Life as an Analyst

by Cathal Melinn

Posted on Jul 10, 2020

Expert analyst Cathal Melinn talks with podcast host, Will Francis, about what data analysis is really all about. In short his message is: "sessions don't buy, people do". They cover educated guesses, soft metrics, assumptions about human motivations, the creative joys of Excel, site optimization and A/B Testing, different social platforms' definition of a click or a conversion, the messiness of the attribution model, and even a bit of Aristotle.

The full transcript is also available below. And check out Cathal's introduction to the power of data, in his webinar on Data-Driven Marketing.

.pdf
Podcast Transcript - Life as an Analyst

TRANSCRIPT

[00:00:04] Will: Welcome to the "Modern Mindset," a podcast about soft skills brought to you by the Digital Marketing Institute, exploring those personal skills that no one really teaches you but are vital to success in your marketing career and keeping you ahead of the game. I'm Will Francis and today I'm talking to Cathal Melinn, a commercial analyst who helps businesses make better decisions through analytics. It's a crucial skill in any line of business, but particularly in marketing where we must measure and optimize as we execute in today's fast-moving digital world. Welcome to the podcast, Cathal.

[00:00:39] Cathal: Cheers, Will, it's good to be back.

[00:00:40] Will: Good to have you. So, you're a commercial analyst. What exactly are you analyzing for your clients?

[00:00:47] Cathal: Well, that's the thing, I mean, because so much of our transactions and different things happen online, I'm managing and I'm understanding consumer journeys and trying to get numbers and turn them into meaningful picture that represents what a human does with their mouse and their computer to buy a product or visit a website and different things like that. I absolutely love it. It's such an interesting world because, you know, you're presented with a suite of numbers - your metrics - and you can split them up by different segments, like the computer device. They're your dimensions. That's all you're presented with. And you have to paint a picture. The challenge is for you to draw on your experience, to draw would you know about the customer, about the client, about the product, and turn these numbers into something that is meaningful and useful for business strategy and other reports on the success of an action that you've already taken or allows you to plan something that you're about to take. So, yeah, it's just turning numbers into something else.

[00:02:01] Will: So, your job is actually, you know, an analyst's job isn't just getting the numbers, it's turning them into a meaningful narrative or story and useful information that the client can then actually act upon. They can not only look out and go, "Oh, that's interesting," but they can actually look at it and say, "Right. Now we need to know what we need to do. And what changes we need to make." That is for you to interpret that?

[00:02:28] Cathal: Yeah. And that's the thing. But, like, the challenge is, I suppose, for all of us and anyone in digital marketing analysis is, you have so many numbers and so many things, and it's, where do you start and what do you do, and what's actually meaningful, and what is low priority? And there's a couple of levers and there's just a couple of straightforward for any kind of business outside of the analytics tools that we use in digital marketing and the different ad platforms and the web analytics. Just how to actually read the tea leaves and turn those into something that's actionable. But there's levers we can pull, there's certain go-to areas that we can start...

[00:03:09] Will: What do you mean by that?

[00:03:09] Cathal: Well, that's the thing. So, you know, I generally, if we pick something very straightforward like eCommerce and I generally have a three-point strategy for any kind of analysis of the numbers. And the first point is, I want to identify what's valuable to me or my client. And what's valuable to me is a sale, or a conversion, or whatever it is. And what I'm interested in is clusters of these or hotspots. So, where are my conversion hotspots? Meaning, where am I getting a low cost per sale or lowest cost per acquisition? Where am I getting high numbers of sales, or where have I got a high conversion rate? And what I mean by that is, now I know what I'm looking for and they're my hotspots. These are my, you know, the areas, if I put more money into a hotspot, I can hope to get more out of it because you know it's active, you know it's demonstrated clear intent, it's given you good numbers in terms of conversion rate, CPA, or just conversion scale.

[00:04:18] So, go after that stuff and then work on the incremental things. So, I begin with hotspots and I say, "Well, where are these hotspots?" And I look at that in two levels. One is, what market? Where are they actually physically located? And this can be for website visitors. Where are your most valuable website visitors or your most engaged social media users? Where are the clusters of those groups? You know, in terms of where are they actually physically located. Because that's the first pivot that I would use to identify how I can activate a cluster in a certain region. You know, so I check, where are my clusters by my market and then identify them and I prioritize them.

[00:05:07] Will: And is that specifically geographic, or could that be gender or age or other demographics? Are you specifically talking about geographic?

[00:05:14] Cathal: Starting point would be geographic clusters. And then, because I just want to know, where are they? Where actually are they? And then I'll start segmenting. Once I've identified the different hotspots, I'll say, "Well, what devices are they using to access the internet?" So, this is still, it's still quite functional at this point. You know, it's just, you know, we're looking for, you know, the high-engaged people. We want to know where they are and we want to know how they get to my social media profile, how they get to my website, which is where are they? And then how do they engage with me? Then it's my job to understand who they're. So I know where they are and how they get to me, and in cluster numbers. And then I go, well, you know, what's their gender, what's their age, what are their interests? And I'm able to segment that.

