ago right so but what’s actually kind of driving the advancement in attribution technology right it’s really this increasingly complex purchasing Journey from Mostly single touch to multi-touch

[09:55] and now multi- channel right uh but also so there is just a lot more first-party data and just real quick first-party data just means the data that you produce just by operating so that’s the data that you own right so as a brand there is more and more first party data being available the platforms are making it easier and easier to actually get the

[10:19] data out so that also then gives the brand a lot of potential and possibility in terms of you know start owning some of this attribution in house and then of course there’s this focus on privacy whether that is regulations or iOS 14 or ad blocks and last but not least is right all of the advancements and Ai and

[10:40] machine learning that has been happening that now also makes this actually easier but also it can be more customizable for the brand with not without an army of data scientists so a quick uh comparison on the methodology itself you know so the best way that I understand it and of course uh I’m not claiming to be an

[11:05] expert this is just my best effort at understanding this and also presenting this to everyone you know mo MTA is a lot more tactical it’s a lot more ground up while mmm is a lot more strategic and a lot more top down so on the MTA side right we rely on actually very granular user level event level data so so that

[11:29] necessarily also means that it’s mostly going to be data from digital channels right because that’s where you can actually kind of collect some of this data uh you know the focus can be more on the short to medium term uh and we’ll get into some of that as we kind of get into the use cases but the what’s cool

[11:51] about that is it actually allows for near realtime optimization whether that is your marketing campaign or your on-site journey and of course it does face tracking and privacy challenges so you know the goal here isn’t to have the ability to track every single user and every single touch point that they have with you but ultimately you’re still

[12:13] going to be able to capture a good percentage of those users and a good percentage of their touch points and you can actually extrapolate from there uh use cases are actually a lot more than just um allocating budget right so uh you know that obviously it does help you with the budget allocation question but it’s actually really good for things

[12:38] like conversion rate optimization personalization you know things like assigning page value uh so there’s a lot of different use cases once you have this data set on the mediax modeling front you know again it’s very top down right so it’s really good for medium to long-term strategic planning it does incorporate data points from online and offline channels right so that could be

[13:03] obviously uh in-person activation event or Billboards or TV or even celebrity activation that you can’t really track with a pixel right uh it does also not every model does it but uh some model takes into external factors into consideration you know things like uh you know the macroeconomic conditions you know and and even weather you know um and then it’s really kind of then use

[13:31] casewise uh res revolves around high level budget allocation but also scenario and Analysis and planning so you can have a good way of understanding you know where can you actually scale and where does that saturation Point reach in terms of Roi you scale a channel so fate can you just quickly go back to that so these aren’t opposing

[13:54] Technologies these are really complimentary right yeah yeah and and the way that I would think about it is you know it’s never going to hurt to have multiple attribution methodologies because there is no Silver Bullet when it comes to attribution and I think um in within those methodologies where you you’re not dealing with models so it also will not hurt to have

[14:21] multiple models right whether that is an off-the-shelf model uh which there are many in our industry or something that is internally developed or open-source models right so as an example both Facebook or meta and Google have open- Source mm models that you can actually leverage uh and deploy internally got it the one thing I wanted to mention before we move on is if

[14:49] you’re used to Universal analytics an attribution in Universal analytics you know back in the day when we had that you’d have these live screens where you could see what was happening with your adver vertising like up to the well I don’t know if it was millisecond but very recently like during the day that changed with G4 you can’t do that

[15:07] anymore there’s a 24 to 48 to sometimes 72 hour delay which means that when you’re doing something like when you’re launching a campaign and you want to know how it performs right away you actually you can’t really see that which I know many of the people we work with are super disappointed about what uh we’re talking about with MTA and using

[15:27] levar’s data and then eventually getting it to Source medium the information is live so you get that Real Time stuff back which is really important and that’s part of elars Pub sub integration which is like a raw data stream and this is what source medium consumes from us and builds on so just uh just important to note that we’re with what we’re

[15:51] talking about today we’re talking about bringing back realtime reporting to you yeah and I think one thing to re emphasiz is that that data is not sampled you know so that is all of the user streams all of the server side streams that is happening via lar directly made available to you without sampling um sampling just means you know

