How to drive business outcomes using weather data

Many brands’ sales are affected by weather patterns, but often marketers fail to understand how and why. Because of that, they may lose sales and fail to capitalize on weather-related business opportunities. In this webinar, you’ll learn:

  • Why the right weather data source is so important, and what technology is necessary to process it
  • How to use weather-pattern insight to get smarter about year-over-year changes in business performance
  • How real-time weather events can impact messaging and timing of marketing tactics

The webinar will include a live demonstration of how the Callahan Intelligence Platform integrates weather data with sales and marketing information, allowing analysts to visualize the business impact and how to respond. You’ll see graphical representations of how weather data is used to drive business outcomes.

Webinar Transcript:

Zack Pike:
Hey, everyone. Thanks for joining the webinar. We’re going to start here in about one or two minutes, as the rest of the group kind of file in here. So we’ll be back in about a minute.

All right. Well, I think we’ve got most of the group on the webinar at this point. So we’ll go ahead and jump in. Thanks for hopping on today. We’re going to be talking about weather data and how to use that data and analysis to drive business outcomes. We’re going to talk from a technology perspective, but we’re also going to talk more strategically, in how you make decisions on this type of data.

I’ll introduce myself first. My name is Zach Pike. I run data strategy and marketing analytics here at Callahan. I’m also responsible for our intelligence platform, which is a piece of technology that I’ll take you through, just so you have some context on how we process all of this weather data and how we use it for analysis.

A couple of housekeeping items before we get started. You should have a way to submit questions through the webinar. So feel free to do that anytime. I’ve got some people in the room here with me that will be collecting those throughout the webinar, and then we’ll answer as many as we can before we hop off here.

The other thing is that there will be a recording of this webinar available afterwards. So you’ll be able to download that at any time and send it on to whoever you think might benefit from this information.

So first, let’s talk about what we’re going to cover. When you think about weather data, you can probably imagine there’s a lot of ways to get it. Some of them cost money, and some of them are free. So we’re going to take you through what to think about as you’re going down this path of how to actually acquire this data, what technology’s necessary, and then how to mash that stuff up against the other data you’re trying to analyze it with. The examples I’m going to talk about here today will be sales information across a big retailer network.

Then we’ll talk about a couple of case studies, and the first one will be using this data to understand the impact of weather on your business, your product, your marketing plan, all that fun stuff. That’s a really important step. I have to know what the impact is before I can actually use it to make decisions. We don’t want to always be working off intuition, because, oftentimes, it’s wrong.

That would be a real case study. So this is a real example that we did for a client, and we’ll show you some views of the data there. Then we’re also going to talk about adjusting your marketing strategy around real-time or near real-time events in weather, and I’m not just talking about hurricanes and big storms. We’re going to talk about temperature and precipitation on a week-to-week basis and how we’ve used that to adjust media plans. Okay?

So let’s hop into the first one here, and that’s how do we even get this data in the first place? I’m going to talk you through the steps that we went through, some of the challenges that we had as we were going down this path, the technology that we use to handle it, but then also tell you the weather source that we use.

The first step, before you even really start processing this data or select a vendor, is we need to prove that we actually can use it to make decisions. So weather … It’s intuitive to think that weather impacts a lot of different businesses in different ways, but there are steps you can take to prove the use case for this, to build a proof of concept, before ever even paying for the data.

There’s a lot of free weather data sources out there where you can download a slice of that information to use in these types of analyses, and our process that we take, even when a client is asking to purchase the data through us, through our subscription, we’ll actually do this as well. So before we’ll put a scope or a contract in front of them, we take a slice of it, prove out that it’s actually going to be valuable to decision-making, and then we go down the path of spending the money on it.

So take that as a first step. There are a lot of free data sources out there to grab weather. If you’re a Google BigQuery customer, there’s actually weather tables sitting inside BigQuery that you can just grab for free and do smaller scale analysis on. The NOAA offers a lot of free weather data as well that you can pull out to do that type of work.