[00:06:03] And what it does is it allows me to focus in on my high-value clusters and know who they are, where they are, and how they get me and then make their journey easier and more relevant to them. So, if they predominantly access my social media with a mobile device, I want to make sure that like everything is fitting in a mobile screen or can be bought with a mobile transaction. If it's in a certain region with certain kind of cultural traits that I am adhering to them or leveraging them. And then, you know, if there are certain kind of micro indicators around like their motivations or their demographics, am I speaking appropriately to those as well? And, you know, I suppose an overlay to all of this as well is, well, what channel do they use to get to me? Is it, you know, an email? Is it just typing it into the browser? Is it a search? Is it through social media? Is it affiliates? How are they getting to me? And just, they're my areas there, that's where I start prioritizing my numbers because we have all the numbers and analytics, all of the numbers are there, and it's just too overwhelming. So you have to know what can I push away and what can I focus on? So...

[00:07:23] Will: So, firstly, it's about focusing on those most-high value, high converting people and identifying and then in terms of geography and working out device usage, culture, location, and source, how they're actually finding their way to convert, like Search versus Twitter or something like that.

[00:07:47] Cathal: And that's kind of...that's the most straightforward template. And it can be adapted, I mean, the thing with this is it can be adapted and you can bring in other metrics. And depending on the maturity of your business, you might want...like if it's a quite mature, you know, business where you're selling loads of stuff or you've got a decent audience and a well-established brand, actually just, you know, those areas are fine. But if you're not an established brand, and you're not selling a lot of stuff online, you might want to look at things like engagement metrics. So, it's your softer metrics. It's not your conversion metrics, because you don't have any. You know. So, it's the mindset of actually, because we can't track stuff, we have to track everything. But that's not...one of my favorite analysts is Kauhis and what he considered is a data puke where you're just throwing every single piece of data down and hoping that that equals insight or that equals value. Like my recommendation for anyone doing analytics or anyone managing an analytics, you know, kind of scenario is, fewer metrics. You know, it's 2 or 3 high impact metrics that give you the 1 or 2 things, actions that you need to take rather than 10 metrics that are all dependent on each other and they don't give you a clear action plan.

[00:09:22] Will: So it's actually, yeah, it's easier to act on a simpler, smaller set of metrics because you're going to get a clearer instruction from those metrics. Whereas if you've got 10 metrics, they could start telling you a much more complex picture that in some ways can kind of disagree with itself or have kind of contradictions or become so narrow and focused that you end up just talking to one person who's on Android and, you know, in Blackrock in Dublin, who's female age 35. And so, okay, so, you talked about a three-step process.

[00:10:01] Cathal: Yeah. So, the three-step process is, so, we're trying to identify clusters. So, where are they by location and device? Why are they there? And that's your things like your motivations and your gender. And then the last piece of the process is your action piece, and that's, what do you need to do to do more of that. So you've identified that there are clusters, you identified how they get to your site, you identified the nuances about those people. Now you have to take a mental leap, which is, it's not so clear cut. This is where we actually move outside of analytics into insight drawing, where you kind of, you draw on your experience and your history to say, "Well, I know where they are, I know who they are, I now have to think to myself, how can I do more of whatever I'm currently doing that is leading to this behavior?" And whether it's an offer discount or do more social media or whatever, but your foundation is to identify who they are and why they are there, and then your final step is, how can I do more of the good stuff?

[00:11:09] Will: Yeah. So, it's like everything in marketing, there's a... you can never, I get a lot of people in my courses who expect that there's a series of steps. There's a formula that you can follow and you will arrive at the result. And, you know, as anybody who's been in marketing for a number of years, you know that's not the case, that all marketing, and including analytics, which you would think is a very kind of formulaic discipline in a way, but it requires that human intervention, that creatively to think, have that conjecture initially of, why might these people be there? There's not going to be a kind of an output for that, that's spat into your lap. It takes a bit of imagination to say, "Okay, I have to use, kind of, my powers of empathy, of imagination and creativity to wonder why they might be there and come up with a good theory, a hypothesis that I can confirm later down the line when we get more data and we try things where we explicitly test their motivations with explicit messages that speak very directly to those motivations and see if that works as a piece of ad copy for instance." Right?

[00:12:23] Cathal: I think you said a really interesting word there actually, when you went, "I wonder why they might be there." Because wonder is approaching a situation without knowing or without having any preconceived bias or opinions, you're just kind of allowing, you know, an open, like an open understanding of what might be. So, wonder is a great word and it's just something I picked up on there that like I would definitely approach it like that, you know.

[00:12:49] Will: It's true. I think you have to wonder and you have to investigate and interrogate and be a bit of a detective. But again, I could say the same about, you watch any great detective movie or detective show and the good detectives in these shows, they have creativity and imagination. They don't just work with cold data. That's a starting point. And you layer that kind of human stuff on top. So, and then once you've conjectured why they might be there, again, you have to become creative about how you can more closely resonate with the kind of people that you think could come in to your site and buying your highest value items, right?

