How to Kill Assumptions and Make Marketing Predictable With John Readman From BOSCO [AMP 246]
Too much marketing is based on guesses not backed by data. Paid tactics, like pay-per-click (PPC) and social media advertising, can burn through your budget when guesses are wrong. How can you use data to make marketing more predictable to forecast performance and adjust to shifts in trends to increase your ROI?
Today’s guest is John Readman from BOSCO, a digital analytics and predictive modeling platform for retailers and eCommerce companies. He discusses what it takes for predictable marketing to be successful. It involves understanding historical data, performance, and trends across a client's channels.
Some of the highlights of the show include:
- What is predictable marketing? Getting all data in one place on an ongoing basis
- Why should marketers make decisions driven by data, not gut instinct/intuition?
- Data and Decisions: Depend on volume, understanding data to base decisions
- Challenges: Digital marketing data is used to scale ROI in one particular channel
- COVID Comparisons: Causal effect of supply and demand during the pandemic
- BOSCO: Helps marketers predict future w/ machine learning, Bayesian statistics
- Out-of-Date Numbers: Forecasting runs scenarios, planning, and model analysis
- Different Data Sources: Connect platforms to quickly predict what could be done
- Wasted Budget? Run and identify data models that are hugely, scarily accurate
- Two Key Metrics: Cost per acquisition and understanding that by channel
- Clients/Conversations: People make decisions emotionally, justify them with data
- Control: People get nervous about not doing things the traditional way
- Predictive Analytics Platform/Practice: Get buy-in by leading conversation with potential results, starting small, and using data to quantify progress and success
Quotes from John Readman:
- “If we've got the right data in the right format, and we understand what is going on around certain targets, what makes it predictable is understanding the metrics and the outputs we are trying to achieve.”
- “Fundamentally, why do people need to make data-driven decisions to really explain where they're spending their money, where are they getting their ROI, and then how can they scale it?”
- “It all starts with getting all your data organized in one place, then looking at what I am willing to pay to acquire a customer, and then maybe looking at customer lifetime value.”
- “The thing to stand out will be a better proposition, a better product, and a better promotion, which is sort of the traditional marketing going around in a full circle.”
How to Kill Assumptions and Make Marketing Predictable With @JReadman From @AskBoscoIndex
Click To Tweet
I think the hardest bit about this is getting all the data in one place, and I think that's why not many people are familiar with predictable marketing. But it's becoming increasingly easier now with APIs and connecting data together to try and do this on an ongoing basis. That's certainly what we believe is the term predictable marketing.
Ben: Sure, and I think it makes sense to plant your flag, so to speak, with that definition. I think that makes a lot of sense. Before we get too far along into the actual nuts and bolts of what predictable marketing really is and how it's executed. Before we even get that far along, why is it important that marketers make data-driven decisions rather than or perhaps in addition to decisions that are driven by gut or by intuition?
John: In the world where there is so much data now, I suppose it's very difficult not to make data-driven decisions. I suppose I'd add to that is to make the right data decisions or the correct data decisions because one of the great things about digital marketing or ecommerce is the huge volume of data we now have access to. I think the challenge is understanding which bit of data is the most important to make the decisions around.
I think, fundamentally, why do people need to make data-driven decisions to really explain where they're spending their money, where are they getting their ROI, and then how can they scale it? I think the challenge often comes when we've been used to scaling, maybe in one particular channel.
What we've certainly seen over many years is people carry on spending more money with Google, keep spending more money with Google, then they think, well, we should try and spend some money with Facebook. And then Facebook measures things in a different way so we're getting some different data points where maybe we're not confident about those.
I think the one question and discussion we have with many people is about the whole picture of all your digital marketing data, rather than just how could I make more money out of Google, and should I spend more money with Google. How can I make more money out of Facebook? It's like where should I spend my next pound across my whole digital marketing?