So that’s a big first step. The next thing to think about … So I’ve proven, yes, temperature or precipitation or snowfall, something is a factor in my business that I need to use for decisions. It’s going to make me smarter for forecasting or media planning or whatever down the road.

The next step is to evaluate all of your potential places you can grab the data, but as you’re doing that, we need to look at the data collection mechanisms that they all use. If you go down this path, you’ll find very quickly that they’re very different from provider to provider.

So something like IBM, IBM’s weather data, which is the service that we subscribe to, has an extremely complex collection mechanism. They collect data from I don’t even know how many points – thousands and thousands and thousands of points, all around the world – and their algorithms to get this data as accurate as possible are extremely advanced. In my opinion, it’s probably one of the best weather data sources you can purchase.

But there are sources on the opposite end of the spectrum, that collect data from only a few weather stations around the country, or maybe 100, 200 weather stations around the country, which, if you’re doing analysis on a big retailer network, that’s not enough. The data is not finite enough to really drive conclusions around.

So think about the collection mechanism, and dig into that as you’re going down this path.

The other piece is does the source provide the attributes and the locations that I need? Okay? So when we think about attributes, you’ll see on the screen here, this is a sampling of some of the attributes that we process, from a weather perspective. But it’s not just temperature and rainfall, right? It’s high and low temperatures. It’s the heat index. Is it rainfall or is it snowfall? Relative humidity tends to be a big one that we look at a lot.

So think about what you need. It’s best to collect more data than less and then analyze it and use what you need later.

Of course, budget is a major concern. So the free weather data sources that are out there are not very scalable to get data out of. You can do stuff for a proof of concept pretty well, but if you need to run this out across six thousand retailers or a bunch of distribution points across the country or across the world, you’re probably going to have to spend some money on the data set itself.

There are sources that are a few hundred dollars a month, and there are sources that are several thousand dollars a month. Depending on the attributes that you need, the level of detail that you need them at, and then, back to my second bullet point here, which is the collection mechanism on how they actually acquire their own data, are big influencers on that, that equation.

Your sources that do a ton of data collection, they really specialize in making sure their weather data is as accurate as possible for as many locations as possible, tend to be the more expensive sources that you can purchase from, but they also are the most accurate. If you’re making decisions around budget and product planning, you need accuracy. It’s very important. Okay?

The last point I’ll make, as far as acquisition, and then I’m going to talk a little bit about technology, is infrastructure. Do I have the infrastructure, the horsepower, to handle this data? Then, secondly, do I have the contextual data already there and harmonized in a position that I can use it up against this new weather data that I’m looking at purchasing?

So, from an infrastructure perspective, if you think about weather data, if I’m needing to pull this for a bunch of zip codes … Let’s say I’ve got six thousand retailers. Well, that’s six thousand zip codes that I’m potentially pulling, I don’t know, ten weather attributes for every day or every week or every hour, depending on the analysis that I’m doing. That can create a really large data set.

For example, one of our weather tables that we have sitting in our platform right now is about 200 million rows of data. That’s definitely not something you’re ever going to handle in Excel. That’s … You’re going to need some horsepower on the data processing side to be able to actually use that and draw insights from it.

So, in that vein, here at Callahan, we’re kind of an open book, as far as the technology that we use and how to do analytics to drive brand strategy, inspire creativity, and do all the fun stuff we do from a marketing perspective. So I’m going to show you the technology that we use to do this, and our clients contract with us. They use this technology to get access to this weather data and all of the analysis that we do on it.

The big piece about this information is, for us, it really helps us on the front end of any strategy that we’re putting in place. So our overall approach is when we get on the path of planning media or a brand strategy or a product strategy for a client, we’re using as much data as we can on the front end of that equation. That’s how we get the best results. If I can identify the opportunity, the impacting factors – in this case, we’re talking a lot about weather – I’m much smarter on what I’m doing once I actually start executing Facebook or paid search or my print media plan, whatever the final plan is.

So that takes some horsepower, right? So we start, of course, with the data. So, yes, I need to acquire my weather data, but I’ve got to have something to use that with. In the case we’re talking about here, it is retail sales information. I want to understand the impact of weather attributes on product leaving the shelf. Once I’m smarter about that, I’m smarter about a lot of stuff.