[00:13:35] Cathal: That's the thing, you know, in my head, when I think of digital marketing and analytics, I think of those old railway station, signal stations that, you know, they have all these levers in it and they change the tracks and all that. And I just think to myself, and I know our listeners can't see it, but, I'm literally, I'm holding a lever with my right hand and I'm pushing it forward and saying, "Do more good," you know, and I'm holding another lever with my left hand and I'm pulling it backwards and I'm saying, "Do less bad," you know. So, you have the data that has identified your clusters of good, do more of that good stuff, and you've identified, you know, possibly what is less effective based on what your initial assumption was and do less of that, you know. So, do more good, do less bad. And that's, in a very simplistic viewpoint, how I would kind of approach analytics. And then it will iterate and nuance and develop as more data comes in. But ultimately, you do need a starting point to get a foothold into the thousands of different metrics we have available.

[00:14:42] Will: Well, that's the point, isn't it? There's a couple of things to pick up there. First one is that, it's an always-moving feast, you know, the game is always on and it's active and it's moving. So you're always optimizing, you're always just improving on yesterday, aren't you? That's the game of analytics.

But what I find is, you know, a challenge is when you start a project, so, say you have to work out a cost per acquisition, it's really finger-in-the-air stuff. How do you go about, you know, if you go to a new client, how do you go about setting a cost per acquisition benchmark?

[00:15:23] Cathal: So, what I do is, and this is an important word, what I do is I do scenarios. So, I do, in an Excel I've got what I consider relative metrics and definite metrics. So a definite metric is, my budget is $1,000. That's what it's going to be. Okay? A relative metric might be, how's this thousand dollars going to turn into users or clicks when another definite metric is, well, the cost per click for this market or this keyword is $2. So, my relative metric is, for $1,000 investment, and a definite or relatively definite $2 cost per user. I can get 500 clicks, you know, or 500 users onto my site, you know. So, what I do is I do all these different forecast scenarios and say, "This is what is likely to happen if my cost per user is two euro or my cost per user is five euro. And this is what's likely to happen if my conversion rate is 1% or if my conversion rate is 0.5%."

[00:16:39] So I paint five or six different scenarios with one or two starting points and just say, "These are the different futures we could have, depending on how the numbers fall." And what we will do is, we will benchmark how close to each scenario we are. And then, after a week or two of testing, because you do have to test in a live environment, is to say, "Well, we are like scenario four." You know, so let's see what happens to scenario four at scale, and then test that, iterate, iterate. But I remember hearing, I don't know who it was, saying that, the only guarantee with a forecast is that you're going to get it wrong. The question is, how wrong is it going to be? So, like it was a financial analyst, it was nothing to do with web analytics or something like that. Like, you know, so, it's literally, I paint different scenarios and, you know, it's just about, well, it's hopefully going to fall within these fields.

[00:17:47] The inputs that are used in a scenario: so you might have a definite media budget, you may not have a conversion rate for your site if you're just launching, but you can literally Google, you know, global benchmarks for your particular vertical, and put that in as your...I generally do that and I say, "This is our ideal scenario based on global benchmarks from E-consultancy, or Google, or the Drum, or whoever for this particular vertical. And then I will work down from that and I'll cut the conversion rate in half and cut that conversion rate in half. So, it's a quarter and I'll get down and I'll have all these different scenarios, and I'll say, "This is our bullseye scenario, our global benchmark conversion rate, but we're an unknown brand." So, the chances, I say, we have to work towards this over the next couple of years, and these are different scenarios that we might get after launch. So, I think it's not a perfect science, forecasting, because you're trying to see into the future, but it's better than promising results because you can't guarantee anything will succeed or, you know, you can't plan for like phenomenal success either. Like it's just, you put your best foot forward based on any input data that you have like pass campaign, global benchmarks, different things like that. But ultimately, I just paint scenarios.

[00:19:14] Will: So you start with industry benchmarks and you kind of work down to, you know, essentially less successful forecasts with the benchmark being a kind of, something to aim for in the long term.

[00:19:27] Cathal: It's a target, it's a KPI.

[00:19:28] Will: And you let the forecasts do the kind of, you know, you let the forecasts present the likely scenarios and you say, "Well, we're going to find out which one it actually turns out to be like and then work on bringing all those good metrics up and the costs per click and acquisition down over time”. And that's then the job of whoever's doing the performance marketing.

[00:19:52] Cathal: Exactly. And you get a double output from it because, when you do your forecasts, you get the different scenarios and then the metrics in those forecasts become your KPIs and targets. So, you can benchmark, "Was I above or below my conversion rate metric? Or was I my forecast?" So, you can start kind of painting, what would you say? Like structured pictures of your historical performance from what, for all intents and purposes, isn't too far off sticking your finger in the air and just guessing a bunch of numbers. But at least you're grounded in...

[00:20:32] Will: It's an educated guess.

[00:20:33] Cathal: It's an educated guess. To begin, you do need some, like in all of the financial institution ads that say historical performance is not a benchmark for future success, but it's pretty close. If anything, you know.

[00:20:46] Will: So, I mean, what are the most valuable tools in your box when you work in analytics?

