Boosting ecommerce sales using marketing mix modeling (MMM) with PlanetArt

Jessica Gondolfo:

Hey everyone. Welcome back to the Marketing Intelligence Show. This episode is with Ryan Hubbard, senior business Intelligence analyst of Planet Art, who is taking us through how they're using marketing mix modeling strategies to increase customer acquisition and market share. I'm super excited to have you on the show today. How are you doing?

Ryan Hubbard:

I'm doing great. Thanks for having me on.

Read full case study to discover how PlanetArt reduced reporting by 20 hours a week and increased ROI with Supermetrics.

Jessica Gondolfo:

Yeah, absolutely. Before we jump in, why don't you tell everyone who is Planet Art and what do you guys do there?

Ryan Hubbard:

Sure. So Planet Art is an e-commerce business, and I'm sure all your listeners are saying I've never heard of you guys before. That's true because it's not our customer facing brand. So we're actually a house of brands. So we have a bunch of different e-commerce websites and mobile applications. So for example, we have cafe press.com where artists can actually upload their own content and sell on our sites for a commission. We own gifts.com for personalized gifting. If you have a Mother's Day around the corner this weekend, a lot of mothers are getting the best mom ever. Coffee mugs. I've seen them and they're arriving today. And then we have mobile application, e-commerce businesses, like free prints and photo books where you can actually print your custom photos on a photo print or on a photo book and we'll print it and ship it for you. And those are our customer facing brands we like to say, we like to help our customers preserve life's important moments and keep their families close with our products.

Jessica Gondolfo:

That's amazing. And senior business analyst, what are you, that's a lot of brands that you're probably overseeing, so what does your day look like there?

Ryan Hubbard:

Sure. So I get to wear a lot of different hats. So I'm working with teams across the business. I'm mainly focused with the marketing teams. I'm looking at customer acquisition strategies, retargeting strategies, collecting data, building visualization reports, and trying to find any sort of insights to help the business improve the customer acquisition and retention strategies. But I'll also work with our product development team, looking at the user interaction data for our websites and apps, seeing where our customers are getting frustrated, why aren't they coming back, why are they dropping off? And then I'll work with some operations and customer service departments to see if there's any insights or unique findings that we can improve in their departments as well. So it's never a boring job. I get to change tasks every single day and it's never repetitive. So it's very interesting.

Jessica Gondolfo:

I love hearing how someone from the business analyst perspective is actually so involved in so many different departments. It just shows how data can be so versatile, how you use it and where you're putting it for best business, the best business outcome. So that's absolutely fantastic. So my next question for you is definitely around how you manage your marketing strategies, marketing campaigns, how they differ and really what you're doing to manage those.

Ryan Hubbard:

Sure. So because we're marketing so many different brands and so many different platforms, so we have paid search ads for Google Ads, bings, we have paid social on social media sites like Facebook, TikTok, one of the importance is getting all of that data in an aggregated format in a centralized repository so that we can monitor them. Because when you have 60 different ad accounts, it's very cumbersome to go into Google ads and monitor these in that platform alone. And it also doesn't give you a good view as the business grows to multiple channels, what you're doing across the entire business because Google Ads is one format, paid social is one format, and you want to see it at a very high level how your spend and marketing strategies are going through those. So I focus on collecting that data and then analyzing it at a brand level because our ad strategies will actually differ for each brand, especially websites versus mobile apps. And so we tend to do a lot of product focus on paid search channels. And then retargeting we tend to focus on our internal marketing files like email and SMS, and those are just some of the high level strategies we deploy.

Jessica Gondolfo:

And my guess, tell me if I'm wrong, is this is how MMM or marketing mix modeling has really come into your business and why you guys are using it?

Ryan Hubbard:

Yeah, exactly. So the media mix modeling is a very important part of our business, which when we have offline spend, for example during the Christmas holidays, we'll send out Christmas catalogs for customers shopping or even run podcast ads. So I'm sure listeners are going to see some sort of ads to whatever platforms they're running for this podcast. And at the end of that ad there's an analyst like myself wondering what is the impact of that podcast doing on the business? And one of the issue is we tended to use last click attribution modeling to determine where our ad spins were more effective. But that's not really effective when you're doing offline advertising on catalogs or podcasts because there is no last click attribution. In fact, there's actually some cannibalization of the attribution by paid search because what most people do is they go on their phone and Google the business clicking on a paid search link and then placing an order.