Then if you understand your data, hopefully, you could answer that question of actually, I should spend it on Amazon, I should spend it on TikTok, or I should spend it on Twitter rather than just keep spending when we Google. If you went to Google and said, where should I spend my next pound? Google would probably say Google, as with Facebook.
Also, I think some agencies are biased based either on a knowledge bias because they may not understand TikTok or they may not understand Snapchat, so they can't necessarily give you the right advice. Or maybe biased because of the commercial relationships they've got with certain platforms or different technologies. I think a brand or a big ecommerce retailer needs to be looking at their data across all the channels to make the best most informed decisions going forward.
Ben: Sure, and I suppose that when you approach things that way, you can start making better decisions for yourself rather than trying to follow which direction the wind appears to be blowing based on what other people are doing.
John: Yeah, there's an example we once had with a previous client who was moving from direct mailing to digital. I won't mention the product or the company to save people's embarrassment. They said, there's always an increase in sales in this particular product at this particular time. I said, right, that's interesting. It doesn't seem to make much sense. We analyzed some trend data, some seasonality data, and it didn't really make any sense. I said, well, what else happens then? It’s when we publish the catalog, send the catalog out, and put the product into sale.
They were actually affecting the demand for the product, which is why the products sold more. There wasn't necessarily more demand. I suppose it's also understanding what is the causal effect of the demand. Again, we saw a lot of different demands in the pandemic across the world in many different countries. People now are really struggling with year-on-year comparisons. We're not going to have the same demand for certain products now because the shops are open again and we can go back to shops versus what we had last year in ecommerce. You're right with all of that.
Ben: Yeah, sure. This is something that you have a lot of experience with your company at BOSCO. For you and for your company, how does your company execute this? How do you help marketers predict the future so to speak?
John: This could take quite a long time. To get a proper answer, we probably need one of our Ph.D. data scientists because I'll probably simplify it too much. Fundamentally, we use combinations of machine learning and Bayesian statistics to use all your historical data. We analyze all your historical data, both performance data, demand data, and sales data so marketing spend. So virtually like a spend cost versus revenue model. What did we spend on marketing, what did it cost us, and what revenue have we got back?
We then looked at the KPI that you're trying to optimize. Are you trying to optimize as a return on ad spend, or are you trying to return at a cost of sale, or is it cost per acquisition? Is it you're trying to optimize your maximum number of units sold, or the maximum traffic, etc? There's a certain number of things you might want to optimize to, then what we do is we look at all your historical data and then we look at also the data of the demand within the market.
We've got connections within our BOSCO platform into the Google API, the Bing APIs, the Amazon APIs, Facebook APIs, all the different platforms APIs. We can go see what the demand for your products is there under keywords within your market, and then we can also see how visible you are within your market compared to your competitors. Then we can start spotting opportunities, but not just opportunities. How to spend more on Google, we could also highlight well, actually, you should pull back on Google, and maybe spend more on Facebook. Or actually, you're overspending on Facebook, you should maybe spend some on Amazon.
We're looking at this across all channels, and also we’ll index you against your peers within that as well. It will say, they're doing slightly better than you in these channels because... It's a comprehensive model and it will forecast by week or by month. It will say this is what we think is going to happen next month if you carry on with things as they are, and this is what's going to happen next week.
Sometimes people find it a little bit more interesting, Ben. The chief financial officer, head of finance might come in and go, well, how much could we spend? What is the optimum or how much could we spend? What's the maximum amount of money we could spend to get back and hit our targets?
Then you might get the chief marketing officer go to the head of digital, head of ecommerce. They go to the agency and they say, how much could we spend? Everybody crunches some numbers, does some stuff, and goes to an expensive agency to find out. Then the answer comes back, and it's this much, but then they don't necessarily break down where or how.
The challenge with that is that it's also out of date. As soon as you crunch the numbers, it's straight out of date. We've built into our platform a forecasting tool where you can run scenarios, and do scenario planning and model analysis. Again, my background before is I used to run digital marketing agencies. Our historical answer to this is, we need to go into a new channel. Let's have $10,000 or $50,000, we'll do a test and learn campaign, and we will see what happens.