So once I’ve identified those data sources, for our purposes, we have a data transfer mechanism. We use a company called Fivetran for ETL. Then storage. This is like one of the most important pieces when you’re trying to do weather analysis on a broad scale, like we do it here, which is where I think you get the most value out of it.

Our intelligence platform sits on top of the Google Cloud infrastructure, and our core database is Google BigQuery. So when I mentioned that table that had 200 million rows of data in it, that’s actually a very small sample of data for BigQuery. You think about Google, they’re processing outrageous amounts of data every second. My 200 million rows of weather data is small peanuts, as far as they’re concerned.

So we’ve got all the horsepower that we need in that platform. Now, if you’re doing this yourself, of course you probably know there are really strong competitors – the BigQuery, Redshift, Amazon Web Services is a really great platform. Snowflake offers a really great platform. There’s lots of core data bases you could use.

Another big piece of this kind of infrastructure equation, as I’m acquiring this data, is how do I get it harmonized with my sales information or with my media and marketing information? For us, of course, you can imagine, there’s a lot of sequel logic that we are constantly working on to get this stuff as harmonized as possible. But we’ll also employ other tools. The Google Cloud Data Prep tool is pretty slick. Sometimes we’ll use Tableau Prep. If you’re into data, you’re probably familiar with both of those.

Another big piece of our platform is the data overlays piece, which we’re going to talk a lot about here. Weather is just one piece of our overall kind of data library. When you start putting all these contextual data sources together, it makes you very smart very quickly. For us, weather is just one piece of that. If you’re going down this path, you might add in some US Census Bureau data alongside your weather and sales information, just to get smarter. We do that quite a bit.

For us – and we’re going to show you some of this stuff today – visualization … Our visualization layer is Tableau software. We use R and a little bit of EViews for our statistical modeling, and then we’ve got some really smart people here that are really what makes all of this technology tick. It’s a really good thing to be acquiring this data and have it harmonized and at a place that’s usable, but if you don’t have smart people to use it, it’s doing nothing for you. It’s just a big expense sitting there that really isn’t helping anything. Okay?

So let’s talk about some case studies. I’m actually going to hop over to Tableau software here, and I’m going to take you through two examples. The first one is a product that is highly seasonal. When we started with this client, we didn’t quite understand the impacts of weather, but we very quickly identified them and are now using them for planning and forecasting.

The second example I’m going to talk about is a much more transactional product that we’re using weather information to run quick, opportunistic media campaigns, moving forward. I’ll talk you through how we got to those insights, the data we use, and then what we’re doing about it. Really, the point of this is just to share information. Hopefully you take a couple of insights out of this and you can apply it to your own business, or when you’re doing your own analysis on this data, you could use some of these starting points.

So this first one I’m going to talk to you about is a product that is distributed in some of the largest retailers in the country. Walmart is one of those. There were two other big retailers that this product has moved through, and this product’s been around for a really long time. Lots of sales history to look at. We really understood kind of what made the product tick, from a year over year basis. It wasn’t breaking any records from a growth or decline perspective. It was pretty consistent.

What you’re seeing on the screen here is the weekly sales trend for the year. So you see the high seasonal aspect to this. We have kind of a gradual buildup to the … about the middle of summer, and then sales kind of fall off a cliff – the point being that the problem that this product solves, once it gets too hot for too long, the product no longer becomes a problem, and that’s when sales really start to decline.

So that product was cooking along every year, very similar sales pattern. So you’re looking at 2016 here in this light blue line. If I add in 2017, you can see a very similar sales pattern. This was consistent over the last several years, until we got to 2018. So this orange line is the 2018 sales trend, and you see a very large departure in the first part of the year here, where we didn’t follow the same seasonal pattern we had been following the past several years.

You can probably imagine … So the swings in revenue … From a week-to-week perspective, this product was down between 20 and sometimes as much as $35,000 in a given week, which, for this brand, is a large chunk of money, especially when you know you have a finite portion of the year to sell the product. If I’m not on track in the beginning of the year, the back part of the year gets harder and harder every week that ticks by.

