How To Make Marketing Attribution Work

Q1 — “Accuracy Is Table Stakes—Actionability Wins.”

The area that you work in, especially in marketing, is very close to my heart, which is really a marketing attribution. And it’s an area that’s getting a lot of attention from CMOs because a lot of money gets spent on digital today. I think the growth in digital has been fairly… I would say steep over the last decade or so. Therefore marketing attribution is an extremely important area for marketers. 

Your platform also is something that is phenomenal. 

What is your definition of successful attribution? Is it accuracy, actionability or both?

Scott Desgrosseilliers (01:21.476)

It’s definitely both because you know, you got to have the accuracy. That’s like table stakes for marketing attribution. It’d be like if you had a sports team and you didn’t have accurate stats on how the players were performing, you’d have no idea if you actually had a good team or not. And so I like to look at marketing attribution like creating a scoreboard for your marketing. And then it helps make it more accessible for people.

So you need, you do need the accuracy because why we exist as a business and why it’s a field is that it’s hard to get accurate data. It’s hard to get accurate data because people have different meanings of what a sales conversion means, even though they shouldn’t. There’s a lot of revenue that is sometimes not captured online, such as recurring subscription revenue or offline sales processes that get driven from online lead activity.

And, so just getting the accurate data in a lot of business models is challenging. Even in a pure Shopify store, it’s tricky because you need to get the first click data. People that started the customer journey is very important to getting value out of marketing attribution. And if the technology is not good enough to capture that, it’s going to mislead you and report first click data that’s really like fifth or sixth click.

And then you’re banging your head against the wall wondering why you’re not scaling your business. it’s because your data wasn’t accurate to begin with, even though the numbers look good. So that’s why accuracy, I’d say, is paramount to get started. But then once you have accurate data and you can trust it, then accuracy doesn’t matter. Because once you can trust you have accuracy, then you move on to you actually have to be able to make decisions with the data. That’s where the actionable comes in, is you have to not even just be able to make decisions, but have a consistent like weekly routine or if you’re the media buyer, daily routine so that you’re in there doing repeatable processes and aligned with a team looking at shared scoreboards of how the marketing’s doing.

ContraMinds Podcast (03:23.982)

Great, so what you’re really saying is accuracy and actionability go hand in hand, right? That’s really what you’re talking about. 

Q2 — “Pick the Right Model for the Job, Not a Holy Grail.”

So for our audience, if you can define what is marketing attribution and if the definition can be in business terms, how do you really look at the definition of marketing attribution from a marketer standpoint, it will be great. I think it will be a good place to start.

Scott Desgrosseilliers (04:19.407)

Okay.

Scott Desgrosseilliers (04:40.987)

Sure, so marketing attribution is the field of assigning credit to your marketing. So it’s creating a scoreboard to say what’s working and what’s not working. So it sounds simple, but it gets challenging right away because there are a lot of different models. Attribution model is a different way of attributing credit. So when someone says, well, I’m looking at a multi-touch model for attribution.

Initially that sounds like a great idea. I’m going to take my revenue I’m going to spread it across all my marketing touch points and then I’m going to see which ones are profitable and We certainly have the multi touch Attribution models a couple of them available because that does give you a general idea of something’s working or not however, let’s just say that your average sales prospect to customer journey is 10 touch points and let’s say a sale for you is worth $100. A multi-touch model will then take those 10 touch points, take the $100, and then divide 100 by 10, and then give credit to each touch $10 each in a balanced model, which is one that’s easy to understand. The trouble is that then is going to mute the impact because different touch points have different value. They’re not all equal.

And so you need to look at all the different ways you can keep score. You need them in the context of what you’re trying to do with your specific marketing campaigns. Rather than look at this, look for one holy grail attribution model, it’s more picking out the right model for the right job you’re trying to get done. Because when you can measure the outcomes you’re trying to get against the intention of the marketing you’re trying to do, you get a lot more value.

The example I can give is the hardest thing to do in marketing that we see, you know, I’ve looked at over 10 billion in ad spend. The toughest thing to get is top of the funnel, which is cold traffic that’s never heard of your brand, how to find the right sources of those that lead to customers in a profitable manner. And the challenge and attribution with that is people don’t click and buy right away, particularly if you’re not a household brand. If you’re a household brand, you have huge brand awareness in your niche, then they will start doing that. But how they got to that point was marketing when you weren’t a household brand. And so for that attribution of what top of the funnel started, that’s where the first click attribution model becomes very handy. And how we look at that from marketing attribution is we take your revenue and then we use it as a signal and put it all at the first click for that one model so that when you and a marketer saying, well, I sold $50,000 today, where did the first clicks come from? We take that 50,000 and assign it at the very top of the funnel, the beginning of all those different journeys. And they won’t match Facebook or Google who have gotten rid of first click and have no visibility, which is puzzling because they’re great sources for first click. It would give them more credit, but I don’t know, they don’t do it… which is great for us. 