[16:15] GA takes a percentage of the data and extrapolates on probably what’s going on right uh and of course there is inherent risk within that um and and the other thing I wanted to add that I love about the LR data stream is that that is a data stream you actually do care about right if that data stream isn’t good then your

[16:38] Downstream destinations will have performance issues so that then gives the brands the incentive to really make sure that the implementation is and the instrumentation is perfect and then of course the downstream impact of that is the data that you get from that is also of a higher quality as a result great point so let’s start with some use cases

[17:05] right where is the ROI at right so on the NTA side again there’s a lot more than the bullets that you see here but both of them will uh both methodologies as you can see in the first bullet will help you measure the effectiveness of awareness channels and brand uh campaigns um you know in different ways but they can give you data points that

[17:31] again is going to be complimentary for you to triangulate what’s really working on the MTA side you know because the data is so granular right it’s at the user level it’s at the event level you can really do very complicated user Journey analysis right so I have a screenshot of the San ke chart below that’s actually something that we’re

[17:53] going to be launching uh relatively soon in the next uh few weeks for some of our customers which I’m very excited about uh but also you know you can now have these Custom Touch point value assignments right so if you think about the landing page optimization use case if I have two or three landing pages in in the middle of the purchasing Journey

[18:17] uh that was actually instrumental for me to make that decision right those landing pages deserve credit right and of course that scales out to all kinds of things like UTM prams and channels and uh things like that as well and ultimately that’s going to help you with conversion rate optimization because only when you understand your user Journey can you and and where what

[18:41] channels they’re coming from right can you truly understand how to optimize those landing pages which can ultimately lead to personalization efforts uh and because also we have the ad ID in a lot of these events you can also now tie that user journey to a specific creative or app which then further goes into that whole personalization effort

[19:03] overall um on the mmm side you know it’s really kind of about understanding your media mix budget decisions right are you spending too much on meta where you’re not spending enough right should you be start should you start spending other channels right you started spending and testing a specific Channel you want to have confidence around scaling that

[19:28] right so the the the example that I have below is you know essentially um the relationship between spend and projected Revenue right so a lot of these mmm providers but also open source models will give you what the likely saturation point is with a Target row ass right so then that can give you a sense for okay we can ramp this up

[19:53] another 1.5x and then the rest of that marketing budget can go to some new initiatives or other channels and then lastly you know it’s about scenario analysis and planning right projecting out different scenarios different budget mixes understanding how that may uh impact your return on ad spend and your Revenue growth overall hey ba I did have a question so

[20:18] I see that for mmm you have the geographic incre incrementality easy for me to say and um not for MTA so I guess my question is why is the geographic relevance not called out in MTA because wouldn’t that also provide me with some important information that may feed the mmm model right absolutely yeah yeah that’s a great

[20:49] question yeah so I think one thing that I I forgot to mention is that the mmm model improves as you feed it better data deeper data and Fuller data you know so um so you know obviously okay you have these touch points and their geographic location IP address and all kinds of stuff but well there has to be

[21:12] a model on top of that that helps you to understand what’s going on right so the incrementality one is uh a more advanced tactic uh that typically larger Brands employ because you know you can run awareness campaigns you can even run direct response to campaigns in a specific geographic area that exhibit a specific customer Behavior right so in

[21:39] that case the other the rest of the country in this case wouldn’t know or be exposed to those messages right so that’s where you can then start running different types of experiments provided that the geographic region exhibits similar behaviors uh and that’s going to also allow you to to have more confidence to scale out to the rest of

[22:03] the country rest of the world whatever the case may be perfect thank you yeah so let’s unpack so we’re gonna just kind of like really focus on unpacking MTA um you know because I think this is what uh I hear a lot as people kind of come in and uh and and wanting to have more advanced use cases with their data

[22:28] so so I have a very simple example here you know this is one purchase that happened over 3 days um and uh you know it started with a view item so these event names are standardized to the ga4 e-commerce event names just uh FYI so view item is the same as view PDP right this um hypothetical company sells t-shirts