[00:20:54] Cathal: So, the most valuable tool I think to any analyst is Excel or spreadsheets. Now, I know that there's a lot of other kinds of solutions, you know, dashboards that you can create and direct APIs, and integrations with websites. But data crunching in a spreadsheet allows you to take data from two separate sources and combine them under the same kind of rule and methodology. Because if you're applying an Excel formula that divides one number by another number, that's what it is. But if, you know, and you've defined what that is, if that's what a conversion rate is, it's one number divided by another number and that your conversion rate. It could differ if you're using conversion rate in Google analytics is your transactions divided by your sessions. But, you know, maybe sessions isn't the metric you want to use. Maybe users is, which would be my preference.

[00:21:58] And likewise, maybe if your E-commerce site, your Shopify or your Magento or your bespoke site says that your conversion rate is the number of unique cookies on a browser or the number of times that the shopping cart opens divided by, you know, into the number of transactions. Maybe that's how they work a conversion rate. So, if you use native metrics to try and tell a story, it can just distort the image you need, what I call a single source of truth. And that's unified data in a spreadsheet or in some kind of table or whatever that you create and you define rather than taking predefined metrics from the platforms. I like to make up my own metrics.

[00:22:49] Will: That's very interesting. So, I mean, you know, I'd go further. I'd say that, you know, anybody in business or marketing should have a decent grasp of Excel or numbers, or Google sheets, or whatever your preferred spreadsheet software is. Because, like you say, you can, you know, achieve so much that you can't do with a native platform.

[00:23:14] Cathal: Well, that's the thing. And it's like, even though you can use a dashboard or any of the native platforms, your understanding of how the metrics are created, what they mean is much deeper and richer if you know how to create that formula yourself. And one of the knock-on effects I've discovered as an analyst, it's something I really want to talk about because it's something that no one really mentions to most, but I'm trying to champion it at the minute, and that is, make up your own metrics. And I'm not saying talk BS, I'm saying, if you want to know what a particular metric is that isn't in the standard suite of metrics in your dashboard, make it up. Like, you noticed in my example that the forecasting I talked about cost per user rather than cost per click. And there's a reason for that. Because users is, I prefer the user metric than a click metric.

[00:24:14] So, a user in any analytics tool is generally a unique browser cookie. So, it's, in all intents and purposes a person, it's a person who is using a browser to access your site. The technical reason is it's a unique cookie on a browser. But let's just like get philosophical about this and say, well, look, that is actually the closest metric we have to person. We don't have a metric in Google Analytics and in Adobe and anything like that called people. We sell to people and people visit our website. So, we need a people metric. The best defacto people metric we have is users. So...

[00:25:00] Will: Because the way people's buying journeys are very messy and they can take place over a number of visits, over sometimes long period of time.

[00:25:09] Cathal: Lots of sessions, it's like one user, one person can have lots of sessions. And if you want to know how much...if you think about it, work backwards from, say your E-commerce transactions. You want to say, what is the conversion rate for a single human on your site? So, the natural conversion rate in Google Analytics is the number of transactions divided by the number of sessions. So, multiple visits by a single user. So, it's blurry because sessions don't buy, people do.

[00:25:46] Will: I can see that on a tee-shirt.

[00:25:49] Cathal: Yes. Well, and the followup is when we get to media. So, rather than using the default conversion rate in Google Analytics, which is transactions divided by sessions, I create my own one, which is user conversion rate, which is a transactions divided by users. Okay, so you just do that in Excel and find out what is that conversion rate. So now you know what the conversion rate of a single user is on your site. So, then you figure out, well, what does it cost to get a single user on my site?

[00:26:20] So, the metric you're given in Facebook, and LinkedIn, and Google and all that stuff is cost per click. Okay? A user will click, a person may click multiple times and have multiple sessions. So you've got this compound effect of blurry metrics that are layering on top of each other and you're meant to use that conversion rate and that cost metric. So, for me, it's not clear cut. So I work out what's my user conversion rate and what's my cost per user, and that gives me a much clearer journey from, what does it cost to get a human on the site and how likely is that human to buy.

[00:27:00] Will: I like that. That's good. So what's your take on lifetime value, LTV?

[00:27:06] Cathal: Yeah, it's great in certain industries, like, say, gambling. You might find it in airlines, you might find it in, anything where there's a significant repeat purchase or software downloads, or subscriptions, different things like that. It is transactional. If you're a bricks and mortar, it can be difficult to gauge your lifetime value if you're just a brochure sites. So, I think it's interesting, but, like everything, things change fast and people aren't as loyal anymore. So, it's important to understand, if you are using LVT as a metric in, say, your retention strategy, or even in your acquisition strategy. If you're saying, "I can get someone into my gambling site for $50 and they're going to spend $500 over the next six months before they defect, it is a 50 Euro cost per acquisition." Okay. So, it's industry-specific and it's repeat-purchase specific. I think it's useful, but it's not yet developed enough to be as meaningful to as many industries as it is to the ones that's currently active in, shall we say.

[00:28:26] Will: Right. So, it's more specific to subscription or...