So it’s creating a different scoreboard. It’d be like if you’re looking at a soccer match and like you win the game four to two, well, okay, who, how was your possession time for the midfielders? How many shots were missed by the strikers or were they accurate? Or did they have 50 shots to score the four goals? The goalie, did he only have three shots and give up two goals? So you could say, well, we won the game as a team… but our goalie is weak and we need to fix the goaltending situation. So it helps you strategically and tactically move forward and improve your marketing and optimize it.

Q3 — “Start from Real Sales, Tag Everything, Build the Timeline.”

ContraMinds Podcast (08:40.344)

One of the challenges in digital marketing attribution is the number of channels that are there because each of them are pretty much walled gardens, right? So therefore what happens is if I have to look at it holistically, the way I allocate budgets, I allocate it across different media. It would be Meta, it would be Google, it could be LinkedIn, it could be Shopify, it could be an e-commerce site like Amazon ads. So clearly, how do you really go about aggregating them because that’s the first challenge, right, which is really the biggest roadblock I would call for attribution is the way you orchestrate and bring the data and aggregate it together and how hard is it and what are the steps if I were a marketer, what should I take?

Scott Desgrosseilliers (09:24.901)

Yeah.

Scott Desgrosseilliers (09:41.318)

Sure. So I had that challenge before I created Wicked Reports with my very first attribution consulting client. I didn’t even realize that’s what I was doing at the time. This was a long time, 11, 12 years ago. So I decided to start from real sales. So I looked at his sales system and pulled out the sales and said, well, this is what I’m trying to get more of for him or track what happened. I’ll start from these actual sales.

Then I pulled out from the CRM all the leads that were generated. And this was a pure e-comm. But you still have people that, you know, they don’t buy right away, but they opt in for a coupon, or you’re doing abandoned cart capture techniques, or you’re running paid ads to opt in to get special deals. All those ways, you’re still actually generating leads, even if you’re not a lead gen marketer, you’re an e-comm.

So I started from there. So you got to start from the raw first party data because then you have data you trust because you can go and look at it. At least you have something you can validate because yes, the walled gardens aren’t going to give you anything. They’re going to give you ad spend and then inflated conversion counts. So starting from there, you know, you have your leads that occur and you get your sales that occur. And lots of times they show up at the same time but it doesn’t mean they bought the same day. It means that they weren’t identified by the CRM until they bought and then they got pushed into the CRM. So you start creating a user, it’s called identity graph is the term for it, but basically a timeline of what you know about the email address. And so from this, just pulling from the CRM and the cart, have sales, hopefully multiple, depending on your business model. And we’ve got when they showed up in the CRM, which may or may not be before they bought.

Then we have to start adding clicks to that timeline. And the way to do that is you’ve got to tag all the ads in a way that you can identify them, not just for what platform they came from, but you’re to want to know which campaign. And then you’re going to want to know which ad group, which is the targeting, if it’s social. Ad group is just a grouping of keywords or audiences, if it’s Google or Microsoft.

Scott Desgrosseilliers (11:58.136)

And then at the ad level, if it’s a social or and then down to the keyword level, if it’s search. So the way to tag those is using UTMs or you can create your own, generate your own ID and look that stuff up. Or you can have some other mechanism, but that’s what generally works. So that’s another challenge because people will have a lot of, a lot of ads. And so now it used to be really hard back in the day, you had to go in and manually uniquely code them all. Now there’s variables so that it’s a little bit easier. So once you’ve got that, when people click, you’ve got a string of text on the end that identifies what made them, where they came from. Let’s just use meta. So it’ll say meta, campaign one, two, three, ad set. I don’t know, male, lives in Boston like me. And then the ad is whatever hook you’re using or whatever image.

You want to name those UTMs so that they’re accessible to you. You as the marketer go, I know what that is. That’s the whole point of the strategy. Now I like to instruct people to name the campaigns what part of the funnel is. TOF for top, MOF middle, BOF bottom. So they have an idea. And it will help us know which model we should use to score it. 

Then the most challenging part, the next work comes in. You got to capture the click. You got to try to identify them, which is not, is more possible if they end up submitting first party data, meaning their email while they’re in that visit, but particularly when they’re new to a brand, they might not. And so then you got to store that click with all the info you can, IP address, device, whatever else you can get your hands on and just store it away. If you don’t figure out who it is. And then that’s where you really make your money as an attribution company is that then when they end up becoming identified in the first party data, that’s a bigger signal to me. Okay, I’ve raised my hand and said, I’m willing to get your bombarded by retargeting ads and emails, here’s my email, or ideally I’m willing to buy and take a chance on your product. Then we use our attribution time machine, we call it. We gotta go back and figure out, okay, now we know them when we didn’t know them, now it’s value, let’s go find out that whole path.

Scott Desgrosseilliers (14:23.577)

And that is like 10,000 lines of code to do that. So a lot of, a lot of just continued, just keep iterating on it and learning new ways to identify people, getting, you know, lot of challenges there. So all of those things are what it takes because otherwise if you’re just trying to stitch the clicks when people buy right away, that does tell you what closes people. And if you’re a household brand or if you have a really impulse type buy item with a small, small average cart value, then that is probably still potentially good enough. But anyone that has some consideration, you need to be able to stitch the clicks before because that’s your competitive edge if you can get it. So I hope that helped. I’ve got a little techie, but that was a trip down memory lane for me.