[22:55] right so as you can see I’ve highlighted the the join key that we can use to resolve this entire Journey which in this case is a user ID that lvar actually assigns right so if lvar can resolve an identity across sessions it’s going to give it that same user ID so as long as you have that you can stitch

[23:20] your Journeys together and that’s why other data sources will actually be complimentary like G4 because G4 also has a user ID um but what’s interesting here is if you look at the journey right with the view item in July 1 the UTM is telling us that right meta a meta prospecting campaign drove that click right with an

[23:44] ad IDE of the you know shirt at number seven right and then they landed on white t-shirts landing page PDP and then they probably close the session from there a little bit later in the day they come back with a retargeting ad from meta right and now they’re landing on a midf funnel landing page which is something that we

[24:09] did a lot in my mattress days uh back when I used to sell mattresses uh the landing page of 10 reasons right so we have 10 reasons why our shirts the best or whatever the case may be right um and you have an ad that is driving that right 10 reasons ad whatever that may be but then not

[24:31] convinced yet but maybe a lead cap happened uh you know so you can imagine another Point here for leap um we just ran out of space uh but then right they leave again so let’s say later tonight later that night they get the clavio abandoned browse pre-purchase campaign right because of that leap event so now they add to cart but what’s interesting here

[24:58] is uh perhaps in that abandoned browse you have some product recommendations and one of them is a black T-shirt like I’m wearing right now right so they actually end up landing on the black t-shirts right uh but now this event have a revenue impact right because now we have the line item information and how much is in the

[25:19] cart but then they go to sleep and then they forgot about you Al together right so then July 3rd comes along and you abandoned cart email triggers and that brings them back to begin checkout but what’s interesting here is you know their landing page now is officially the checkout page right so that’s where you’re going to lose the value of the

[25:44] initial PDP view the 10 reasons page and the black T-shirt page but also the value that meta actually drove right of course meta will take credit because of the attribution window and things of that nature and of lr’s tracking right so but ultimately um it doesn’t really give you the type of data that you need to be

[26:06] able to say yeah 10 reasons landing page actually has a monetary value to us of X but in this case you can’t um and then lastly the purchase finally happens and as you can see we have now our um settled IDs right so settled ID just means you can now point that to actual customer and an actual order right uh

[26:32] but the user ID stay consistent as one two three throughout the whole journey and um but we can also now add additional data sources like zero party attribution from our partners at Fairing and no Commerce and maybe they actually say I actually originally heard about you from Tim Ferris okay so without that piece then you would have assumed that

[26:54] the awareness came from meta but it’s actually from an influencer I hope that’s clear is this clear for everybody okay awesome um so what did we learn from all that let’s just do a quick checkin for this particular order with the order ID of 789 right it took a total of five sessions for the conversion to happen it took three days

[27:19] right it has a revenue impact of 100 bucks and then we actually have multiple conversion channels starting with Tim Ferris on the awareness front followed by The Meta uh prospecting and retargeting ads followed by the final conversion driven by clavio your email efforts uh there was actually Four landing pages involved right the two pdps a 10 reasons page and the final

[27:45] checkout there was two ad creatives involved here right the shirt ad and the 10 reasons ad two claval flows involved and then there’s of course other inputs like referral domains promo code used etc etc um you know but we can spend all day basically digging up all the metadata there but what’s cool is that now that you’ve linked it to an actual

[28:07] customer you have all of the customer level data right so are they a first or repeat uh purchaser right if they a repeat purchaser is this their first second third fifth order what is their LTV right what are their previous order attributions you know so now you really really truly have like this holistic picture of how this customer actually

[28:33] got to know you so I’ll just do a quick crash course on what are all the different MTA models that is like popular right you might find this on a textbook right but of course with the advancement in uh machine learning you know you can now have much more nuanced uh credit assignment uh logic of course you don’t need machine earning to

[29:06] necessarily do that it can be developed internally um so starting with linear right really this is saying every single touch Point gets the same credit right so what that means is out of that 100 bucks of this Revenue impact we had five sessions so each session gets 25 bucks for 20 bucks right if we’re just kind of