So you can probably imagine, if you’re a brand manager or a CMO or a sales person or the CEO in this company, people are concerned, right? There’s some people running around there with their hair on fire, because we think we missed the season, right? Something’s drastically wrong with this product that has performed so well for us year over year and fairly consistently, all of those years.

So that’s about the time we got engaged. The question, to us, was really, “Hey, I’ve got this really weird swing that happened in year-over-year performance. We had a really slow start to our season this year.” There were a lot of theories. So some theories were our marketing program missed the mark. There were theories that the sales program was not on track. There was a lot of talk around retailers and the shopper marketing program, and was product on shelf when it needed to be on shelf? Was a competitor taking all of the share that we had?

There was lots of stuff thrown around, and weather was always kind of an intuitive thing behind the scenes. They thought there was some impact of weather, but we didn’t really know what it was or how much it really affected things.

So when we got engaged, we went through the normal process that we always go through. We started collecting a bunch of data. We got all the sales history on the product, where it was sold, how often it was sold, and we started mashing that up against weather information. Okay?

One of our … If you attended our last webinar, one of the things we talked about a lot is that not every store is created equal. So if you’re distributing a product in, let’s say … In this case, it was four or five thousand stores across the country. You have good stores and you have bad stores, and there are a lot of reasons for why those stores are performing the way they are.

When you think about weather, it’s the same thing, right? So, on the screen here, you’re seeing rain patterns across the country for this distribution network of stores. You’re seeing weeks tick by every time that that view changes, and you’re seeing, basically, the darker the blue, the more rain is happening that week in that zip code.

This kind of tells a story that, well, jeez, yes, I have this kind of crazy distribution network. Every store’s performing differently. I need to understand that, but weather’s the same way. I can’t look at weather information across the country and expect to draw any insights from that. I can’t just say, “Well, it was … The season this year was hotter than it was last year, so that probably impacted our sales,” because the nuances in that weather information get washed out the broader you get, from a scale perspective.

So the way that we do it is we pull everything at the retailer level. I’m pulling the weather around Walmart number 45 – excuse me – and I’m looking at those weather patterns over time in conjunction to sales. Now, you’re not doing this on a one-by-one basis. This is a broader statistical analysis. But this view kind of helps show you that there are drastic swings week to week in weather patterns.

If I hop forward here one more, what surfaced really quickly is a really strong predictor for this product was temperature. Okay? So you saw that seasonal trend that’s like we have this kind of gradual buildup to the peak of summer, and then it really falls off the end, and, again, temperature is a more consistent metric across the country. You don’t have these drastic swings, necessarily, like you do in precipitation.

But even in this view you’re looking at on the screen, you’re seeing temperature … basically, the heat rise from the bottom as it gets to the summer, and as we get into the winter, the cool fall down from the top of the country.

This is all real data. But this ended up being a really strong factor in what was going on in their stores. In 2018, in these markets that we were having a problem, the temperature pattern for 2018 was drastically different, the beginning part of the year, than it had been over the last several years. It took much longer to heat up, and once it did heat up, it was really hot really quickly. We didn’t have this gradual build to temperature.

So what you’re seeing on the screen here is the weekly temperature trend for 2016. I’m going to add in 2018. These look kind of familiar to the charts we just looked at for sales, and then now I add in the sales trend, and, like holy cow, it looks almost identical to their revenue trend. Obviously, we don’t have this cliff at the end in temperature like we do in their sales, but the beginning part of this year – right? – we see this drastic swing in 2018 vs. the prior years.

If I put these up against each other, you’ve got temperature on top, and you’ve got sales revenue on bottom. This, again, tells the story. I’ve marked the period that kind of caused the most heartache in this organization, but you see … This tells the story. There’s … You didn’t need a big presentation or a statistician at the client to understand, “Holy cow, yes, it’s pretty clear temperature impacts our business.”

In this case, it was average temperature for the day, what we analyzed. We’ve seen some examples where it was the minimum. The minimum temperature for a given week had to be at a certain level for a certain number of weeks before we saw an impact in sales. There’s all kinds of ways to look at this data.