ContraMinds Podcast (15:12.11)

No, That’s very, important, Scott, because clearly what you’re telling me is, so if I’m a marketer, the importance of categorizing and building my data types for effective marketing attribution is the first step as much as you are thinking about your campaigns and where you spend the money. So therefore, tagging the data type by the campaign type, by the media type, whether it is top of the funnel, middle of the funnel, bottom of the funnel is as important, if not critical in your process, as much as your creative and as much as your multi-stage campaigns or multi-touch campaigns that you run. And therefore, as a marketer, giving that attention is very, important is what you are saying.

Scott Desgrosseilliers (16:07.055)

Yeah, mean, the challenge at Top of Funnel isn’t going to go away because Meta has some super Andromeda black box AI. No matter how good they think they can target, it’s still not going to make someone buy until they’re ready to buy. Because there’s already all this fantastic creative out there. So there’s still going to be a path and a journey. And so you still need to measure. And the only way to improve it, you got to measure it so that you can then iteratively improve it.

ContraMinds Podcast (16:21.944)

Yeah.

Scott Desgrosseilliers (16:34.947)

Because otherwise, now, so once you’ve done all that tagging, if you don’t have top of the funnel measurement, you’re to go into like Google Analytics 4 and see a lot more direct than meta because of the time window that Google Analytics uses and because Google Analytics 4 data is shared with all their properties, while they’re advertising, they can’t do certain types of tracking that are perfectly legal if their first party data just kept for the brand’s use because they’re using it to make money against other brands for all their bidding and all kinds of smart business reasons, but not maybe beneficial to the brand using Google Analytics 4. So that means that you need that alternative score. 

Otherwise you’re going to go in there and say, geez, or the person with the budget will say, hey, CMO, I see all this spend at 50% of the budgets at top of the funnel. I don’t see any sales in meta and I don’t see it in Google Analytics 4 and my Shopify automatic tagging isn’t showing it either. It’s showing these bottom of the funnel and these brand campaigns. So why are we spending more there? So ideally you have something that keeps score so that in the CFO and the CEO are like happy. But if not, you need to get alignment on that because I mean, if you’re not spending at top of the funnel, you’re gonna be flat. Even though your RoAS in platform looks like three to five to 10 on all your retargeting.

It’s retargeting all your existing people that might be closing off your SMS or email, and then all the different channels are taking credit at the same time. And that’s why if you try and aggregate the in-platform ROAS and conversions, there’ll be more sales than you actually have, almost always.

Q4 — “You’ll Never Get 100%—Stack Signals and Move On.”

ContraMinds Podcast (18:18.082)

Perfect. So clearly, when we started this conversation, you talked about the way you tag, you talked about how do you give credit to a campaign, right? So therefore, how do I know that I am the same customer coming across different journeys? And is that cookie based? Does your system help build these identifiers, the identity graph that you spoke about? So can you talk a little bit about that?

Scott Desgrosseilliers (18:49.819)

Yeah, so we do use first party cookies that they can load in from the brand’s domain. So it loads in as the brand because we’re only using the data collecting for the brand, using it for the brand, not sharing it, selling it, any of that stuff. So that allows you to load in under the domain. So that does help. IP address helps even if people are masking their IP because if they are opting in in that session, you can still use the fake IP.

It doesn’t help for location and demographics, but it does help for identifying where they clicked from. So, and then there’s device, there’s other device ID. There’s one or two others, but, and then there’s people that sell, they’ll grab the TV IP and use that, which is very valuable. That one doesn’t change a lot. There’s a few others, but that basically covers it. And so you’re trying to… collect them all, and then there’s certainly use cases that no one can get 100% perfect attribution. Because like if I am on my phone at a coffee shop and I’m watching a YouTube video, but then I don’t take any action, there’s no way to know, they watched that video on their phone and then here otherwise. Now Google might know it and aggregate it as a view.

But for me, I’m trying to get, that one’s really tough to get. Now with Facebook, we pull in all the view data related to this, and then we allow the customer to set how much credit we wanna give it. Because it’s virtually never gonna be 100%, but if you have a strong video campaign, you can drag a slider to give it more credit, but not the extremely optimistic credit that Meta likes to give it.

The for me, the the table stakes has always been well, I we need to be looking at this as we’re getting signals from everywhere and we’re trying to turn it into actionable data that makes people more profitable with their ad spend and stops wasting their money. And as long as we do that, if we miss a few, if there’s a few untrackable situations, as long as we’re still growing their top line revenue or making them more efficient with the return on ad spend or cutting their ad waste, then we’re doing our jobs. But there are places that you can’t track perfectly.

ContraMinds Podcast (21:16.888)

So therefore what you’re saying is once a marketer subscribes to your platform, clearly when the campaign gets set up, every creator, every, I would say identity is actually built from your system into the brand’s domain. And then every data comes in to your platform to really look at how would you allocate attribution for each of those campaigns up to the conversion. That’s really the model and the benefits of using your platform.