[00:28:29] Cathal: Yeah. Like a bank maybe or a subscription or a gambling...

[00:28:31] Will: ...service-based, gambling, things like that. Okay, well, how reliable are the metrics that we're getting out of these ad platforms and analytics platforms?

[00:28:40] Cathal: I think, within the world of the platform themselves, they're perfectly reliable, but it's when you start trying to compare different data sets across each other. So, if you compare Google metrics to Facebook metrics or, you know, email metrics to Instagram metrics, you start seeing that they just don't line up, you know, the platform sometimes record things in different ways. So, and I'll give you a very simple example. So, for a long time, Facebook recorded clicks as anyone who clicked on a post, liked a post, shared a post, commented on a post, or clicked through to your website, but the common understanding for people when you see clicks is, click through to your website. And it wasn't that Facebook were being dishonest, they counted clicks as anything within the Facebook sphere, which is, you know, perfectly acceptable. Now, later they've introduced a metric called link clicks, which is obviously linked through to your website.

[00:29:44] And that is because there was a certain amount of confusion over it. The other thing is things like conversions. You may find that a conversion on Facebook is what's called a post impression conversion, which means that if you see the ad in your feed and you either click it or don't click it, but generally you don't click the ad, but you've seen the ad, just like you saw an ad in a newspaper or you heard an ad on the radio and you couldn't click the radio ad or the newspaper, you went to another channel and bought it. The radio will say, "Oh, we're after getting a serious uplift in your sales." So, that's the model that the post-impression model is. They're saying, we showed you something which had an effect on your sales down the line, but they record it like a sale.

[00:30:29] Will: Well, this is about attribution, isn't it? So, on Facebook, the default I think is seven-day post impressions and one-day post-click by default, right? So, yeah, like you say, if six days after seeing an ad on Facebook, I just Google something and then end up buying it. Facebook can see that because their code's on the commerce website and they claim the sale and will report that back. What kind of attribution settings should we be looking for? Is there a good way to track our campaigns do you think?

[00:31:03] Cathal: Well, that's the thing. I mean, like, for the longest time it's always been the last-click attribution model, which is the last channel before the conversion gets the sale. Then Facebook introduced… Facebook looks terrible on the last click attribution model. So, they had to by default apply a post-impression model. Or people will say, this Facebook stuff is doing nothing, and they wouldn't invest in it. You know, when it clearly is doing something, we are less clear about what it's doing, but it is, if you measure against baselines and benchmarks, you can see that if your standard conversion rate is 2% and you get a thousand brand searches a month and then you do a Facebook campaign or social campaigns really successful and suddenly your conversion rate just goes up to 2.5% and suddenly you're getting 2000 brand searches a month.

[00:31:55] They're good indicators that Facebook has shifted the dial ever so slightly. And because brand search isn't going to create increase unless there's an outside impetus to say, hey, look for this like a TV ad or some kind of outbound communications. So, attribution is still a bit of a minefield because this is, again, it's a philosophical moment that, because you can record stuff on the internet and on internet advertising, people think you can record everything and you know everything that's going on, when simply there is an internal human process that's involved that can't be measured with clicks. It can be kind of trended maybe, but it's very much an internal experience. And unless you're beside the person and you literally say, "What are you thinking or feeling right now?" You don't know everything, but there's an expectation that because there's numbers, you know everything. And attribution is still suffering from that.

[00:32:57] Google will claim last-click model does everything, but again, it's just happened to...clapping the ball in over the line at the end. Facebook will use its post-click attribution or its post-impression attribution model, programmatic, different platforms like that. Display platforms like that might also use view through impressions or post impression, post-click. And this is where our problems begin as an analyst. They're all using these different attribution models to report success. You know. And they're all valid in their own right, in their own sphere. But there's still no harm to apply a standard model to all of them. And the reason is not to say that, like if you take last click, say we apply the last-click attribution model to everything. All you're doing is saying that, for every thousand euro I spend on Facebook, I get two sales, and for every thousand euro I spend on Google, I get 50 sales.

[00:34:05] And what, you know, so what you might say is, if I spend 10,000 euro on Facebook, I will get 20 sales, and I'm happy with that, because that's all that Facebook can do. And you're able to use a unified model between what cost investment is in Google versus cost investment is in Facebook, but the overlay is the understanding of what the channel can do. Facebook can't drive direct sales, it just generally doesn't do it unless it's something under $20 and it's an impulse buy. It generally doesn't do it for flights, for different things like that, considered purchases. I would apply a standard model to all of the different channels and just see how they work under one model, and then I might change the model and just see how they work under another model. But I'm applying the exact same model to them all, because if I just download a CSV from Facebook and a CSV from Google, my numbers are all off. Facebook metrics are fine within the world of Facebook. It's when we start comparing them to different data sets that we just need a unified model. So...