So there’s, of course, some other kind of cool views we could do on this information. This view is showing you a literal heat map for each year. So you’re looking at weeks in each of these little boxes here. You see the summer months, but this one does a bit better job of explaining how many weeks you stayed cold, right? So this green part goes much deeper into the year than it had in the prior two years.

But then another factor for this client is the heat started much sooner. Basically, we missed … We didn’t have spring this year, in these areas of the country, which spring is a very important time for this client.

So, at this point, now you know, right? Now I know, okay, temperature impacts my business. I can use that for forecasting. I’ve explained, at least broadly, what happened in 2018. I’ve armed myself, one, to be able to handle conversations internally, but also to bring some comfort to some people who are probably a bit concerned about the overall strategy, from a marketing perspective.

But what I want to challenge is think if we would have known this in 2018. Think if this analysis would have been done two years ago, and we did a five- or six-year study to say, “Okay, I need to understand the impact of weather.” I do that study. Now I know temperature impacts the business. If I would have known that heading into 2018, the year might have been very different, right?

So this client, when this started happening, they really didn’t make any changes to the marketing strategy. They absolutely did not pull budget back, because they thought they were missing things. If anything, they were probably considering pushing more money into the market to try and knock off whatever competitor had kind of moved into their territory.

But think if I would have known that this weather thing was a factor here, right? Early in the season – let’s say, like, week number four or five of having really low sales, in comparison to last year, and low temperature – I might have thrown an alert, right? The data guy who’s watching this stuff might’ve walked over to the brand manager and said, “Hey, we’ve got problems coming up this year. I think that the season’s going to start later. I think it’s going to be a shorter season. We need to start thinking about what we would do if that happens.”

A few more weeks tick by, and now you’re six, seven weeks into this season. You’re seeing that, yes, sales are down, but I’ve got an explanation for why sales are down. Now that I have the explanation, I’m smarter about what I should do, right? I might start pulling back marketing budget. If I’m advertising in this part of the year, this early part that’s down, I’m advertising at a point where the fish aren’t biting, right? They’re definitely not biting at the rate that they were in prior years, and it’s due to this kind of weird season we’re having from a temperature perspective.

So I might start to pull back, at least in my most impacted markets, that budget wherever I can. If I’m doing TV, that might be difficult, but if I’ve been doing digital campaigns, that’s definitely within the realm of possibility.

I might also be talking to my retailers, right? I might call my buyer at Walmart and ask them what they’re doing about this. One, I want to make sure they know that this is a problem. You would be surprised how often they don’t, and that’s not specific to Walmart. That’s every retailer. They’ve got a lot of stuff they’re worried about, but … and then talking to them about what we can do when the season actually comes back, because, at this point in the year, I would know, one, I’ve got a weird season on weather, but the most important piece of this is I know I’m going to have a really short selling season this year. So all this money I’m losing early in the season, I need to figure out a way to extract that value during my peak, when the season finally comes back.

So that’s going to be a team effort between retailers, the sales team, my marketing approach, and everything. So I’m having those discussions at this point in the year, before this is really a huge problem.

But, then again, if I’m monitoring this weather data, if I’ve got an analyst on my team or I’m working with a company to help me with this, we see this peak here in week 11, where it looked like the variance was starting to shrink, right? I’m starting to get back on temperature. You might get a little bit of faith here to say, “Okay, sounds like the season’s coming back. Let’s think about what we might do next week to start turning things back on,” ’cause you actually see it in sales information down here as well.

But then we fall back off a cliff, right? So this … We just dropped temperature again, and we kind of hung out here for several weeks. At this point, you’ve already pulled back on marketing where you could. You’ve already started having conversations with retailers. At this point, it’s like, “Jeez, I’m really glad we made that decision,” right? Because, at this point, without this weather data, I’m just burning cash that’s not producing anything for me or is producing at a much lower rate of efficiency.