Basically, we'll go use your money to find out what is good in the situation. The good thing for you is you'll also find out and the good thing for other clients. We'll find out for them whilst we’re using your money if you're an agency. I thought with the availability of data nowadays, we shouldn't have to do a test and learn. We should be able to run a model or a simulation to go hang on, if we took $50,000 and put that into Facebook against a [...] target of three or whatever the number is, what is going to happen? Is there enough volume? Is it the right average order value? What amount of impressions or clicks are we going to get?
Our software will give you a profile of exactly what you need to do if you're going to spend an additional $50,000. But it would also say, here's what you could spend, and here's what you could return for your category for your product in your sector across all those different platforms.
I think what this is going to encourage is a healthier conversation and debate between the C-level executives and the vice presidents who are running brands and ecommerce businesses, their agencies, and/or their digital teams. Because hopefully, everybody's now looking at numbers where they can try and drive performance and make better decisions. Whereas before, everybody's just had to trust the numbers that either come from the marketing team or come from the agency. You haven't necessarily had all the numbers in one place to be able to challenge or make better decisions.
That's how we do it. It does rely on connecting a huge amount of different data sources. But nowadays, because we have partnerships with all the platforms, it's a relatively straightforward process. It takes less than 48 hours to onboard somebody. You could connect all your data sources and have a very simple view of what could we do next month relatively quickly. That's how we do it.
As we get more and more clients on board and more and more data, the algorithm should get smarter and smarter and smarter, and we're constantly learning. That's how we do it. That saves getting into the whole regression modeling and all the different statistical things that I could go into, but that's not my background. I'm more on the commercial side of it. We've got Dr. James, who's our Ph.D., data scientist who leads the science on it.
All the models we've been running for all our clients have been hugely accurate, scarily accurate, and I think the biggest thing is about identifying wasted budget. Because if you do speak to the networks or the platforms, as long as the campaign overall is profitable, they're very rarely going to turn around and go, you're wasting money there, you're wasting money there because that's profit for them as the advertising platforms. I could go on about this all day, Ben, sorry. Hopefully that answers the question.
Ben: I think it's super, super interesting. I think that's a very, very thorough answer. I will say that our audience is not data scientists, and probably not PhDs either, so I think that explanation actually works just fine.
If someone listens to the show—a marketer or a marketing team—wants to move into a more predictable direction with their marketing tactics, strategy, data, and how they apply that data. Whether they use your platform or they do this a different way, where would you recommend they start? Let's assume they may be using data the way normal marketers use data.
John: This is a great discussion. Even if people need to start looking at this anyway, and obviously, I'd want everybody on the planet to sign up for BOSCO, that would be amazing. But realistically, if you just start thinking about how you acquire your customers and your marketing data in a different way, then it will help you make a better and more predictable future for your budget requirements and how to invest.
The first thing I would say is, ideally get everything into one place. If you can get all your marketing and sales data into one place, whether that be a massive database in the Google Cloud, or just one big spreadsheet, especially depending on how big your business is, and then what we really want to be looking at is what was the source. Where did that come from?
Then I suppose ultimately, there are two really key metrics I always encourage all business owners or all heads of marketing to look at. One is cost per acquisition. What am I willing to pay to acquire a customer? That might be different for new customers or repeat customers, what am I willing to pay?
In order to understand that, you really have to understand all the costs of running your business, and then if you've got say a $100 profit in that product, how much of that profit are you willing to spend to acquire that customer? You need to get your head around that. Then I suppose if you have multiple touchpoints with the customer, a product that might be able to be sold once every six weeks, or is it a more repeatable purchase, what is the customer lifetime value?
Because again, you might decide, well, I don't need to make a profit on the first sale. I'm okay because I know if I get a customer breakeven on the first sale, they're going to buy six more things over the next however many months, and that's when we'll get our profit. I would say, first of all, is to understand cost per acquisition, and then ideally, understand that by channel. What is that by channel? That would start you off on the road to understanding and getting to be more predictable of the way you should be investing your money or moving your money around.