[00:35:14] Will: Yes, that's true. But I think we also, it's right that we also have to, as marketers, have to recognize the multichannel experience of someone who goes from being...goes from not knowing about us, to knowing about us, to thinking about our products, to kind of wanting our products, and then to actually purchasing them. And I can see why Facebook have that, you know, one-week window post impression attribution model because, much as they want to make their numbers look as good as possible, it actually does make sense. And I have personally seen, I've seen that happen, exactly what you just said. Social ads driving a clear uplift in branded searches. And there is a definite common behavior where people see a social ad. They don't act on it, they think about it, they sit on it for a bit and they suddenly decide they're going to go about because they're at home, it's Sunday, whatever, there's some reasons, some personal reason why they've got time to go through with that purchase and give it the headspace to think about it. And they Google it and that's the brand searches come from.

[00:36:24] So, I guess we have to recognize that, yeah, humans, as you said, we're messy, we can't be tracked in a really kind of neat, predictable, report-friendly way. We all make our decisions in very kind of messy, unpredictable. We never stick to the path, we kind of wander through the undergrowth and the bushes and God knows where, and yeah, we might end up at the destination of purchase, but it's very hard. And that's the goal of an analyst, isn't it? To try and build a picture of what people do.

[00:36:59] Cathal: What might be happening.

[00:36:59] Will: Yeah. What's going through people's heads, what they're actually doing on that kind of messy, windy path to purchase.

[00:37:06] Cathal: Yeah. So, there's some words that I have in my common repertoire and I've always talked about scenario building and stuff, but also, it's inferring. You infer what people are happening, which is meaning, you're using data as your kind of grounding but there is a gap. Now, generally, what they say is, assumptions are gaps in data, but it's beyond an assumption. An assumption would be kind of, it can have bias in there, it can have all this stuff. When you're inferring stuff, you're meant to remain clean, and back to that word that you used, you're wandering.

[00:37:43] Will: But, I mean, that's why more behavioral tools have become popular in recent years, isn't it? So, things like HotJar where, you know, they're trying to offer as an insight beyond click-through rates and show us things like scroll depth, show us mouse cursor behavior...

[00:38:03] Cathal: Mouse movements, hotspots where people click.

[00:38:04] Will: And really, so to try and get inside the heads of our modal or average customer and how they use the site, identify friction points, identify those hotspots of success quite literally with the heat maps. How much store do you put behind that kind of analysis?

[00:38:24] Cathal: It just helps me and I think numbers, as we said at the start, like, your job is to turn those very, very, very explicit, you know, numbers into something that's a narrative or a story. But when you see this older layer of data, like, you know, to what level people scroll down your page, where they click, it allows you to just build another layer of insight onto what those numbers might actually mean for the average human. You know, again, the starting point is always, what is a person doing? Because a click isn't a person, it's a click. It's why, you know, a click. So, you have to think like, what made them do it, you know, what did they do beforehand? That might be your social conversation, all of that stuff. So, it's always to keep that human in mind. And those nice tools like Hotjar really bring that human factor in, where you can see that like, at the top of the page, the color's all red because that's where people look. And at the bottom of the page, it's blue because only 10% of people look there. And it's very, very, very intuitive. And I find it a really great tool. It really helps with just adding, I know it is obviouslý a heat map, but adding a bit of color onto those kind of basic numbers.

[00:39:45] Will: It brings out my voyeuristic side slightly because you can watch replays of people using the site, and moving their mouse and scrolling and clicking from page to page. And it's when you do that, that's the real thing for me with a tool like Hotjar, is when you watch someone using your site in that playback mode, you kind of go, "God, I wouldn't have used the website like that." And actually, sometimes it looks like they almost don't know what they're doing or what website they're on. They're kind of moving about it in a way that it just seems so unintuitive and not the way that you would have guessed. And I think that as anyone who's even an armchair analyst, you know, not someone who's got analyst in their title like you, but all of us who are essentially analysts, it's our job to allow our assumptions to be challenged, isn't it? And to remain open-minded and understand that people aren't us, you know, they're not like us, and people are different and unpredictable. They're an unwieldy bunch.

[00:40:44] Cathal: I know, and like digital marketers are a terrible sample because we're cynical and we don't click on the paid search ads and, you know, we discount everything and we use websites much more efficiently because we're around websites all the time. But the standard person has a different set of priorities and a different day. But it is kind of people watching what number's. It is definitely a, it's an interesting game to get into. And one of the things that I want to emphasize is, I trained in my undergrad as a photographer and then I moved on to being a copywriter. Right now it's for TV and radio in traditional advertising agencies. And then I moved on to search marketing with Yahoo and Overture and into the agency and all that stuff. And now I just crunch numbers. So, moving from visual art to Excel spreadsheets was kind of natural in some bizarre turnaround way.

[00:41:41] It's not just a gang for mathematicians, and for Excel nerds and stuff. It's actually, I love the creativity behind it, and I love that you can challenge yourself to find something new or a solution to something new each and every day. And like I still make up new formulas for new things that I'm trying to predict or, you know, I just want to find out something, and it doesn't exist, so I have to make it up. And it's empowering to know that you can do that, and you should do that, and not just take what you're given.