Then, of course, things start to get happier, right? Temperature comes back. It really surges very quickly, so, in a matter of about three weeks, they made up all that loss in temperature, which is a very critical three weeks, because that’s when sales start to come back as well. If I’m already primed to be watching this every week – or, depending on the business, maybe even every day – to say, “Okay, I feel like the season’s coming back. We’ve got a good amount of temperature history, to where we’re trending in the right direction,” now I can turn on marketing right at the point where customers are going to be starting to think about this product again, right?

‘Cause once we know, once those temperatures hit, the problem presents itself, customers realize the problem, and then they go into the retailer to purchase it. We want to be in market when they’re having to make a decision about what product they’re going to purchase, ’cause they’re going to purchase something. We want them to purchase ours. So if I’m smart about this weather, I can start turning that media on at that point. Okay?

So hopefully this gives you a little bit of view into how this can be used from a scheduling perspective – right? – to make me smarter in campaign and in the season. To a guy like me, it’s really cool and hugely valuable, and our clients have found a lot of value out of it as well.

So one more example, before we hop to questions here. This next example is a much more transactional product, and it’s kind of a problem solution product. It’s not something that you’re going to stock up on. Think about a product like Off, right? So Off is designed to keep pests away from you, from biting you – mosquitoes and any other kind of really annoying, biting pest. Think about how you purchase that product, right? You’ll purchase it at an event.

So I’m going camping, or I’m going to be at a picnic, or I’m going to be trick-or-treating or something, and it’s warm, and I know mosquitoes are going to be out. So I’ll purchase the product.

The other way you purchase that product is after you already have the problem. So I’m at a baseball game in an evening. I’m getting eaten by mosquitoes. I go buy the product. Okay? It’s much more transactional, and this client came to us with a question. Basically, it was, “Hey, I’m seeing these kind of weird peaks and valleys from a week-to-week basis in sales. I really have no explanation for why that is. We’re not doing any marketing things that could be causing that. Our shopper marketing program is pretty consistent. Retailers aren’t changing shelf position or doing in-store signage or anything that could be causing this. I really don’t know why it’s happening.”

What you’re seeing onscreen here are those peaks and valleys. So these are sales trends for four of the roughly, I don’t know, 2500 stores that this product is distributed in.

So, again, we followed our same process. We started looking at a bunch of different data. We, of course, pulled all of the media and marketing data that was happening. We, of course, already had all of the sales information that we needed to do this analysis, and, again, for the topic of this webinar, we looked at weather data. We didn’t know what the factor would be in weather. We just had a hypothesis that weather may be somewhat of a factor here.

So if I throw rainfall totals by week up against those sales trends, it starts to bring forth some really interesting and useful insight. So just looking at this chart, it probably doesn’t make a lot of sense, but you’re seeing, again, your revenue totals by week in the lines and then your rainfall totals by week in the bars. This is for four of these stores.

If I call out some of the examples, what actually ended up happening is when we see rainfall elevated in a given week, for the one, two, and sometimes as far as three weeks after that elevated rainfall, we see elevation in sales activity. Okay? These are products leaving shelves. So this is retailer consumer sales activity, and it was super consistent.

So I’m showing you four stores here. This actually proved true across many – excuse me – many more. If I scoot forward in the year here, you see I see an elevation in rainfall, an elevation in revenue. Elevation in rainfall, elevation in revenue, and the same holds true down all of these. These little bumps are almost always explained by elevated levels of rainfall.

Through this analysis, we actually identified what that magic number was. So it’s not enough just to know, okay, yes, I had more rain. I should expect more sales. It’s how much more rain do I need? I’m not going to tell you what that threshold is, because it’s extremely important information. But there’s a very specific number. Once we surpass that number, we know to expect a certain level of incremental sales for the one, two, and sometimes as far as three weeks afterwards.

So this proved true – again, I’m kind of clicking forward here – through the year. But with this product, specifically, when we got to the back part of the traditional season for this product … You think about flying pests. When it starts to cool off, you tend to stop having … You tend to not need products to solve that problem anymore. The pests kind of go away, and they’re not anymore.