Then ideally, and this is a difficult conversation than to have with finance or to have is, I genuinely believe you shouldn't have a marketing budget per se. You should work on a performance metric. If we want a 10 times return on ad spend to grow as we just keep spending $1 to get $10 back, and actually, there should always be enough more for $1. I think the mistake a lot of people make in performance and predictable marketing is they say, well, I've got $100,000 to spend and that's it. Go get me as much as possible for $100,000, which is actually a very different strategy to go get me as many $10 for my $1.
I would just get people to think about scalability, but you need to understand your metrics first. Then I suppose the next step on from all of this—which may or may not, it might be a whole separate podcast—is understanding attribution. Because once you start looking at this, I think the days of people making a decision by thinking about something, taking an action, and then making a purchase, booking a lead, or doing whatever it is they're doing on whatever platform they're doing it, those days have changed, they're gone.
Now I think there are multiple interactions or touch points between a customer and a brand before they actually go ahead and make that purchase decision. This then opens up a whole plethora of questions around which bit of marketing should get the credit? There's a whole evolving world all-around attribution that I think people need to start thinking about because most people are still going on last click wins or last non direct click wins model.
Actually, if you really understood the impact of upper-funnel display, Facebook, video, or programmatic, you might look at your budgets and be able to predict the future in a much better way. It all starts with getting all your data organized in one place, then looking at what I am willing to pay to acquire a customer, and then maybe looking at customer lifetime value.
If you're really getting into the data, start looking at attribution or then building an attribution model. I would say, maybe don't just rely on Google Analytics because there are better models that might fit your business. I suppose, just to finish on the attribution point, all attribution models are wrong. It's just trying to find the one that's least wrong. There isn't a right answer, unless you sell one product to exactly the same person all the time in the same way, it's going to be wrong.
What we need to do is try and find a model of best fit for your business. I think that once you understand that these models are flawed, an attribution model is better than no attribution model, then you're off to a reasonable start.
Ben: Yeah, you're trying to make that data directional rather than definitive because it reaches a point where it's impossible to maybe effectively attribute every single time someone interacted with your brand, saw something about your brand, or interacted with something and try to put that all together. There are too many variables.
John: It just gets harder and harder and resistive. The other thing we've been working with some clients on now is they're sort of saying, we can't spend any more money on Facebook and it works within our constraints and our targets. We're now saying, I think we need to improve click-through rate and conversion rate, and that's only going to be about improving creative.
Actually, what the data tells us when we get to a certain point is right, you need to go back and review the creative because I think every fifth post now on Facebook is an ad. It's just like, how do you stop that scroll? Is that about targeting and bid optimization, or is that more now about creative?
Which I probably shouldn't be saying as someone in the predictive analytics business, but I think creative is going to become more and more important as more and more important people get good at predictable marketing and predictive analytics type modeling. Then the thing to stand out will be a better proposition, a better product, and a better promotion, which is the traditional marketing going around in full circle.
Ben: Like with any new tactic or tool that you want to implement with your team, if you're interested in diving into a predictive analytics platform or practice, you're going to need to get buy-in. The way that you get buy-in is by leading the conversation with the potential results, starting small, and using data to quantify progress and success. If you can do those three things, you can ensure you'll be more successful and make your own marketing more predictive, or enjoy whatever other change or benefit it is that you are trying to achieve. Now, back to John.
Something else to maybe touch on for a moment here is you say that you believe marketing teams or companies shouldn't have a marketing budget, which I think makes sense. A couple of jobs ago, I worked at a company that operated that way, at least as it pertained to pay-per-click advertising, there was no budget. As long as it was making money, they just shoveled more money into it. So the ad spend was incredibly high, but so is the ROI.