[29:30] doing really really simple math here time decade is really about um having an emphasis on the more recent events right so of course last click is just like the only thing that happened right before the purchase but this is still putting more emphasis on the recency of the events and giving those events more credit because they’re closer to the

[29:55] purchase you shaped is interesting because it essentially gives the most credit to the first touch and the last touch so you know that’s giving credit to however the awareness may have happened and then also the thing that’s closest to the conversion and then evenly spread out the rest of what’s left and W shaped is essentially uh so

[30:19] the last two is a little bit more uh nuanced you know so w-shaped it’s about giving credit to the first touch and then giving credit to the lead capture moment which you know is going to be somewhere in the middle of the funnel where it could be in the beginning and then again giving the most credit towards the end of the purchase funnel

[30:40] um a full path is probably not really necessary I don’t think for most brands but this is giving credit giving the most credit to the first touch the leap and some kind of a activating or key purchasing pre purchase event like adding to cart right and then the the last thing being the uh actual purchase um you know of course sorry um I just

[31:11] want to reiterate back to what we were talking about earlier where is lovar an attribution tool do we do attribution this is the stuff that we don’t do we we just don’t make these kinds of decisions and this is the kind of stuff that a company like FaZe at source media May there’s another aspect to this though um with a channel like meta you

[31:33] probably everyone seems to be on one side or the other in terms of do they take too much credit for what they’re doing or too little credit I think from our perspective I think we’re on the opposite side of the merchants and we oftentimes think that a channel like meta or Google ads should actually receive more credit than they do and

[31:52] they should take credit for absolutely everything that they do because they’re not necessarily an natur distribution tool either they’re just trying to tell you when they were involved and you want to know that so where this is all leading is meta has a tool called Channel lift studies and those Channel lift studies allow meta to kind of get

[32:14] in the mix on attribution a little bit so you can send them information about how conversions happen let’s say you have a Google ads conversion somebody buys last click was Google ads you send them the information along with who you think made that conversion happen in this case Google ads and meta will say we were involved in a click like a

[32:37] day ago and we think we deserve some credit here and and they’ll take credit in the platform these Channel LIF studies are a fairly new feature of met I think they’ve been around for a year or so they’re relatively hard to implement lvar makes it push button so if you’re looking for a new perspective and you have a meta rep that you can

[32:57] talk to you can get that New Perspective with like no work you enable it in the app you talk to your met rep and tell them it’s on and they will provide you kind of sort of similar data to what fa’s showing you here from their perspective they’ll take credit for what they think they deserve so I just I

[33:15] think it’s the easiest way forward with this stuff is usually what’s gets done and The Meta Channel live study is if you’re interested in like dipping a toe in the pool it’s an easy way to sort of start exploring this stuff yeah thank you for that I think that’s really exciting and I think um Geo incrementality testing is uh also an

[33:40] option so uh you know we helped uh one of our customers ALS overnight uh a few months ago you know getting that data set ready um with dma level information and you know you can imagine a situation where they’re testing a new campaign type or way of doing your ads but just for Texas right so then your whole

[34:02] business doesn’t tank right if the thing doesn’t work so um so you know there’s a lot of different ways of again understanding attribution you know and uh you know is incrementality part of an mmm methodology or is it its own thing right like ultimately it’s it’s it could be all kinds of different things it’s just about first of all having all of

[34:26] your first party data ready and at a highest quality possible but then second of all drive that from the use cases which um move on from this I just have a couple of questions you might be getting to the answer but so one I just want to confirm these percentages that you have for all these different models those are industry

[34:44] standard percentages so if you choose that model that’s the way it’s going to be applied right okay and then yeah um do you see or so do you see that clients merchants typically go with one of these over any other for a particular reason um and do you have and John to answer do you have one that you think is the most

[35:11] relevant and would give them sort of the the most um realistic picture of who should get credit for what that’s a great question um I’m happy to go first uh so yeah I think most of what we see right now is uh you know folks are using more of an off-the-shelf Tool uh where it’s not really clear you know and there may be

[35:39] documentations but it’s sitting with a vendor um so so that’s I think if you’re using one of these tools it’s important to understand how they’re doing credit assignments and all kinds of stuff like that and you know because it may or may not be the right fit for you and the other thing is it’s really uh specific to the business type you know