[00:42:14] Will: I agree. No, I think it's a very creative discipline. Unlike you, I'm a creative at heart, but I think, to me, analytics and the analysis of behavior in marketing, what you're actually trying to do is almost crack the formula of human behavior, human desire, and the way that humans respond to your creative work or your agency or your business' creative work, you know, how do people respond to that copy line combined with that image, with the cheeky emoji at the end of the line? And that call to action versus that one? And the red header on the website versus the blue one on the form on the left and the right? And all those kinds of things. You're almost trying to crack the code and step inside the matrix of humans. And I think that's what makes it so fascinating.

[00:43:02] Cathal: It's great. I mean, it is like, you are Socrates. You're like, you know, Aristotle, you're just philosophizing with data as your grounding force to figure out, well, what and why. And it's always those questions. It's just why, why, why? And turning that into human, a person, not click, not session, not conversion rate, not anything. This is, I suppose my other philosophy on this is, if you don't have humans on your site, so if you don't have users on your site, you don't have transactions, you don't have cost per click, you don't have number sessions, you don't have average time on site, you don't have pages per session. All of those other metrics are diagnostic metrics for how to understand human. I have a person on a site, how long did they spend? How many pages did they go to? If I have no people on the site, all of those other metrics don't exist. The starting point is user. So the starting point is human. And you use these other metrics to diagnose the human condition.

[00:44:12] Will: Indeed. Wow. Well, if you don't want to get into analytics after hearing that, what do you want to do with your life?

[00:44:19] Cathal: I'm very philosophical now, Will.

[00:44:22] Will: But it's true. I think that's what we're doing. It's great. It's fascinating. So, just to kind of really ground what we're talking about. What I'd love to know is, in really concrete, practical, explicit terms, if we've got a cost per acquisition, a certain amount, what are the kind of steps, practically, that you take to try and drive that down?

[00:44:49] Cathal: Well, I mean, back to clusters, I focus in on the channels and the media splits, if I'm talking about media investment, that replicates my ideal scenario, or gets close to my ideal scenario, and then I just, I test, I repeat, I change maybe one variable for a week or two weeks or whatever it is. But like the idea of testing is absolutely important, from website to your Hotjar inputs, to forms, to using things, free tools like Google optimize, which is a free AB testing tool that naturally integrates with Google Analytics and Google tag manager and will allow you to AB test different versions of your website. And if you use something like Chrome, you can get a plugin that will allow you to just change elements of your site and Google optimize will serve that. So you don't need a big, expensive AB testing tool. You can just say, I've got the Chrome plugin and I'm going to change the color of this button to red and the color of this button to blue and Google optimize will manage everything for you.

[00:45:59] So, to bring cost per acquisition down, I focus in on what does good and do more of that stuff. And I reduce my effort in trying to convert difficult customers, or difficult regions, or like use channels that are vanity channels or different things like that. It's all about prioritization and focus. There's no exact exact formula. But what I do use to kind of drive my strategies would be, obviously, you know, seasonality is one way of looking at it, but sometimes can be a bit of a card, is things like anomalies in the data. Where do things get different and what caused that note of difference?

[00:46:46] Will: Yes. You're looking for the exceptions, anomalies, aren't you?

[00:46:49] Cathal: What are the anomalies and what do you think caused them based on what you know happened around that time and what might, you know, have happened within the wider market, your experience, all that stuff. But ultimately, something changed. What changed it? And, you know, there's a good analogy, and this is like, anomalies aren't just things going bad. It's, why did my conversion rate go up? You know, and how can I make it go up even further?

[00:47:14] Will: Yeah. Why was Sunday afternoon this big spike?

[00:47:16] Cathal: Exactly.

[00:47:17] Will: Or, why do people on Android phones seem really predisposed to buy this certain product?

[00:47:23] Cathal: And we do a terrible thing. We do a terrible thing that when things go well, we don't investigate why. And that's because we're just happy to take the good, you know. And an old friend of mine said to me, if you don't know why you're going up, you won't know why you're going down. So, it's important not only to know, so when things take an upswing, to understand why and what are the potential reasons that they took the upswing, can you get more out of those? Can you push more into those reasons? And if you start going down, have any of the reasons why you went up changed? So, that gives you an indicator for how you can save a situation or at least minimize the impact of a change in direction in a negative field. You know, so, that's important. You know, I think.

[00:48:17] Will: So, to drive down something like that all important metric of cost per acquisition, it sounds to me like there's a few different levers that you can test different pages on your website, different versions of those pages, different creative messaging and ads, different products, different targets into different users, geographically and demographically. What are the main levers to you?

[00:48:45] Cathal: Yeah, I mean, I think we have to figure out what channels work, what times work, days, different things like that. So, because it's a human experience, we have to operate in the world of time. And time means, when I'm in a position to take action, that's a valuable action on the site. Whether it's a time of the day, or a day of the week, or a part of the month where it's close to payday or something like that. So, time is a major lever to understand, when can I and when should I increase visibility? Or add a discount code, or change my messaging, or do something to get in front of potential customers at this key time to ensure that I enter what, you know, the old consideration set. And, you know, in that world of marketing where when they're ready to take action as a personal experience that I'm front of mind.