How would you recommend or maybe if this is something that you do with clients right now, if somebody is in a position where they're feeling constrained by their budget. They have the data to show that if they just increased their spend by X%, they know that the return would be X%, but they can't really maximize profit or maximize return because they've been told this is how much money you get to spend and you’re cut off after that.
How would you recommend that they approach the conversation—whether it's with the CFO or whoever—to maybe change that thinking and view marketing as something that can just grow money rather than something that you cap like it's an expense? You're just going to spend this much money and that's it.
John: Well, I think some of it comes down to the perception of what marketing's role is. That's interesting. It's like, are they the chief marketing officer, are they the chief revenue officer? Where do they sit and what's their role? I think, fundamentally, and I've been involved in conversations like this before and helped clients manage this sort of conversation, it's almost like asking. Because most things in life we make decisions emotionally and then we start justifying them with data and logic. That's just the world we live in.
I'd almost be tempted to go see the CFO and say, how would you feel if I could deliver this much in sales uplift? How would you feel and what impact would that have on the business? And get them to think about it. And then they'll start saying, oh, that'd be great. We'd feel amazing. We could go in this direction, what would it mean for the business? Well, we could create all these new jobs, we could do this, or we could do that?
Then it's like, well, what if I could show you a way we could do that and sort of taking them on a bit of a journey? Then you could go, right, this is how I believe we can do this. If you look at the way that performance marketing channels work, particularly pay per click, pay for social, or affiliate marketing, the biggest channel I think that's underutilized is affiliate marketing because that is your only pay on performance.
It got such a bad reputation in the past with maybe discount codes or all those sorts of things, and people have shied away from it. Anyway, I'll come park that. But going back, if we position an emotional story and then say, well, let me show you how rather than going, I want to have an unlimited budget and go just keep spending because people think that's irresponsible.
Ben: Yeah.
John: Also, people are nervous about control. People are nervous about it's not the traditional way of doing it. Normally, you’d put your budget in and then you tell people what you're going to spend your budget on. Then if you stick to your budget, that's what happens. Whereas what we're actually saying, we're going to do it on a return on spend model all across the sale model and we're just going to keep scaling.
The companies that really get this and also, the fact that we calculate all their costs of sale every month or even every week, based on the exchange rate, delivery costs, returns rates are the companies that really managed to scale. I would sit down with the person who's got the purse strings and talk about well, what if we could do this? And if I could show you how we can do 4X uplift, how would you feel and what would that mean to the business? Then agree to a small trial, that sort of thing.
Rather than just saying, let's just go ballistic and spend all our money, maybe say, well, why don't we take two products, three products, or two categories of services that we offer and why don't we try it? Then let's prove the case and then we can roll it out across the wider business.
I think it's about getting the emotional buyer to agree to some sort of test, then agree to some measurement, then prove it could work, and then roll it out across the rest of the business. Hopefully that's a useful approach. It'll be interesting to see if people listening to this try that and it works.
Ben: Yeah, absolutely. I think that's a very smart approach and I think you could follow a very similar framework. I think really for having any kind of budgetary decision within your company as it pertains to marketing really regardless of what it is you're trying to get done.
The last question, I'll throw your way, maybe just to kind of drive some of this home a bit. What kinds of results can marketers expect when they start using data and analytics in a more predictable fashion over what they're doing before? That's a very broad question, but generally speaking, what does the before and after tend to look like when marketers start to move in this way?
John: I won't mention brands, but I'll give you some examples and some numbers. We work with many, many different companies who came to us and said, we spend this much a month. Maybe you're looking down at the long end of the telescope, you need to be understanding your cost of acquisition. Could you physically make enough widgets and dispatch enough widgets if we could sell all the widgets? We can get you all the demand for the widgets in the world, could you sell them all?
They're like, well, we often refer to this as a champagne problem if you get me too much demand. We have worked with clients and as long as you've got the delivery capability in them whether the logistics, the fulfillment, the service, and the people to deliver whatever it is you're selling, as long as the rest of it is [...], it is almost like turning on a cash machine or a conveyor belt. As long as it's done correctly and done within the constraints of KPIs and the agreed targets around [...].