[36:02] so one of our share customers elix Health uh elix healing.com they have a um more intensive sort of pre purchase journey of you know evaluation and personalization and things of that nature so in that case right you have to have certain things happen before you can really even buy the thing um so you you know so what are the touch

[36:31] points in that pre- purchase quiz or evaluation uh you know another one of our share customer CPAP has something similar right when it comes to uh a prescription right so without those things a conversion cannot happen so you know in that case hey maybe the full path makes a little bit more sense right or maybe the W shape makes a little bit

[36:53] more sense but if the purchasing journey is relatively straight forward um you know it’s just a great product another one of our share customers that’s here perfect gnyc right people want jeans they need uh shirts like it doesn’t take a whole lot of evaluation um if it’s a good value and a good product um so in that case I think

[37:18] a more a more simple uh model is would probably suffice I’m going to cheat here and say data driven attribut bution and that’s the only reason I cheat here is just because it’s something that you can explore if you have a G4 property set up um Google it I won’t go through all the details here but Google it if you’re

[37:38] interested it’s a little bit different than what we’re talking about here where you let an algorithm decide where the weights should be placed so something you can explore pretty easily if you’re interested and there’s a question from um a participant that I think is is fitting here so if you don’t mind I’ll just ask it here um and it’s how much customization is

[38:00] possible in terms of either crediting or penalizing one of the different sources so they have a source for example that they would like not give any credit to are you able to do that in yeah yeah uh like for example what we do is we actually just unify and get the data right up to the point where you can

[38:23] start doing the customizing um and I have a a good graphic in a couple different slides um that will kind of allow you to visualize that uh but I think the key is you know I mean that is a reason why as a brand is scaling to start owning this in house because only the brand would understand

[38:48] those nuances um that an off-the-shelf provider may not be able to provide you know and and ultimately one size fits all is is tricky when it comes to modeling just in general you know any kind of modeling so and and the the data D attribution piece with GA is also a great Point like that is an example of using machine

[39:14] learning to essentially figure out hey this is how I want to spread out the credits and it could just be any percent right I can give 1% to One Touch point and 98 to another because the machine said so right so that’s where you’re going to give up some control as a result of that but it’s important to

[39:37] experiment like what I would do really is to um you know create a reusable like data template where you can have all of these different models and compare right and using that’s where the human actually comes in to sort of evaluate the outputs relative to what they know to be true and the human ultimately have to be the one that decides on the on how

[40:02] to do the credit assignments and but it’s not when andone because your business is always evolving all of a sudden you’re in Costco all of a sudden you are you know in Canada right so all of those things is good reasons to also just be always rerunning these models and always be Rec comparing and re-evaluating what you should be making

[40:23] your decisions with so what are the building blocks of attribution right and this is generally uh I would say this is generally applicable to uh like mmm as well but some of it is more specific to MTA on the on the technical side so there is two aspects right one is organizational or cultural right so this is human

[40:55] beings have to sitting room right computers can help you with that right and really the most important thing is start with the underlying business objective of why you need to do attribution and what Northstar kpis should be the one that you’re looking at in terms of optimizing towards right because if the initiative is successful and you’re your object your objective is

[41:22] to grow Topline with a you know media efficiency ratio of three right like that should continue to be the case and the model should be giving you all the information you need to make that happen but it could be profitability right it could be other things as well uh or R purchasing right then the other thing is

[41:42] to just really map out your customer journey and to be continuously updating that as you add different sales channels right so you obviously have your website but you you might have a mobile app that you know you build build with something like tap card or other Technologies um you know or you know other sort of ways of essentially

[42:05] generating sales right uh that includes Facebook shop right and Instagram shop understanding the privacy and compliance that has to happen there right because obviously if you’re collecting pii then there is a privacy implication and how that data needs to be stored and then aligning on measurement framework what are the events you actually are tracking that is in your funnel right that’s that

[42:32] goes beyond pdpv at AAR begin check out and purchase and what can you actually do with that if you did uh uh start tracking those events and then standardizing your inside generation practices reporting templates and ultimately actually what is the plan around activation of all of that once you have the outputs what are you actually going to do with that on the