So you may look at your advertising and spend and saying, oh great, Google search ads are doing great, but really that attribution should have belonged to a direct mail catalog or a podcast listen, but you don't get that because you're looking at a click attribution. So what we do is we aggregate all that offline and digital marketing spend in one location and we compare it to our internet traffic as well as our purchasers and revenues. And the model can actually pick up when we see those catalog drops or when we see those podcasts impressions and say, you know what? Paid search may have gotten the last click attribution credit for that order, but most of the order and revenue that impacted the business should have been tied back to those ad spends on the podcast or direct mail catalogs. And that's how we sort of make sure we maximize our return on investment by making sure that those channels get credits for the orders that they deserve.

Jessica Gondolfo:

This might be more of a technical question, but how does your model determine that weight of all the different places it touches? How are you saying, you know what, this catalog really should have 40% and the ad should have 10% and multiple emails between?

Ryan Hubbard:

Sure. So for direct mail catalog, we actually have estimated days of when they will arrive in the person's households, and that's when we see the greatest impact because nobody's, very few people are sitting on these catalogs for weeks before they purchase. They're going to purchase very quickly so we know when they're going to arrive in the household and then we can feed into the model the spend data on the days that they're arriving in-House because then we know that day deserves the spend and the model's going to pick up those spends on those particular days, and then it's going to give credit at a higher rate to direct mail even if it detects a paid search increase.

Jessica Gondolfo:

That's super interesting. Is this completely different across all three brands that you have or are these very similar models?

Ryan Hubbard:

It's very different across all brands because we don't do much direct mail for our mobile app side. A lot of the direct mailing is on the personal creations simply to impress sides of our business. And so that's why we have to make sure we have different model parameters depending on what brand we look at because the advertising strategy is different and if you use the same model, you're going to get very different results.

Jessica Gondolfo:

Do you have any organic stuff that's coming into these models as well? Do you guys focus on organic strategies?

Ryan Hubbard:

Yeah, we do have organic. We do have our search data coming from Google that we feed into models. It tends to be smaller because a lot of our traffic is coming from paid search. And even then when they do come through organic search, a lot of that is branded where people are searching up our brands. But the organic search tends to be more of a retargeting system because we need to get customers familiar with our brand and products before they start searching us.

Jessica Gondolfo:

So if that's from the acquisition perspective, I guess if getting people familiar enough to take them down a buyer journey to make that purchase, how do you guys use this modeling in-House to fuel your retention strategies and your returning customer base?

Ryan Hubbard:

Sure. So retention, we tend to look at internal channels like email, SMS and push. So SMS is one of the big ones where it's an expensive format to market to customers. SMS, because you have to pay the carrier fees, you have to pay the SMS send cost fees. And so we have to be very selective about who we reach out to on SMS. And as customers stay on your SMS file for long periods of time, they are less likely to click on an SMS advertising link. So we may have to look at that span and saying it's not as impactful on the business anymore. So we will actually reduce the number of targeting send SMSs we send out so that way we can reduce our costs without losing a lot of the impact on the business. So that's one of the areas. So if we see that we're spending a lot on SMS, but its impact on the business is dropping, we will actually look at our targeting parameters for our campaigns and tighten them up just so we can reduce our cost.

Jessica Gondolfo:

That's interesting. You guys have been using Supermetrics for a while actually.

Ryan Hubbard:

Yeah, I think it was here when I started over five years ago. I think since I've been on board, I've definitely expanded our uses Supermetrics because we had it in a very limited format. And for me, I want as much raw data as possible and Supermetrics has been able to increase the amount of data I come back to.

Jessica Gondolfo:

Yeah, I wanted to understand how are you guys using supermetrics today and how does it help you with some of these bills, some of these data models that you guys are building internally?

Ryan Hubbard:

Yeah, so our implementation is mostly around our paid marketing advertisers. So we have connections to Facebook, Google, Bing, TikTok, and a couple other platforms. And we're bringing in either every couple hours or every day all of our impression data, our click data, our spin data, which is very important to us, and we're bringing that into an S3 bucket where we can then query in our internal database in Redshift and join it with our internal order data or our internal user interaction data and we can implement monitoring tools and quick alert tools to see if any errors pop up. And so we mainly use it as a data pipeline tool where we just centralize it all in one location.

Jessica Gondolfo:

That's great. And what is the business impact of having tool supermetrics look like for you guys internally?

Ryan Hubbard:

Yeah, I mean obviously there's the labor cost because when I first signed up at Planet Art, my first task was to copy and paste these values from Excel sheets going through each ad account, copy paste, copy paste, and then compile that into a centralized reporting. And so as we've grown, that task would take normally 20 to 30 hours a week. So if you're looking at a full-time employee basically to collect this data for reporting, and then it's also a very slow, cumbersome experience. So we get the labor cost right away by pipelining our data through Supermetrics, but that's not where I see most of the value that came from Supermetrics. What it's done for us is it's shortened the timeframe at which we get the data into our internal systems and we can build visualization and we can iterate faster on our marketing spend because of the insights we gained from that data being in our systems and aggregated faster.