[00:49:44] So, the first lever is time. You know, what is the time to do it? And then the next lever will be, well, what channel and channel mix to do? So, you know, what is your converting channel? How does that help? And what do you need to do from a supporting channel perspective? That's your social media, your email, your organic search, things like that. And influencers and maybe any offline, you know, that, you know, helps those converting channels do the job of getting the valuable action over the line. So, time first of all, then channel. And then different experiences on the site. Once we get them to the site. So, what's the layout? What's the journey? What are the friction points we can remove? And when you talk about testing, don't te...Like, if you're doing a 50/50 test, this is a guarantee, an absolute guarantee, that one will be better than the other, and that is a fact. And this is why testing is important. Again, it's a philosophical moment that if you put all of your faith in an assumption or a hypothesis, there is a 50% chance that you've chosen the wrong option, because, with an AB test, one of them is going to be better, it can be better by loads or better by just a small amount, but one of them is going to be better than the other. So if you put all of your faith in an assumption without testing, you are by default adding risk of 50% that you've chosen the wrong thing. You know, so, testing doesn't mean success, but it does reduce your risk by half, because, if you have a 50/50 test, at least half of your traffic is going to the better option. Whereas if it's just one option, like you put all your faith in one thing and you get it wrong, you have automatically reduced, you know, the amount that you could have done with that by 50% on top of the poor performance on this side. So, it's about managing risk as well. So, what I think about testing is, people say test that, but test that doesn't mean you're going to get everything right. Your test might be wrong. And the answer or the the learning from a wrong test is, dump it out.

[00:51:58] Will: Well, that's, I mean, that's what I tell a lot of my clients and delegates, you know, testing is not just about finding success, it's about learning. It's about improving your marketing instinct, about improving your instinct for what works. Because, over time, I do think that has a positive effect on you as a marketer creatively and analytically. You know, when I run social ads for clients, I run massive multi-variant campaigns. You know, there was a campaign last year I did for a client. They spent about £120,000. I ran well over 4,000 different ad variants. Now, I didn't make them all myself, you use a tool to do that. But, the amount that I learned about that industry, about the way that people respond to messages about that kind of product, time, all those factors, those pivots that you talked about, those variables, time, location, demographics, creative, product, and you missed this massive opportunity I think to learn and to become a keener, better marketer.

[00:53:08] Cathal: Yeah, I agree.

[00:53:08] Will: So, you know, I think, and that's part of being the analyst. I think that we all need to kind of become, as marketers, it's not just about digging into the data and being able to understand and interpret that. It's about being able to test and grow a bank of learnings within your own head and improve your instincts.

[00:53:27] Cathal: Yeah. What can we do here that could be interesting in terms of other, an upswing in conversion or learning output. You know, that's generally the way I'd approach things. But, it's interesting that you said like, so much that you learned about gauging the response, how people reacted. You didn't say, I was really interested in the amount of clicks that this ad drove or, you know, you were talking about human responses to things and that's, for me, I mean, that's everything, and that's the starting point, and it's just, like, there's no metric in any of the tools called human, or person, but that's what we're dealing in here. So, it's always important to remember that that is your ultimate goal is to unlock that secret.

[00:54:10] Will: It is. That's our job, is to get humans to respond to in a positive way.

[00:54:13] Cathal: Yeah, humans to do stuff in a positive way, exactly.

[00:54:15] Will: Yeah, it's what we're doing. Well, that's absolutely fascinating, Cathal. Thanks so much. Thanks for all your insight and experience in this space. It's fascinating to hear just what your mindset is when it comes to analytics. So, thanks so much for coming on the podcast today.

[00:54:33] Cathal: Cheers, Will, it's been great. Yeah.

[00:54:35] Will: It really is a pleasure. Thanks a lot.

[00:54:37] Will: If you enjoyed this episode, subscribe wherever you get your podcasts. And for more information about developing your own soft skills in marketing, head to digitalmarketinginstitute.com. Thanks for listening.

Related Articles


Cathal Melinn
Cathal Melinn

Cathal Melinn is a well-known Digital Marketing Director, commercial analyst, and eommerce specialist with over 15 years’ experience.

Cathal is a respected international conference speaker, course lecturer, and digital trainer. He specializes in driving complete understanding from students across a number of digital marketing disciplines including: paid and organic search (PPC and SEO), analytics, strategy and planning, social media, reporting, and optimization. Cathal works with digital professionals in over 80 countries and teaches at all levels of experience from beginner to advanced.

Alongside his training and course work, Cathal runs his own digital marketing agency and is considered an analytics and revenue-generating guru - at enterprise level. He has extensive local and international experience working with top B2B and B2C brands across multiple industries.

Over his career, Cathal has worked client-side too, with digital marketing agencies and media owners, for brands including HSBC, Amazon, Apple, Red Bull, Dell, Vodafone, Compare the Market, Aer Lingus, and Expedia.

He can be reached on LinkedIn here.

Upgrade to Power Membership to continue your access to thousands of articles, toolkits, podcasts, lessons and much much more.
Become a Power Member

CPD points available

This content is eligible for CPD points. Please sign in if you wish to track this in your account.