[42:55] technical side uh the most important thing really is about uh aggregating your first party data right and for some data sources you can’t go back in history uh and I think lvar is a good example of that we can start collecting the data the moment we start collecting the data and look back is difficult uh for some data sources so it’s always

[43:19] going to be worth it to start your data aggregation Journey uh as early as possible uh so you have all of the data points for when you need to make certain decisions later on uh and then of course you actually actually have to in instrument these tracking uh instrument elevar instrument ga4 instrument all of that on your website and other on your

[43:39] app right all all kinds of stuff standardizing the data format and the taxonomy which I’ll have a slide on in a little bit unifying your naming conventions like UTM promo codes landing pages and then ultimately just like the example that I showed connecting your user Journeys with the anonymous ID which was the user ID from the previous

[44:03] example with persistent IDs right persistent IDs being your order ID and customer ID that Shopify generates or any other platform and then getting to the actual modeling bit uh you know at the very end um the reason that I kind of wanted to show this slide is because I think most of us uh probably have experienced this right a customer comes

[44:26] for a prospect comes and says I need to do attribution right but it’s like okay why though right what do you actually want to do and do you kind of have some of these foundational building blocks because if you have incomplete data or uh low quality data and you model on top of that it’s going to give you bad

[44:45] outputs and you’re going to be making bad decisions which ends up being worse than actually just doing nothing at all right I just wanted to give you a heads up that we have like 10 minutes left and I know know you have some great um customer examples to show so I wanted to make sure that you’re able to share

[45:03] those yeah yes I’m actually at the end um just because of the timing constraint so uh after this slide we can move into Q&A so this is a diagram of essentially all of the um event streams that a brand likely already has access to today you know so so from the website you have your ga4 event stream you have your lar

[45:32] event stream right obviously combining the two can make it even more powerful right whatever lar may have missed ga4 may have picked up and vice versa right from your mobile app also you have ga4 but also tapar we are able to ingest their event stream as well uh but what’s also interesting here that I added more recently is physical retail right I know

[45:55] buckon has a lot of uh physical retail data there but also some of our partners like novel and Bridge right whether that’s using uh an apple wallet pass or a QR code allows for that sort of like offline Redemption offline self-identification so we can also resolve that to a customer record and then ultimately also the zero party

[46:21] piece with uh you know our partners at Fairing and no Commerce so now you have all of these different data streams that’s basically going to be fire hosing 24/7 right well you need a way to essentially centralize that so that’s where a data warehousing technology comes into play so whether you want to adopt something like bit query or data

[46:43] bricks or you know AWS or whatever snowflake right where you can use a vendor like Source medium to just like simplify that and and really kind of just get to the part where you need to just work on the modeling right um and then what’s key here really is about producing a unified event schema that unifies all of these different event

[47:07] streams into one event stream right so identifiers right your session ID user ID customer ID but also in the physical retail case what is the store ID or even what is the clerk ID that actually run up the order um hey do you have some questions to ask about that today maybe not but might you have a question about

[47:28] that a month from now maybe yes right so that’s why you want the data then the next piece being the attribution right the utm’s landing page referral domains the ad ID Etc and then standardized events right so the events should all be any PDP View events from any of these sources should just say view item right they shouldn’t have different names

[47:52] because every vendor does have a different naming convention so unifying that’s very very important and also Event Source like did that come from your mobile app from your website from physical retail the revenue impact and of course the timestamp which is going to allow you to have that um timeline view of you know how a purchase happens

[48:12] and ultimately pii right name email phone Etc and then in that data warehousing environment or in that data infrastructure environment is where you can then go ahead and do the modeling piece at the end uh so that’s kind of what I meant by like what we do is we bring the data to 90% of what you need

[48:32] to do the boring stuff essentially but that’s also very scaled and it’s very hard to do for a brand so then you can essentially focus on the part where it’s just about deciding on the model right and how you actually want to model that data that is it so would love to take some questions and also I’m going to have our