Scott Desgrosseilliers (21:48.868)

Yes, and so we have all these user journeys captured and then we have all these models and you can on the fly change your funnel and scoring. Now we want you to set up a few standard template. We have this like Netflix interface where you pick different view templates based on your marketing goals and then we’ve taken care of all the gory attribution settings in the background for you. But then you want to get these few standard ones that the teams get aligned on so that it’s a shared scoreboard.

And then, but then when you want the deeper dive, probably CMO or the data team can go experiment with changing all the different settings we have to change what the top of the funnel definition is, change, look back and look forward windows for actually, because we can look forward and backward lifetime forever. But some people don’t want to do that for strategic reasons, so they can change that.

So it’s highly customizable. But yeah, we have a default way. like people to look at it unless they have their own because marketers are generous. You must spend a lot of time on marketing strategy. They don’t spend a lot of time on measurement strategy. That’s where we come in. We spend all our time on measurement strategy and aligning it to your marketing strategy. So you have the right measurement without spending weeks and weeks trying to configure it.

ContraMinds Podcast (23:06.798)

I think CMOs spend a lot of time on marketing strategy, but it’s critical for CMOs to spend time on measurement strategy is a fantastic quote. Very, very important, I think which allows me to get into the metrics and the KPIs. 

Q5 — “ROAS & CAC Aren’t Enough—Track NCAC and New-Visit %.”

So if I were using your platform and if I was doing marketing attribution, what are the common KPIs and metrics like return on ad spend? So can you talk about a couple of metrics and what do those metrics mean to a marketer?

Scott Desgrosseilliers (23:40.636)

Sure, let’s talk about four actually.

Return on ad spend, it’s very clear. It’s revenue divided by spend. You gotta get to one to break even and that’s just on the marketing spend that doesn’t take product costs into account. 

So generally you wanna get probably at least two, which means you’re gonna double your money before you take product costs into account. 

CAC is another one, customer acquisition costs. That’s how much did I spend divided by how many unique customers I got. So I spent $10,000, I got 100 customers, I spend $100 to acquire a customer. That’s often a good metric to use as a goal. 

And then those are two standard ones. 

Two more advanced ones that we provide that people really enjoy and hire us for is NCAC, which is New Customer Acquisition Costs. So this will expose if your marketing is mostly just re-buying customers you already have, which in some cases is a great strategy. 

In some cases, you didn’t mean to do that and you’re actually wasting the money. And so what we do to measure that is when we like, we use Shopify, we have like 40 or 50 sales integrations, we’ll use Shopify. If you have Shopify, you have a hundred different sales in a day, 40 of them might’ve been from new customers and 60 repeat. Could be a different spread, but that’s a… let’s just go with that one. So what Wicked Reports does is we take just the 40 new customers and then go figure out where you acquired them and then look at the ad spend there and then calculate the new customer acquisition cost. So in our case where, you know, that we spent the $10,000 for the 100 customers, well, only 40 of them are new, we spent $250 for the new customers. And so then that gets very, that gets CMOs attending like, crap, that’s… I thought I had $100, I was acquiring, they go, yeah, you are acquiring customers at 100, but the new customers are 250. And then that’ll really, at first alarm people, but then it’ll be like, well, now you have the evidence, now we gotta go improve it. And then when we go look to improve it, we go and say, well, let’s look at the traffic that’s coming in. Is the traffic potentially a new customer or is it already on your customer list? And you’ll find that a lot of times people just accept the default.

Scott Desgrosseilliers (26:02.273)

settings which mean target wide and then the algorithms learn, it’s a lot easier to target these people and what do they have in common – they’ve already bought from you before. And so you’ve got to take great pains for the ad platforms to not just retarget your existing customers. 

And that’s where we come to another measure we have called new visit percentage. New visit percentage is out of the clicks. How many of them do we feel are brand new visitors?

And you don’t want that on any of your retargeting campaigns because as retargeting you want it to be zero. 

But at your top of the funnel you want it to be as close to a hundred as you can get which is impossible to do because the as powerful and as accurate as the ad platforms are they can’t always handle your exclusions perfectly. They still people slip through. But we’re trying to see how your top of the funnel bringing in new clicks which then when you retarget you’re just your new clicks.

Scott Desgrosseilliers (26:58.541)

are then you have more percentage chance that they could be new customers, which then are you showing the right products that new customers are more likely to buy, which then should lead to a lower NCAC. And then you actually have the accurate data at the top of the funnel leading to the accurate closing. And then we can look at all kinds of other things to help improve that. But that’ll generally start to get growth going once you solve that puzzle. It’s a challenging one, but.

People do solve it all the time and then it can get growth. But then you have to align as a CMO with your media buyer and your budget holder that new customer acquisition takes a lot longer than a retargeting campaign. We like to set it at 30 days if we’re blind and have no data. But then we can look how long do your customers actually take to convert and you have reports that will show. You need a whole customer cycle to evaluate that. You go and you run a top of funnel and after a week you like blow it up, it’s not performing well.

No one’s is, no one’s is. Those are the four that, know, generally those four are gonna get you a lot. There’s other advanced ones you can do, but those four are gonna get you in really healthy place.

Q6 — “Let Intent Decide the Model.”

