Podcast

How to discern between interesting data and valuable data

Gain confidence in your media planning efforts by looking at the data that’s readily available. While the data you have at your fingertips may seem uninteresting at first glance, or even second or third, it can be extremely valuable. Combining different types of data sets can result in impactful results for your media strategy — and bottom line.

At Callahan, we create smart, efficient media plans by combining seemingly unrelated data sets, and you can, too. Consider the sources available, including Census Bureau data, the Bureau of Labor statistics data and Google search trends.

In this episode, we discuss three case studies where overlaying different sets of data revealed new insights, strengthening and informing our clients’ media plans.

Listen here (or subscribe on iTunesStitcherGoogle PlayGoogle PodcastsPocket Castsor your favorite podcast service):

Welcome to Callahan’s Uncovering Aha! podcast. We talk about a range of topics for marketing decision-makers, with a special focus on how to uncover insights in data to drive brand strategy and inspire creativity. Featuring Jan-Eric Anderson and Zack Pike.

Jan-Eric:
Hi. I’m Jan-Eric Anderson, Chief Strategy Officer at Callahan.

Zack:
And I’m Zack Pike, Head of Data at Callahan.

Jan-Eric:
Thanks for joining us today. It’s going to be a fun episode. I’m looking forward to this one. We’re going to talk about one of the more exciting things I think about analytics and some of the projects that we work on. We’re going to talk about how we land on different combinations of data when we’re seeking insight for clients, and the reason why this is a fun topic is because this is part of the job where we get to plug in some creative thinking to think about what would be an interesting way that we might be able to get into some insights.

We’ve talked about this on some past podcasts before where we’re kind of talking about data that’s right in front of you, and how you recognize data that might be useful. Sometimes the data that would be useful is not obvious, and so part of the game is to think of some interesting combinations of data that when brought together, if we can combine these things, these different types of data, would they start to tell a story that maybe we haven’t seen? Do they start to paint a picture?

Zack:
Yeah, it’s thinking outside the box and getting creative. It’s so easy in data to get in your lane and stay in your lane. I look at search data, I look at sales data, but when you take a step back, and think, “Okay, what else could be out there available to me?” It can get pretty cool.

Jan-Eric:
So we can talk and kind of generalize this and talk hypothetically, but it might be good to just jump in a couple of examples and tell some stories of some clients that we’ve worked on. We’ll remove the names of the clients, but we’ll talk in general terms about kind of what that process is. So Zack, pick a client that you could think of where we’ve combined some fun sources of data together to paint a picture. Let’s talk about that.

Zack:
Yeah, so we’re kind of fortunate that we have several clients that have seasonal businesses. There’s certain times of the year where they have increased sales activity versus others, and when you’re planning marketing and media campaigns around that, I mean you have to take that into account, so you’re not fishing when the fish aren’t biting. One of the coolest examples is we have a scenario where we had to help assist this client in planning for a product that was designed to take care of a problem with mosquitoes, and all of you listening, you know mosquitoes have a pretty finite time of the year where they become a problem. Sometimes that fluctuates from week to week, and it’s highly impacted by weather and temperature and location in the country.

So in this scenario, we didn’t have a lot of history to go on. This wasn’t a product that had been sold for a really long time, so you couldn’t just look at all your prior activity and draw your conclusions that way, so this is where we all took a step back and said, “Okay. What could be out there that we could pull in and start to predict when I should market, when I shouldn’t be spending money, and really where I should be in the country from a distribution standpoint, from a targeting standpoint with the media?”

We found a couple of data sets that ended up being very valuable. The first one was weather. So we’ve talked about weather quite a bit on this podcast. We subscribe to the IBM Weather API, which is the same data you get on the Weather Channel app and most of the other weather apps on your phone. It’s the best weather data that’s available. So that was valuable.

The other one that ended up being pretty cool was Google Search data. Google, many of you probably know this, but Google has their tool called Google Trends, and what a lot of people don’t know is you can actually extract the data out of Google Trends. It’s kind of cumbersome to do unless you write a script around it do it in an automated fashion, but in this scenario, you could just pull mosquito search activity for every state. Even a data set as simple as that, and the point of that was to say, “Okay there’s states that have more of a problem than others, and I need to know the mosquito season in every state because if it’s not mosquito season in Kansas, and I’m advertising, that’s a bad thing, right? I’m probably not being as efficient as I could be.”

So the way that data comes out of Google is not total volume. It’s of all the searches happening in an area, what percentage of searches are associated with the terms you’re asking it for? So the way that we looked at it was, “Okay, for the whole year, in every state, when do these mosquito terms become a higher share of all search activity in that area?” And that is a really good proxy to the problem that people need a solution to help solve. So in the middle of the season, search activity’s high. That’s a really good time to market.

But what you find very quickly is that the season is very different in every state. We pulled the data all weekly for every state, so once we laid all that stuff out in a heat map, you now have essentially your media schedule that you can use to plan. Bump that up against weather data to one, validate that search information because these are just search activities. Google is giving you really high level numbers. This stuff is not super finite, so you always want to balance it against actual data, and we know roughly why mosquitoes breed, and when they develop based on weather patterns, so we bumped that up against IBM’s data at the state level to say, “Okay in times when the weather is more rainy, do we see elevated search activity from a mosquito perspective just as a gut check.” The answer was yes.