We've had companies who've scaled from maybe $5 million a year to $50 million a year inside a year using this type of marketing approach, understanding the data, and understanding where to invest it. The challenge then is they're saying, we want to get to $500 million. Then you're like, well, is there enough demand in your niche, your sector, or in your vertical?
Is there enough demand and there will become a point of diminishing returns? You'll get to a point where actually, and I think the other interesting thing people obsess about and get addicted to when they get into performance marketing is how much more can we spend? How much more can we spend? We want to keep this working.
Actually, that is the right approach to start with, but then people stop focusing—and we had this conversation only on Monday last week with a client. They were obsessed with traffic at the top of the funnel, but actually, the biggest lever you can control is your conversion rate.
If you move your conversion rate from 2%–4%, that's going to double sales, whereas we're not going to double the volume of traffic very quickly. What we've been trying to work with a lot of clients is it's great to have predictable analytics and marketing, but you also need to understand the data that's available to you across the whole piece.
Actually, the biggest lever you can pull is to improve, making sure that the number of people that come to the site actually engage and interact in the way you want them. Whether that's by improving your UX or your CRO, and if they're making sure every click matters that come through to your site.
I would say if you look at the marketing in this way, it can hugely affect your business, and also I'd say, actually looking at this sort of advanced version we talked about. One thing is the cost per acquisition, but actually looking at which channel is having the biggest effects, agreeing on an attribution model, and getting senior buy into the attribution model will unlock budgets that probably weren't there before.
If you can prove to them, look, when we spend on programmatic video on YouTube or whatever, this happens. That also has an uplift effect on paid search, shopping, or whatever other channels you’re marketing with it. I wouldn't say profound effects, but it can rapidly accelerate performance because if anything, the pandemic across the world has accelerated people's ease and confidence to transact online anyway.
Now is the time to get into the data, really understand how we can accelerate this further because there's a whole new load of customers that have come online in the last two years that probably aren't going to go back to the high street.
Ben: Yeah, certainly. It's really probably permanently changed a lot of consumer behavior that way. In ways that we probably don't fully know yet like how things are going to shape out. It makes sense that it really makes it all the more important to try to use data to paint as clear a picture of what's coming next is what you can because nobody really knows.
John: Yeah, and that's it. The more accurately we can predict the future performance of the business, the better we can plan. Also, I think it can also help. If you're a retailer buying products, if you know I expect we're going to be able to sell this in the next three, six months, you can then be more efficient with how much stock you'll buy. Which then probably means you have to then put less into sale, which means you hold up a higher margin so you then make a higher profit. It has an impact across the whole business, the better you get at this. There are huge volumes of data out there that can help you do all of this now.
Ben: Yeah, absolutely. That does it for all the questions I had prepared for you, John. Before I let you go, do you have any parting thoughts or just anything you'd like to throw out there that we haven't touched on yet that our listeners might find interesting before we wrap things up?
John: We talked about lots of interesting things, certainly not the England football team. We have another view around performance marketing, and part of our company also helped people in-house their digital. Then our technology platform helps them monitor, measure, and optimize across all different channels.
One thing I would suggest I think the future of performance marketing and predictable marketing is actually executed in-house. Rather than paying people a fortune to go away and do the heavy lifting and the thinking, then they own the IP to that thinking. You need to learn how to do this yourself, you need to get some training, or you need to recruit somebody who can sit in your team and give you advice within your four walls. Because I think it's very easy to get an instant answer by paying some experts, but then every time you need the answer, you need to pay again.
I think 15, 20 years ago, nobody knew what was going on at all. Now, it is relatively commonplace to understand there's plenty of experts available that you can recruit within your team rather than always having to rely on different agencies doing different things. I think it's often too easy to just go to another agency. They're going to help us with this, they're going to help us with that. That's one thing I'm pioneering and pushing is encouraging brands to take control of their data and bring it in-house and really understand it.