ContraMinds Podcast (28:11.502)

So one of the questions on marketing attributions that I have is, you got the last touch attribution, you got the first touch attribution, you got the multi-touch attribution and that’s also weighted and you also have algorithmic. So therefore, can you talk a little bit about the four and which one amongst these in your opinion marketers should adopt and practice?

Scott Desgrosseilliers (28:29.594)

Yes.

Scott Desgrosseilliers (28:41.635)

Yeah, so I spent 11 years worrying about that. So I feel very qualified to answer. Only because of the time in the trenches. I didn’t figure it out at first. It took me a long time. So I created a system called the five forces because a lot of people were like, which attribution model should I use? And then what should I do with the data? And what I came upon was that…

the measurement strategy should match the marketing strategy.

But in a day-to-day practice, that means when you create a new campaign, you set an intention for that campaign. And that could be something like generate top of the funnel traffic that converts. I it always ends up being that ends up buying. But is it top of the funnel or is it new leads? Is it retargeting traffic? Is it I’m just trying to make the sale right from this ad, direct response. And based on what you’re trying to do, determines the right attribution model to use. 

We already talked about a couple of those, but let’s take an example of generate new leads. 

If my intention is I need to get new, you know, this campaign is to generate leads, it’s okay. So then from a measurement perspective, that means when what we’re measuring is not the clicks, not the cost per click, we’re measuring how many leads get generated that are brand new in your CRM. And then we hold on to that source of lead generation not just source, but campaign, targeting and ad. And then we look forever into the future and whenever there’s revenue, we give that lead gen credit. Because like a common funnel can be generate a lead and then I’m gonna do a sales call, excuse me, or I’m gonna do a webinar or all the different razzmatazz us marketers do. And so then down later in the funnel, you can say, geez, we’re making more money. Where the heck did these leads come from? Well, that’s specifically what that attribution model solves. And the why we were using it is because you had a strategy where you were trying to generate new leads to begin with. And so we match it to the strategy. And so that’s why there’s no one, everyone tries that, I got a different weighted one that’s going to work great. Anything that’s weighted almost always goes, a bunch of credit at the top, a bunch of credit at the bottom and the middles like evened out somewhat. That’s how they all basically work.

I like just changing this. You have one score when you’re trying to match your revenue from a CFO perspective, then you use a linear balance touch, which means I can’t over count whatever the sale was is the maximum credit I can spread throughout the journey. But where the light bulb for me was not trying to use the CFO’s view for the marketer.

Scott Desgrosseilliers (31:29.049)

And for the marketer, in using the revenue as the score rather than trying to always make it perfectly add up to your sales for the day, gave people greater insight and ultimately more efficient ad spend. And so that’s what guides when we put a new model in there, a new idea is, all right, how are people gonna actually use this? How are we gonna have a template based on a common outcome that a CMO wants?

I have a job to be done. I have an outcome I need. How can I click a square on our thing and we’re going to have the measurement set up that actually gives you insight and then you can do actions. And I got to get off my mathematical high horse to get there, you know? But I could be like all right all day and show like, you know, standard deviation, this and that, know, stats was my passion. And that wasn’t helping anyone. I was just showing that I was good at math.

ContraMinds Podcast (32:12.791)

Hahaha

ContraMinds Podcast (32:25.656)

Ha ha.

Scott Desgrosseilliers (32:26.863)

So it really helped to just change it to how strictly looking at jobs to be done framework.

ContraMinds Podcast (32:34.52)

Perfect. And this is a course that you offer. And if I were having a marketing or a digital marketing team, the team can take this attribution course so that they start to get familiar with and comfortable with marketing attribution, right? That’s the 5D course that you’re talking about, correct?

Scott Desgrosseilliers (32:53.411)

Yes, so it’s called the five forces and there’s three ways to access it. So you can take the course on your own at five forces.com. You can use the process inside of wicked reports by tagging intentions and then running the rest of the process or our AI is trained up on it. And that’s the only agent we use. So I feel this is why attribution exists. What we created with reports for. So you can set it up and then hit a button and then RAI every morning will generate, or on demand if you need, will just generate and run the analysis for you so that you can just start plotting the, when you see the output, you can just start taking action.

www.5forces.com (URL typing Animation) + Website screenshot

Q7 — “Think TOF–MOF–BOF: Different KPIs at Every Stage.”

ContraMinds Podcast (33:39.928)

So one question Scott I have is customer journeys are very complicated. It’s not the way Google or the meta models really talk about because it’s not as if every time I I know I’m going to buy a product, right? I might actually consider it. I might probably search for it, then probably I will buy it. So there are several stages in the journey cycle that I go through. So which means that the intention that you are talking about, and your campaign have to match with your identity graph and therefore that framework becomes very very important because every time I may not be in an intense signal to buy, I may be in an intense signal to search, I may be in an intense signal to you know compare. I may be in an intent signal to you know look at the features. So the fact that you need to look at intent signals with the purchase cycle or the phase that I’m in and therefore which campaigns are aligned to those intent signal is a framework that marketers need to be prepared for and that’s the measurement strategy that you are talking about, right?

Scott Desgrosseilliers (34:57.815)

It is, for those different cycles, we broke it down to top, middle, and bottom. 