And of course using that with business data is where it all ties together. We used those three data sources to help plan, to help empower this client with better prediction on when the season was, where it was going to be heavy and where it wasn’t going to be heavy. That was a year ago almost that we did that, so then we still had all that data. Now the question is, “Okay, well you guys thought X was going to happen. Did that actually happen. Were my seasons lined up to when the search activity took place, and then what can you tell me from a weather perspective?”

From my perspective that’s when it gets really exciting because we can, at the point we’re at now, is we’re using the weather data to predict hot and really hot periods for this mosquito activity, and this is not secret, but mosquitoes breed in pools of water. So in times where you have standing water around, you have elevated mosquito activity in the weeks following that. Standing water is developed by precipitation, and so when you have times of precipitation, there’s a certain amount of time after that precipitation happens that you see spikes in sales. That lines up with the search activity, and now you’ve got a really strong planning model one, for forecasting, just the business in general, but think about the applications for media. Think about turning on and off media based on weather patterns. It gets really efficient.

Jan-Eric:
Well, and I think that that’s where it comes full circle. It takes it from being just interesting insight to something you can actually do, so in a nutshell, we saw that search activity correlated to sales activity. People were searching when there was a presence of mosquitoes. When people are searching, we know that sales came up. The thing that’s interesting, it’s a little bit of a no brainer, but when you then layer in the weather piece to find the triggering events around weather that consistently were present just before and during times of the heightened, or spikes in surge, it becomes a lead indicator that there’s going to be heightened levels of interest and it becomes a triggering event for media. In that regard, you’ve taken weather data, search data, and business data combined to develop a media plan, and that is the beauty, it is the perfect example of how the power of combining seemingly unrelated data sets can become very actional.

Zack:
It’s your secret weapon. It’s shaving off that little bit of incremental sales on every little opportunity, right? If I’m in the market with media, that’s a good thing. But if I’m in the market heavy at points when there’s a lot of people with a problem that I solve actively looking for a solution, that gets, you know, that’s really where you’re optimizing things to the maximum extent.

Jan-Eric:
So let’s shift from mosquitoes and talk about a home improvement product.

Zack:
Yeah. Yeah, so this is another cool one. This product is sold in places like Lowe’s and Home Depot. It’s a home improvement that’s used both by individuals like me who are not a professional by any means doing anything around the house, but I figure it out, and also professionals. Trade workers use this product a lot as well in their jobs, and the challenge with this one is that this product had been doing really well from a sales perspective, growing year over year, everything is really positive. But, like with anything, we want to do better, and we want to help the retailers understand this product’s going to keep growing. The velocity is going to increase over time so that they feel better about keeping it on shelves and really giving it better shelf space in the future.

But the challenge with that is, “Okay, I know it’s not going to keep growing like this everywhere. Where are my areas that it’s growing, and why do we think that’s happening?” That was really the question, because once I understand why, I can go find other areas that match that criteria to try and figure out why they’re not growing at that rate, but also I can bump advertising into these areas where I think I have additional opportunity.

Jan-Eric:
Well, that’s probably a good time to interject also that with this particular client, this is a client that has a limited media budget.

Zack:
Right, absolutely.

Jan-Eric:
And needs to be really really smart. So this is not a client, while this is a product available nationally, it is not a brand that has a media budget that allows it to be supported nationally, so part of the challenge of this was, “Hey, I want to build a really smart media plan. I need to demonstrate impact on sales, so can you help me figure out where I should be advertising, and assume that I’m going to be in digital channels where I can geotarget to specific places?”

So that was part of the context of this as well. Where geographically do I get the greatest opportunity for the biggest return-

Zack:
Absolutely.

Jan-Eric:
And how can analytics inform that?

Zack:
Yeah, never have enough money. We never have as much money as we want to spend. So in this case, there is a lot of data that’s available around this type of stuff. I mean, this is where you’re looking at the types of customers and the trade areas of these stores. We looked at a lot of stuff, but the two that ended up being valuable were building permits that were issued in different areas, so in areas where there’s a lot of building permits, you have elevated levels of tool purchases. There’s more workers in that area. There’s more activity going on. There’s more tools breaking, and more tools needed, so it’s reasonable to think in those areas you would see that activity.

The other thing I was actually surprised was available at the level that we used it that ended up being cool is trade worker populations. So you can look at least over the last five years at growth and decline in things like plumbers in different areas, electricians, finish carpenters, and then you start thinking, “Okay, well I can put these together in areas where there’s a lot of building permits, or where building permits are growing over time, and then areas where … let’s just say electricians, there’s a high electrician population, there’s probably a lot of activity happening there. I should make sure my stores are performing well in those areas, and if they’re not, that’s a really good testing ground. I know the demand is there. It’s gotta be. I just don’t have a high enough share of that pie.”

That’s where the sales data comes in, so then we bump that into the sales data. Now I understand my areas where I have opportunity and where I’m probably maximizing what I’ve already got, and that’s your media plan.