But there were some areas of the country where we were actually seeing our best weeks, from a sales perspective, that we had seen all year, and that was really uncommon. We didn’t have history to support that that would ever happen. Oddly enough, what was happening in those areas is they were having their biggest weeks of rainfall.

So, in store number 120 here, you see my biggest week of rainfall. One to two weeks after that, I see my biggest two weeks of sales activity. This is real data. This is not made up for this webinar.

Same here in store 765. I see three really big weeks of rainfall. This area just got hammered with rain. Then three weeks of my best sales activity. Okay? The pest that this product handles – again, kind of a parasite-type pest – is highly dependent on rainfall. When it rains a lot, they start breeding, and then they become a problem. So it made sense that this stuff all kind of lined up.

So think about now that I’m armed with this information. Okay? Now I know what my threshold is, I know how much I need, and I’m processing this data every day and every week to monitor it, moving forward. So think about if I was building my marketing plan for next year. This type of information probably isn’t impacting the broad plan for the year, right? I’m not going to communicate to my board that I am predicting what’s going to happen from a rainfall perspective in 2019 when we’re not even into 2019 yet. We all know how our weather men do at forecasting what’s going to happen next week, let alone next year.

But if I’m monitoring this data on a weekly basis, I know that if I get a certain amount of rainfall this week, I should expect consumers to be out looking for a product next week and the week after. I know that’s going to happen, like clockwork, every time in almost every location for this product. I can build marketing programs around that, right?

So think about something like Facebook Ads. It would be very easy to pre-build Facebook Ads campaigns that were designed to be turned on just to take advantage of this increased consumer activity, trying to steal share from my competitors in those one, two, and sometimes as far as three weeks after that rainfall happens. Okay?

Many times, I believe that your competitors are not this smart. They’re not doing this. So by you taking this approach, it gives you a leg up from a competitive perspective so that now, when that rainfall happens, I get this influx of consumers walking into Walmart, for example. They’ve got a selection of five products they can pick from. If I’m heavy in that store’s trade area with a really strong, digitally targeted message that’s extremely relevant to the problem they’re dealing with, that’s not just my normal, typical media campaign, it … I mean, I think we would all agree there’s almost no way that couldn’t have some impact on what’s happening, from a decision perspective, for that consumer.

So that’s, really, the approach that you would take with this type of data. Okay? So I think this is a really cool example. Hopefully, both of these kind of give you an idea of how we use our intelligence platform for analysis and processing of this weather data.

Now, I’ve got a couple of questions that I’ll read here that have been submitted, and then I’ll answer them for you here.

So the first question is, “When you were looking at this data set, what other sources did you look at?” It looks like it says, “What other weather service providers did you evaluate?”

So that’s a really good question. Like I said, at the beginning, we ended up going with IBM’s Watson product for weather. This weather data is actually used … You probably are consuming it on a daily or weekly basis, and you don’t know it. It’s used in a ton of applications – the Weather Channel, stuff like that.

But we looked at a lot of sources. So Accuweather was one that we spent a lot of time evaluating. They have a really good product. It’s similarly priced to IBM. The good thing about IBM is we hook into an API. So we pull the data that we need when we need it, where some of the other weather providers want to ship you files for very specific requests that you have and that you need filled.

Then there’s a host of kind of smaller weather data providers out there that are less expensive. There are some providers that you can get for a few hundred bucks a month that will solve many of the challenges you might be dealing with.

But, again, before we even went down this path, we used some free weather data sources. Like I said, the NOAA, lots of their data is already sitting in BigQuery. Google’s kind of curated some of that data. Then there’s other websites out there that will give you weather data you can use to pull a proof of concept. Okay?

So the next question here is, “It seems simpler to overlay past weather data with past sales events, but the predictive part seems like the bigger challenge. How accurate is the forecasting?” Then the followup here is, “How do you plan a marketing plan if the forecast is unreliable?”

So I kind of touched on that earlier. So in the weather that we process, we get historical data, and then we also have access to all of the forecast data. I think it goes out ten or fifteen days, and, yes, the forecast data, you have to take it with a grain of salt, because there’s only so much predicting we can do from a weather perspective.