Top for us is we don’t know who the person is. All the clicks until they appear in first-party data. And that’s customizable, but that’s our default view at the start. 

You can make it just the very first click. You can make it the first two clicks, whatever you want. People usually just take our default because it works, because you use that and it scales. 

Then the middle of the funnel is…whatever between they become identified in the CRM, however much on average you need to communicate with them, either email, SMS, or retargeting ads, until you consider it’s time to buy. 

And that time to buy bottom of the funnel generally would be a week. It can certainly be longer for a higher ticket item. But if you look at it as the bottom of the funnel’s the last week, the top of the funnel’s until I know who you are, and the middle’s all the stuff in the middle. 

And then, then you align your campaigns to that. Then we measure according to that. And then it means different KPIs per funnel step.

Q8 — “Bridge Online & Offline: Tie Every Sale to an Identifier.”

ContraMinds Podcast (36:03.95)

So many organizations really are hybrid, right? So for example, not everything is digital, right? So I have offline and I have online. If I have a hybrid attribution need, how can I go about it?

Scott Desgrosseilliers (36:22.117)

So there’s a couple ways. A couple ways that we’ve helped people is when they’re using, as an example, meta ads to book in-person office visits or demos. So in that way, how it would work is they’re running the paid ads using our lead gen tracking. They get the person to book online and pick through like a Calendly or whatever, and you book the appointment. Then their online activity may or may not continue. They may still be retargeting them, but the sale’s gonna happen in the store. And then what needs to happen, which people do it, is then the people upload the sales with the email that they booked the appointment on, and then we can stitch on an online. It doesn’t have to be email, but that’s an easy way to explain on a podcast. So then we do that, and then that’s how we stitch online and offline activity together. That’s a clean way that works every time. Example could be your booking appointments for your med spa or you’re selling something home improvement and you gotta go meet them. Anything like that it works fine. 

It gets way more challenging if you have a retail store. In that case it only works if you’re doing some online promotion that they’re gonna go remit offlineand they’re gonna have some loyalty so you’ve got the connecting email address or phone number. You need that online, offline, if you’re creating this customer journey, you need the identifier and in some cases it’s very smooth. A lot of times it gets clunky or data accuracy, my teenage daughter works at a clothing store, maybe she didn’t ask for the email address and then you lost the trail. So that’s…

where just data accurate. So then your data accuracy, where we started from, you don’t get the data accuracy, so it’s lost. There’s some cool ideas around geo-fencing people’s phones and, they were on that street, so they went to the store that day. But it’s more for Fortune 500 brands that can buy all the credit card data and match it up and geo-fence. It’s possible.

Scott Desgrosseilliers (38:39.267)

We’re not a good use case for that because our use case is only if it’s very straightforward. They’re going to book something online and then when they show up at your appointment, you’ve already got to print it out who they are with their email and then you can upload to us if they buy something eventually. That can work fantastic. We can see people with 90 day lags that are making a factor of seven to 10 X more return on ad spend than Facebook will show because Facebook won’t see any of that.

Q9 — “Offline Ads? Start from Test Drives, Not TV Ratings.”

ContraMinds Podcast (39:05.453)

Yeah. So if I’m an automotive brand where I have dealerships and I have television commercials and I have radio commercials and things like that, then how would you build attribution? Because very, very difficult because then I need a QR code, I need a promo code. You know, I get people to call me into the call center. So what are the ways in which you believe if I want to build a measurement strategy for my offline ad spends, would you be the best platform to do it or would you say you will probably work with the marketing team to build the attribution?

Scott Desgrosseilliers (39:51.12)

Yeah, we’d be a partner, but that’d be a custom build because of that. I was just on the Cox Automotive podcast actually two weeks ago talking about this. It’s not out yet, but it will be. And they were asking about that. Well, I mean, so then you got to pick your most valuable signal and work back rather than trying to nail the whole journey at once. Because there’s going to be some, radio ad is just untrackable, as frustrating as it is. Even though the only thing you could do is hold out tests where you run the ad and if you have multiple dealerships, you run the TV ad for one or the radio for one and not the other and look at the lift and everything else is kept the same is the only pure way to measure it, which would be tough to do, but it’s possible. 

My thought would be to look back from test drives, because that’s a really high signal intent and then work back from the test drive journey or when you capture them in your CRM a lot of tracking to try to get people into your CRM. So a bigger focus on first party data capture. Because if you get that, then you’ve got it. Now I believe Cox Automotive also has an identity graph product of their own because they have AutoTrader, they’ve got Kelley Blue Book, they basically have a monopoly on the data. So if you can get your hands on that, they’ve got a lot of that identity graph built and then you layer it with our models and then you’d have something. Cause you could still use top, bottom change of definition, but the graph is the toughest. The accuracy piece is the toughest. Because I I bought a truck in Halloween last year and it was Auto Trader. Auto Trader was what I prowled around on, Kelley Blue Book. And then I started, then I would submit my info to a couple places once I kind of figure out my price range and mileage and what I could get. So once I’d done the research, I submitted. I submitted my info, so that’s when they’d have it. That’d be my first choice would be get Cox Automotive’s data layered into our attribution, which we’re in talks with them, so that might happen. Right now it’d have to be a request to the person there.

