I-COM Data Creativity Awards 2019 / Analytic Partners (USA)

all right so we're gonna take you through a case study on how we leverage whether based storytelling to impact business decisions so show of hands how many of you checked the weather before you packed for a Malaga good and it was relatively accurate I'd say but what's fascinating though is just how complex weather is it's contextual we don't feel weather at 75 degrees we feel it as hot and humid or cold and damp it's also pretty personal what I would consider a nice beautiful day as a New Yorker visiting Malaga might be different from somebody who's a local and experiencing this all the time and it's also temporal but I would consider it a nice beautiful day in July is going to be very different from what I'd experienced in February so it's very very complex and there's lots of nuances involved with weather and its influence it can also be very significant in terms of impacting businesses the weather can have a direct impact on sales it can also change the tenor of an ad or make a marketing or sales Stratus strategy successful or not and for one of our clients weather was particularly important it was one of the most significant drivers of business performance so significant complex and they had absolutely no control over it to add to that this particular client is in the lawn and Gardens category so they have a very short window in which to hit their sales roughly 80% of their business is sold over 100 a window added to that they have a portfolio brand so their lawn fertilizer products for example can't be spread when it's raining they also have bug and weed prevention products which do quite well when it's hot and muggy so there's a lot of complexity there in terms of how to manage their portfolio with these different weather conditions added to that is the regional dynamics the United States is a big country and you've got a lot of different weather conditions happening all at the same time so for this client weather was a pretty big challenge and for most of the organization a pretty big headache because they were having problems predicting it they were having that no control over it the challenge for us was to how would we take this problem and turn it into an asset how do we take a lot of data and leverage data science to separate signal from noise and make actionable insights and capitalize on the opportunities and minimize on the waste due to weather and we needed to do this in a way that we could transform how the whole organization is making decisions and Jesus all on a real-time basis as the weather is evolving and the business is moving forward so in order to solve this challenge it's led us to the creation of helio which is a weather driven demand solution that has really transformed the way that our client thinks about the weather the way they make decisions and the way they inform their strategies so one of the things that we have coached them on what they've been able to see with Helio is that the weather it can be a lot like the stock market think about it the stock market goes up a lot of people make money the stock market goes down small group of people still make money because they know how to respond and that's exact same type of attitude they have a job towards the weather now good weather can be an asset bad weather can also be an asset if you know how to respond to it and hylia delivers that to them by give them two things one it gives them the numbers that they need but more importantly it gives them the narrative so at a snapshot they can see what is happening with the weather and they can see how their consumers are and going to respond to it so they have this narrative that they can take to their internal teams their executive teams their strategic partners and go in there and say like hey here's what's gonna happen with the weather here's how our consumers are gonna respond so here's what we need to do about it and of course all this comes through the machine learning and the algorithms that we develop because we're minding a lot of very complex data and a lot of very significant interactions that are happening at the hyperlocal level one example here this is for specific market for specific product in a specific week were able to see what is a consumers response to changes in temperature or precipitation for that fact and each one point in time for each product and what's great is that these insights have been used across the whole organization media marketing team they know when to plan the media start dates when we're too heavy out director consumer can better optimize app experience you know which products to target sales they know where to put the boots on the ground farther in advance where did the play promotional staff product improve the customer experience supply chain can know for months in advance now where they need to have the product ago so they avoid wastage and Finance has better lens into what it's going to happen with their earnings to inform those quarterly projections for Wall Street so what is the actual impact of this now when we launched in January of last year we knew that April historically their biggest month there are facing some significant headwinds and looking at a double digit losses in their biggest month of the year and in fact played out almost exactly as expected can't give these specific numbers but the difference between those bars it's about a half a percent so thankfully having four months warning for this they were able to take action they were able to shift their media start dates or TV start dates to later than the season in to make well they knew the consumer demand was going to be there they were able to pull back on some their assets they weren't wasting the marketing assets at a time when the consumer demand simply wasn't going to be there and so how did it play out biggest sales month ever in the history of the company represented a 30 million dollar opportunity just by capitalizing on the weather that was there and so the key takeaways of this the number one the weather is extremely complex we experience it as consumers not in terms of specific numbers okay and that any weather outcome can be an asset if you have the right intelligence and the valued data is in the story that it has to tell data by itself is just information the value of it comes from learning's that you can attract remit the story that relies in there and that's what you can get to if you have the right approach working on this project we were constantly thinking of the scope by Louis Pasteur where it said chance favors the prepared mind and when you're planning around the weather and how people are gonna respond yeah a chance is a pretty big part of your calculus right so we're very proud to say that with Helio now our client is well-prepared for anything anything that chance has in store for them thank you [Applause] thank you very much and again are there any questions quick question regarding the daytime weather what kind of variables did you use if you can share more precisely was it temperature was it humidity was it what so weather is a kind of a bunch of things if you could elaborate on that so in our historical data set we had everything from temperature precipitation solar radiation dew point all that but looking forward all we had was temperature and precipitation that's really what we focused it on and so was really critical then just when we were building out our analysis did a lot of feature engineering we spent a lot of time just looking at creating these different weather profiles so not only does temperature but what is the temperature and this week and what what's in the last two weeks has there been extended warm period of cool period there's a nice day but has this been after a succession of nice days things of that nature so with just temperature and precipitation we were able to create a lot of significant features that ended up creating these these weather profiles across the continental US hi can you talk a little bit more about the geographic segmentation like how what was the deepest level you went in terms of geography and the second question is in terms of dependent variable did you look at anything else besides sales sure in terms of geographies so this was all built at the DMA level which is the Nielsen marketing regions in the United States and we had the ability and we started the weather data that we can get can go down to a half kilometer grid so we can get super super accurate and hyperlocal at these forecasts but then also roll it up to in this case we ended up at the DNA level because that's how they do their planning that's what's most relevant for them so we have the ability to go super local but we roll it up to the levels where they could actually take action on it and for the dependent variable we look primarily at sales but now we're also thinking about more things there's a concept I'll call that growing degree days which is like how many like nice like seasonal days are there and which plans can grow and so now we're working with their science teams to better help inform those predictions as well so they know what when products can be growing are they gonna be going spreading earlier or later in the season and so they can better target then with things like products that fight these types of weeds so they know when those weeds are gonna be growing so now they know when to target them sooner in the season so thank you very much Katie and Brendon [Applause]

Leave a Reply

Your email address will not be published. Required fields are marked *