[48:58] offer here as well okay so oops Darren you’re on mute oh Darren um so um for anyone that wants to scan these QR codes for if you’re still on the line you get to take advantage of this um 50% off our expert install which Bas basic basically is the endtoend implementation of everything for elar for client and serice side tracking um

[49:34] and it’s with the purchase of our Essentials planner above and for Source medium a free data audit um so go ahead scan those QR codes we’ll just leave this up while we take your questions um here’s a question Fay what would you say is the hardest part about adapting a solution like this for a brand yeah so the hardest part is

[49:58] probably just building the the the left side of all of this right so you got to think about you know every vendor has a different way of pushing data to you uh obviously then you need to uh adopt an actual data warehouse right without accidentally spending $110,000 a month which has happened to us you know more than once and um and just like the data

[50:25] modeling of it right so that’s just tons of SQL should I be using DBT or should I be doing something else right so like that is the part that is expensive and risky to do um you know but then of course the model selection process as well you know and I think understanding which ones of these models most closely

[50:51] correlates with the the the the mechanics of your business right but that necessitates understanding the mechanics of your business right things like seasonality things like right uh a campaign blast what that actually does right or promo codes or your LTV motion right because only when that happens especially when LTV comes into the question it becomes really tricky

[51:18] because well are you a do we start assigning Credits based on LTV right uh some might say that’s a good idea but that sounds kind of hard right so that’s like where some of these things can become tricky on the on the other end of it we put that last slide up again because we so um first of all I want

[51:38] people to have the offer but also um there was a question about um Source medium pricing and I I think we’ll um what who should people contact Fai they want to get a hold of you guys yeah so you you can email me directly a f sourc medium.com that’s f as Frank EI sourc medium.com uh our pricing is pretty

[52:00] straightforward we look at your trailing 12 months uh Top Line uh you across essentially uh e-commerce channels that we support so we currently support Shopify Amazon charge B and stripe um so like everything else from there like if you add additional data sources into our hosted bit query or whatever that’s all going to be included and I’m going to ask one more question

[52:28] there’s there’s nothing else from the um floor but so when you were saying about you know you want to standardize on naming conventions and things which makes total sense right and anytime you’re building any kind of database that’s what you want to do you want to be forward thinking thinking ahead for how how might I use this how should I

[52:49] name this right and John um add in some color here because I may be confused things but I know that a lot of times Brands will have their own custom naming convention for a variety of things so how does one balance right having a bunch of custom names versus being able to apply something like multitouch attribution properly when

[53:15] you’re going to need for obvious reasons not just have those events named but probably also promotions and other things that are going to be very important in that model yeah yeah so this is like something around the Realms of like marketing Ops or ad Ops right so you know a lot of Brands come in with to into our you know Universe with naming

[53:40] conventions that was messy right where maybe the first 10 million was kind of all over the place all of that can happen but in a data warehouse environment that’s why it’s important to have access to the underlying data you can fix that but then from there uh what’s important is to have planning right uh relative to the channels and

[54:05] relative to the campaigns right unifying your UTM values unifying your promo code values making sure you know the Cs team is creating you know Z reshipment promo codes and things like that in a consistent manner so that um you can actually then understand what those values are right and then ultimately connect that to the channel right so you

[54:33] know Google CPC is probably the one that is like the most default and no one really customizes it that much but once you get into Affiliates or influencers and all kinds of stuff right if you have a good UTM uh coverage on that and the influencers are using it and they have a dedicated landing page which you know

[54:54] one of our share customer elements does you know quite a bit then you can actually maybe even if you didn’t get the I heard about you from Tim Ferris you can still get to Tim Ferris in that attribution Journey right so that’s kind of where the naming convention can really come in but it’s never too late to start unifying your naming

[55:15] conventions uh we have a UTM template for example that uh you know I’d be happy to share as well well I think that’s everything super interesting I want to thank everybody that attended um for taking time out of always busy days we appreciate it and please reach out to us with any questions um you can reach me

[55:38] at Darren d a r Ren get el.com uh John is at Jonathan gar.com and Fay is Fay at sourc medium.com thank you everybody appreciate it fa thank you guys thanks much thanks guys bye-bye all righty byebye