ContraMinds Podcast (41:46.413)

Yeah.

Q10 — “Black-Box Models? Hard Pass. I Want Transparent Journeys.”

ContraMinds Podcast (41:55.138)

Nothing. Nothing.

Yeah. So the other thing is really there’s a lot of debate around which technique is better. Is a Markov chain better? Is a Shapley algorithm better? because each of them have different ways in which, so if I was a data scientist listening to this conversation, what would your experience and advise me on which ones to pick up and which one do you think has worked better for you.

Scott Desgrosseilliers (42:30.555)

So we don’t use either one of those because we’re dealing with, we need actionable signals, not black box. And all of our attribution is transparent. So if we say that a Google campaign had 22 sales that day, you click on them and you can see the exact order IDs and you click on the order IDs and you can see all the click timestamps. So it’s completely transparent. It’s, it’s maskable with GDPR and whatnot if need be, but otherwise it’s the brand’s data so they get to see transparently why we’re saying what we’re saying So those models come in handy is when you’ve got you know household name brands like Yeti, Coca-Cola When you’ve got brands where when you’ve got a challenge that you’re not going to be able to build a graph. So and ours is always if you’ve reached the stage where you’re beyond individual journeys, and you’re just modeling based then that’s when you’ve outgrown wicked reports. If your scale is so big that you can’t be journey based. And that’s where some of those models might more come in handy on massive data sets. So yeah, we don’t engage in those at all. So I wouldn’t be the right person to speak to them.

Q11 — “Use Last-Click When It Fits—Save Heavy Math for Real Journeys.”

ContraMinds Podcast (43:45.87)

Perfect. So, so largely you are a D to C brand, e-commerce, Shopify, straight through campaigns to conversions on CRM. That’s really where your sweet spot is. But obviously, you know, as brands, as e-commerce brands become physical commerce, neighborhood commerce, quick commerce, probably the fact that they may want to do attribution around these is becoming more and more critical. Would you agree with me on that?

Scott Desgrosseilliers (44:18.777)

Yeah, so for each one, if you can get the raw data, you can trust it. And if not, you have to start getting more black box and inferring the conversions. Like, for example, if you’re trying to infer DoorDash conversions, that’s tricky unless you can get the customer data. Because if not, you’re just going to have this sales ring that is against DoorDash that should, if it gets piped eventually into your final sale in all your final sales, then it’s just gonna be more challenging if there’s a journey. Now DoorDash is more a direct response, so you can look at just ads. Hey, click here and buy, add me to DoorDash if I’m a restaurant. So you can actually scale that type of advertising, because it’s click and buy right away. There’s no real consideration. I’m hungry, what are the options? Click, do it. So that’s where a last click model works totally fine. 

So you could…get attribution as simple as that. But it is going to become more challenging. Yeah, the more places where you’re collecting revenue and there’s an intermediary, then it gets more complicated because then the attribution company has to work out pulling that data in, potentially deduping it against something else. There’s just a fair amount of work for each new source. So it has to be big enough for someone to be willing to do it.

So if you’re just making a few hundred bucks a day on this other source and your sales are otherwise a couple hundred thousand, it’s just not worth it to even spend the time to do it. So it has to reach the right scale. I don’t know if I answered your question fully, but that’s what I have.

ContraMinds Podcast (45:55.8)

Perfect.

I think it’s a good way of looking at it because I’ve been talking to CMOs and when you look at billion dollar firms, large offline online brands, they are looking at marketing mix models. Again, they don’t have all the data and they try to actually allocate expense for each of those.

Scott Desgrosseilliers (46:15.995)

Mm-hmm.

ContraMinds Podcast (46:26.368)

spends that they have. So clearly, you know, I run promotions in Texas or maybe in Mumbai in India, but I may run a different promotion in Chennai or in say California, which means that different promotions, different geographies, different spends, different segments, then obviously in offline, online, then the complexity of the marketing mix model becomes very hard to attribute the final sale, right? So therefore that’s the reason I asked that question.

Scott Desgrosseilliers (46:58.169)

Yeah, you got to start ground up. So each channel needs to have its own. You need to get down to the campaign level, set the intention, measure them all at the granular level as best you can. You can’t look top down in my opinion. I mean, that’s something I do a lot of stats. I used to be a financial systems trader. Before that, I was a baseball stats fanatic and it always comes down to baseball stats. You come right down to the pitch and work your way up.

And now baseball stats are taken over by all the stats guys and all the different, you know, approaches that the plate are now different because they broke it all the way back down to there in terms of what’s the percentage chance we’re going to win on each pitch based on how you’re swinging, what the angle of your swing is. So we’ve got reverse engineered all the way back to how you’re going to swing the bat on a pitch to get, and that would be my advice to them is to get, you got to get granular to get results at the macro.

ContraMinds Podcast (47:27.64)

Perfect.

Q12 — “Experiment Boldly. Be Accountable. Do Both.”