Jan-Eric:
So the overlays here are the presence of trade workers, which is an indication of people likely to be using your product. There’s a population who would have a need for your product. Overlaying that with building permits is an indication that there’s a lot of jobs happening for those workers, and then overlaying that with sales information to validate the hypothesis where in areas where we see high concentration of trade workers and a high volume of building permits, do we also see sales being high?

Zack:
Right.

Jan-Eric:
And that analysis proved out, which then indicated a, do you find these hot spots and this is where you target your media.

Zack:
Yep, yep. It essentially identified our testing grounds.

Jan-Eric:
Got it. That’s great. So let’s shift gears one more time. I think we’ve got one more example we want to talk about because this one’s a little bit different. In the prior two examples have talked about different sources of information bringing into … we talked about weather and search laying in with the biz, and then we talked about building permits and trade workers with the biz. This third example is really going to a single source-

Zack:
Yep, one data source.

Jan-Eric:
Of all places, the IRS, so let’s talk about that.

Zack:
Yeah, so this one was a client that was trying to plan, well needing to plan events that were targeted at children, and there was a very specific type of child that these events are targeted at.

Jan-Eric:
It’s a camp.

Zack:
Yeah, it’s a camp.

Jan-Eric:
It’s a summer camp type thing for kids.

Zack:
Held all over the country. I mean, really you could have one anywhere as long as you have a building. But the primary customer for this are higher income individuals with kids, and you know, yes, you can build media targeting models on that, but what we needed to understand is where the hot spots of these types of people were across the country and then could a camp be supported in there or multiple camps because there’s several ways you could kind of go into that territory.

And so this was one where the IRS surprisingly was really helpful. The IRS data is all public information, and so they publish the totals … I’d have to look and make sure, but I think all of this stuff was to the county level, well, yeah city, and we know … so I know how many people in each tax bracket submitted a tax return. We know the value of that tax return, and I know if they submitted with the child tax credit attached. So what you’re trying to avoid is picking an area that has a really high income concentration of people with no kids, right? A retirement community where you’ve got higher income individuals all living together in a city kind of skewing results.

The child tax credit helps normalize for that, and so we picked areas that were high income and lots of kids, and that essentially built a planning model, both from a media perspective, but also this was kind of a one-to-one marketing program that was being developed as well where they were going into the areas with specific messages. Once you have all that stuff, you lay it out across the country, and now you know your areas that you have, at least with some assumptions in there, we feel like we have more opportunity. This is a relatively new one, so it’s going to be a good six months before we know if it worked or not, but-

Jan-Eric:
Yeah, and this is a bit of a start up concept, so there’s not a ton of history in terms of business performance, but really trying to guide them to areas that are most likely to give them the best chance to get something going.

Zack:
It’s definitely making you much more educated on your guess, and that’s what a lot of this stuff is. It’s improving confidence. We talk about that all the time. It’s not always answering the question 100%, but if I have the choice to make the decision based on just what I know anecdotally versus use IRS data, I’m going to pick the data every time.

Jan-Eric:
And we’ve given three examples here, but it doesn’t speak to the trial and error nature of this in that it’s a bit of a fishing expedition. You try to be logical about what might be relevant to this, but just because you think it is, doesn’t mean it’s always going to be there. You can’t force it. There’s a bit of trial and error with this.

Zack:
We look at hundreds of metrics and pieces of data, and if we look at a hundred, we throw out 95 of them because usually you’re looking at stuff you think will make sense, but it just doesn’t work out.

Jan-Eric:
Well, again, this is one of my favorite parts about what we do, is coming up with these combinations with these insights. Any final thoughts before we wrap up?

Zack:
Yeah, so the only thing I’ll say is if you’re in a position where you’re needing to make decisions like this, or you’re doing the analysis, go out and search for data. This stuff is freely available. The Census Bureau data is all available. The Bureau of Labor statistics data, it is a gold mine of information for planning. So much so that Google just, this was probably I don’t know a couple months ago, that they released this tool. They now have a search engine just for data sources. So it’s called … it’s at toolbox.google.com/datasearch, and it’s just like the Google search, but anything you search is going to return a data set. So if I search mosquito season in Kansas, they’re going to go out and search and find any data source that’s going to give you some type of insight on that.

Zack:
If you want to see crime statistics in a certain area, they’re going to go grab the crime statistics from that county or state you’re looking at and give it to you in whatever format it’s available.

Jan-Eric:
So if you’ve ever wondered if data exists, that’s a great place to start.

Zack:
Exactly, and lots of this stuff is freely available. Sometimes it costs money. We spend a lot of money on subscribing to data sources and buying data sources as needed, but lots of the best stuff is just out there and ready for you to use.

Jan-Eric:
Well that’s a very valuable tip. It’s like a treasure trove for data. Well, Zack, thanks for your time today. This was a great fun topic to talk about. Thanks for listening.

 

Thanks for listening to our Uncovering Aha Podcast. Callahan provides data savvy strategy, and inspired creativity for national consumer brands. Visit us at Callahan.Agency to learn more.