Larger events, like hurricanes, are much … You can predict those sometimes as far as several weeks out, but that’s … Honestly, that’s not something that you would need a really complex weather data set to really evaluate. Where it comes more into play is in the minute changes of temperature and rainfall that impact your business, and, honestly, the forecast data there is sometimes reliable and sometimes not.

The examples I talked about here were picked specifically because, in those scenarios, to predict what’s going to happen, we actually only need this week’s data. I need this week’s actuals to know what’s going to happen next week. So in the rainfall example, I know, once I reach that threshold … So if I pull my data, let’s say, on Friday, if I’ve surpassed that threshold for the week in the trade area of a given store, I know, next week, I should expect increased sales. I don’t need the weather forecast to be able to make that prediction. Okay?

Now, that doesn’t mean the forecast data isn’t valuable. It doesn’t mean that it’s not useful. But this is kind of how we tend to use it, again, because that forecast data is only so reliable.

All right, and looks like another question we’ve got here is, “You’ve talked through two examples specific to temperature and rainfall. Are there any other weather-related environmental factors that you’ve seen impact a business’s sales?”

The answer is yes. For example, I think I mentioned, on one of these slides, relative humidity. Oddly enough, relative humidity is well correlated to a lot of the products that we analyze. So if you think about parasites, parasites tend to like humid environments – not only from just an overall geographic perspective, right? So you think about the South, right? You expect a lot of fleas and ticks and mosquitoes in the South. But even on a week-to-week basis, you will see parasite activity elevate when the relative humidity elevates, ’cause they’re more active, and they’re just out and about more, bothering people.

So that’s one example. We’ve done a lot of work around products where the minimum temperature has to be at a certain level for a certain amount of time before a problem will present itself that a product is there to solve. If you think about one kind of easy to understand example is if it gets below freezing, you’re killing off stuff, right? Several types of bugs. You’re killing off several types of plants when you get under that freezing threshold in an evening.

If you have several weeks where you’re above that freezing threshold, things change, right? You start growing those insects and those plants, and you’re not killing them off, which there are products designed to either help that growth or kill off those insects that are pests that you want to get rid of.

Then another big factor, oddly enough, is sunlight. So there’s a lot of research to suggest that – to really prove that – consumers are more active when there’s more sunlight. When the days are longer, people are shopping more, and they’re buying more stuff.

But even if you look at things like visibility and UV index, that can give you insight into that on a weekly or daily basis, right? If you put that up against rainfall, you can start to predict storms and what might be happening inside of a market that might be the reason it’s keeping people inside, right? You think about if you’re a restaurant owner, it’s raining, it’s really stormy, people just don’t want to leave their house.

When you have that context, if you’re analyzing a big restaurant network, when you have that context and know, “Okay, this week was a really stormy week in these five markets. I should expect lower sales activity in those markets,” it gets really powerful.

All right. So I think that’s all of the questions we’ll answer today. Before I wrap up here, I’ll just give you kind of the ten-second elevator pitch on Callahan, and, really, at our core, we use … We drive business results with data-savvy brand strategy and inspired creativity. That’s kind of how we describe ourselves, and on the screen here, you see a few bullets that outline some of the services that we offer.

I said this in the beginning, but our approach is to use data on the front end of anything that we do, because we think it makes us smarter on what the opportunities are that we can tackle with marketing or creative or media or any of that, that fancy … those fancy approaches to marketing. It’s data that empowers all of that stuff. Weather is just one of those factors in our library that we’ll dip into when we feel like it’s going to bring value to our clients. Okay?

So if you want to learn more about us, you can go to callahan.agency. There’s lots of good information on that website.

So, again, thanks for attending the webinar. The recording will be available shortly after the webinar is complete. You’ll get an email with, I think, the recording in it, and then you’ll, of course, be able to download it on the website anytime.

Feel free to reach out to us with questions if something pops in your head around how we would approach an analysis or how we would handle a problem that you’re dealing with. Feel free to reach out. Those requests will get sent to me, and I’m happy to help out however we can. Thank you.