ContraMinds Podcast (47:57.774)

So if having spent so much of time, a couple of last questions that I have, if I were a marketer or a CMO, what are the most important skills that I need to have in myself and in my team over the next, say, few years? Which do you think are extremely important having seen and worked with large brands? What are the skills that you think are important?

Scott Desgrosseilliers (48:25.723)

I guess there needs to be two things that sometimes are at odds but shouldn’t be. One is the ability to experiment and experimentation mindset. You got to be able to take experiments and fail as long as you learn from them. 

Cultivate an ‘experimentation mindset’ – experiment, fail, learn.

And so if you’re coming in, it’s a high pressure, hey, we need results, we need results. That’s when people will default to, okay, let’s just pour money on retargeting to pump up the in-platform ROAS so it looks like we’re doing well. And that’s not the solution.

Additionally, you also need accountability because otherwise if everyone’s just doing ideation and there’s no clear number, eventually it’s getting aligned on scorecard of a person’s job role, I think is really important as well. 

I run, you know, I’ve seen inside a lot of marketing teams and clear lines of accountability, but the willingness to experiment are two things that lead to success, but they both gotta be clear or they’ll compete against each other and you won’t get either. That’s what I say.

Q13 — “Start Local, Get Wins, Build a Portfolio.”

ContraMinds Podcast (49:32.786)

What would be one piece of advice you would give to a young marketer entering this business?

Scott Desgrosseilliers (49:44.188)

Well, there’s a lot of opportunity because it’s a complex field. Even though at the end of the day, just trying to get people to buy stuff. Well, I would say start with some local brand where you can get some…

…the sooner you can get work experience, the better. And the beauty of the field is it’s not really a academic base. Like I’m not going to say, Oh, I want to hire a marketer because they’re from Harvard.Who Who cares? 

Are you good at running market? 

Can you fit the jobs to be done that we need as marketers, as a brand is more important. And it’s easy to get real world experience. 

And then that’s when you learn is on the job. 

So if you could even just approach your local smoothie place, a clothing store or what have you, and start working maybe at a lower rate than you aspire to, but you get the experience, then you build a portfolio because as someone who’s hired and fired a few marketers, but also seen a lot of agencies inside of them and a lot of brands, it’s the values in getting quality work done, communicating clearly, and ultimately driving more revenue. It’s not always about the big aha idea. 

It’s about just getting people to methodically do the work that they should be doing and aren’t a lot of times like email sequences. A lot of people don’t look at their, just can you write a couple of good emails and get them automated for someone and then track, look, they made more sales. Small brands and shops would be thrilled with that. So I’d start local because the challenge is the ad spend gets bigger and the scope gets bigger, but it’s still about strategically talking to the brand owner. What’s unique about what we do.

How do you sell it in person that enabled you out of this business and then figuring out how to do it online? Still the same challenge as you go further in your career.

Q14 — “Skip the Textbooks—Follow Practitioners Who Ship.”

ContraMinds Podcast (51:52.236)

If I were a marketer doing the stuff that you’re talking about, having budgets, spending the money, looking at marketing effectiveness, are there some books that you believe you can recommend or websites that where I could go there and probably start to learn this? One is really your course. The other one, what are the other sources of…information or knowledge which you have found useful which you would recommend.

Scott Desgrosseilliers (52:25.627)

Not a lot of marketing books, actually. I follow a couple people on LinkedIn. Ralph Burns on LinkedIn has a lot of good strategies. John Moran, those are two off the top of my head from this agency, Tier 11. They’ve got a lot of good content they put on LinkedIn. Usually it’s LinkedIn thought leaders. It’s not so much books.

ContraMinds Podcast (52:44.992)

Okay. Perfect.

Fantastic. 

Q15 — “Persistence Beats Brains and Budget—Focus on Outcomes.”

What’s one piece of advice that you got? I’m sure you would have got a lot, but something that’s remained with you till date.

Scott Desgrosseilliers (53:02.107)

Probably around persistence from my father-in-law. He was like, you know, he’d get me a plaque, it’s around here, it’s right over here. But it was around that, you know, 

…it’s not the person with the most skill or the person that has the most money or the most intelligence. It’s often the most persistence that ends up winning the battle. And I found that to be true running a SaaS. Like we might have some great idea and the market doesn’t take to it.

Or there’s always a lot of ups and downs, but just persisting and continuing to attack the problem. And then I realized, just focusing on what are our customers’ outcomes and are they getting them or not, that orientated us really well once I took that mindset. And that was a persistence of staying at it. 

So I’d give him credit for that.

CLOSING

ContraMinds Podcast (53:58.654)

Thanks, Scott. With that brilliant piece of advice, which I completely agree and it resonates totally with me. It was a fantastic conversation. was an hour of, I would say master class on attribution, marketing attribution and thanks a lot. And I really think there’s so much to be done in this area and we’ve just scratched the surface. I think…

This conversation will inspire people to know more about what you’re doing, the course that you are offering. And I’m sure this is a place where people will go to if they really want to start off and then get better at marketing attribution. Thanks a lot for your time and lovely talking to you.

Scott Desgrosseilliers (54:47.161)

Yeah, wonderful interview. really appreciate it. Thank you.