Jessica Gondolfo:

What do you think in terms of the revenue side of the house, what kind of impact does that have for you being able to make such on demand decisions?

Ryan Hubbard:

Yeah, it's had some moderate improvements, not at the top line revenue benefit for us, but as far as the profitability, because what it's allowed us to do is look at our advertising spend and saying, okay, these channels that we're spending a lot of money on aren't as efficient as other channels. So we can reduce the spend there and shift it to other channels where it might perform better. And so our ROI has gone up even though we're getting less from our spend.

Jessica Gondolfo:

So it's really like you're optimizing, you're spending less, you're getting more, you have that strong optimization there from using Supermetrics.

Ryan Hubbard:

Exactly. And as we get that data so much faster now, it allows us to iterate and test faster. So you get the benefit of you're getting better performance, but you also get to try more things even though they're going to fail. Most of the time you get to iterate and find the ones that work faster

Jessica Gondolfo:

And you're using our data blending features. What products are you using?

Ryan Hubbard:

So we're data blending our Google ads being ads with our internal order revenue. So that's where we tend to, and that's fed into our media mixed modeling because we have our impressions and spins that come from our advertising and then we can blend it with our internal order and revenue tracking because one of the issues with these ad partners is they tend to over exaggerate the number of orders and revenue that they attribute to their campaigns because they want you to spin as much money as possible in their platforms. So it's very important that you look at your internal systems, you don't want Facebook and Google ads taking credit for the same order, then they're going to convince you to spend more on their platforms when it's just one order you're getting in the business.

Jessica Gondolfo:

Yeah, that's really fair. Since you have all of these models, have you particularly seen specific trends that have been happening? Have you guys been like, Hey, we would definitely want to shift budget or create new campaigns and new channels based on how you are seeing your audiences shift?

Ryan Hubbard:

So what we've seen is that we can be more selective because we're a house of brands about which brands we're going to start paying for. So one of the things we have is personal creations.com, which is a gifting and personalization site for e-commerce, but it's got a very big product catalog. So we can actually see that if we adjust spin to our sites like Parker and pip.com for kids gifts or legacy lane.com, which is our memorial category gifts, we can be selective and push people into a more targeted category site where it might perform better because they don't have to sit a shift through all the different products in a large catalog. And we tend to see that's where we're seeing a lot of performance benefit, where we can actually shift our spin depending on what brand we want to promote at what time.

Jessica Gondolfo:

That's great. And seasonality is a big one for your business. Are you doing a lot around looking at trends year over year for seasonality and pushing that out?

Ryan Hubbard:

Yeah, because we're a gifting site, we often follow the holiday trends. So right now we're in graduation and Mother's day season, but the biggest time of year for most businesses like us is Black Friday Cyber Monday weeks during the fourth quarter in the holiday season. So that's where we sort of ramp up our advertising and we see a lot more impact of our mixed media modeling. And it's actually hopefully going to save us with super metrics because have a little anecdote about how supermetrics may have saved my job. We got an alert one day that our ad spin had dropped to zero for one of our ad accounts and we investigated in one of our saved form of payments had expired. So we were able to quickly resolve this issue before it negatively impacted the business, and we were able to get those ads up back up and running really quickly. But I hate to think that in a very seasonal business like ours, if that were to happen on a Black Friday or Cyber Monday, and nobody caught it because we didn't have this data in our internal tools, those are the types of mistakes that cost businesses millions of dollars. And so I was very glad to see Supermetrics actually help us save our business on that day.

Jessica Gondolfo:

I'm also happy to see Supermetrics to help you save your job there, but that's actually excellent. I do have a lot of people who have referenced before the different alert systems that they have up, whether it's a brand or an agency for these specific reasons, but that's really awesome to hear that you were able to get that kind of insight as quickly as you were.

Ryan Hubbard:

Yeah, and that's where the benefit I see of supermetrics is, is that you get the data quickly and it's clean and aggregated in a central repository.

Jessica Gondolfo:

That's amazing. Ryan, I want to say thank you so much for coming on the podcast today. Loved this conversation. Media mixed modeling. I feel like I might've said marketing and I think is it interchangeable? That interchangeable?

Ryan Hubbard:

I've heard it. It's technically media mixed modeling, but I've heard it go both ways.

Jessica Gondolfo:

You can correct me. It's okay. I am not technical, so you go ahead and tell me that's wrong. I will correct that going forward. But thank you so much. For those of you who are listening, go check out Planet Art or personal creations.com if you need something personalized. They are incredible. They're amazing. And yeah, it's been really great.

Ryan Hubbard:

Great. Thanks for having me on.