In his new book, Consumer-Based Forecasting and Planning, Predicting Changing Demand Patterns in the New Digital Economy, Charlie takes this experience and combines it with world leading analytics powerhouse SAS to delivery an incisive and novel approach to forecasting demand for retailers.
Welcome to the The Voice of Retail , I’m your host Michael LeBlanc, and this podcast is brought to you in conjunction with Retail Council of Canada.
Charles Chase has spent his entire career working with high end analytics to help understand and predict consumer behaviour.
In his new book, Consumer-Based Forecasting and Planning, Predicting Changing Demand Patterns in the New Digital Economy, Charlie takes this experience and combines it with world leading analytics powerhouse SAS to delivery an incisive and novel approach to forecasting demand for retailers.
In a book and this interview right for the times, we talk about how the changes in people, process, analytics and technology plus machine learning can address our most challenging forecasting opportunities.
Thanks for tuning into today’s episode of The Voice of Retail. Be sure to subscribe to the podcast so you don’t miss out on the latest episodes, industry news, and insights. If you enjoyed this episode please consider leaving a rating and review, as it really helps us grow so that we can continue getting amazing guests on the show.
I’m your host Michael LeBlanc, President of M.E. LeBlanc & Company, and if you’re looking for more content, or want to chat follow me on LinkedIn, or visit my website meleblanc.co!
Until next time, stay safe and have a great week!
Charles Chase
Charlie Chase, Author at SAS Blogs
Senior executive with experience in the Consumer Packaged Goods industry specializing in Global Strategic Marketing, Demand Forecasting and Planning, Market Research, Market Response Modeling and Analytic’s, Knowledge Management, Data Warehouse Design, Systems/Applications Development, and Project Management.
Specialties: Global Strategic Marketing & Planning, Market Mix Modeling & Simulation, Global Strategic Forecasting, Market Research & Analysis, Forecasting Systems Design, Forecasting Process Design, Project Management, Supply Chain Management, Data Warehouse Design, Data Warehouse Tools
Michael LeBlanc is the Founder & President of M.E. LeBlanc & Company Inc and a Senior Advisor to Retail Council of Canada as part of his advisory and consulting practice. He brings 25+ years of brand/retail/marketing & eCommerce leadership experience, and has been on the front lines of retail industry change for his entire career. Michael is the producer and host of a network of leading podcasts including Canada’s top retail industry podcast, The Voice of Retail, plus Global E-Commerce Tech Talks and The Food Professor with Dr. Sylvain Charlebois. You can learn more about Michael here or on LinkedIn.
Michael LeBlanc 00:04
Welcome to The Voice of Retail. I'm your host, Michael LeBlanc. This podcast is brought to you in conjunction with Retail Council of Canada.
Michael LeBlanc 00:10
Charles Chase has spent his entire career working with high end analytics to help understand and predict consumer behavior. And his new book, consumer based forecasting and planning, predicting changing demand patterns in the new digital economy. Charlie takes his experience and combines it with world leading analytics powerhouse SAS, to deliver an incisive and novel approach to forecasting demand for retailers. In a book and this interview right for the times, we talked about the changes in people process analytics and technology, plus machine learning, and how that can address our most challenging forecasting opportunity.
Charles Chase 00:45
So, there's always going to be collaboration between people and people, and people and machines. So, it's never going to go away, it's going to be minimized. So, machines are going to do more of the heavy, heavy workload working. So, let me explain what I mean. So, what you need to do is first take the letters AI and flip them to IA. It's really about Intelligent Automation, supported by machine learning. So, what does that mean, that means we want to automate all the mundane non-valuated work.
Michael LeBlanc 01:13
Let's listen in now. Charlie, welcome to The Voice of Retail podcast. How are you doing this morning?
Charles Chase 01:17
Very good. Thanks for inviting me on, Michael.
Michael LeBlanc 01:19
Well, thanks so much for joining me. I have your book in front of me and we're going to get into it. It's a weighty, tomb of information and it's actually a great read for the novice and also the experts. So, really excited to have a conversation. So, let's start at the beginning. First of all, tell me about yourself, your background and what you do for SAS.
Charles Chase 01:39
Okay, well, I'm the Executive Industry Consultant for the SAS global retail and consumer packaged goods, practice at SAS, and when, SAS about 18 years. Prior to coming to SAS, I worked for 20 years. In the consumer packaged goods industry. I started out as a Demand Analyst, Manager, Director, Senior Director, and at several different large corporations and then ended up at Heineken as the vice president, Chief Knowledge Officer for Heineken USA. Prior to that, I worked for Coke and Johnson and Johnson consumer products and Reckitt Benckiser.
Michael LeBlanc 02:14
and, and when you were growing up, is this something you always oriented towards business or analytics or what drove you in the, in the, kind of, the general direction that you took?
Charles Chase 02:24
Well, I was always in analytics. In fact, my undergraduate degree was in economics and computer science. And then, when I was working on my MBA, I started working for a private company called the Mennen company, which was still owned by the Mennen family. I ran across the manager of forecasting and I had happened to be taking a course called business forecasting MBA. And I realized I can apply all the statistics that I learned from economics and econometrics and that's what I majored in. My MBA was econometrics and management, information management.
Michael LeBlanc 03:00
Right on.
Charles Chase 03:00
and that's when I became an analyst.
Michael LeBlanc 03:03
Right on, using those University skills or college skills to put them to good use. Tell us about SAS so certainly a well known brand name, a well known name, are very well respected, but perhaps might, you know, might and perhaps not everyone might know or appreciate the full scope and scale give us an overview of SAS.
Charles Chase 03:22
Well, first of all, SAS is the largest privately held software company in the world. We're at 3.3 billion in annual sales. We've also are number one, in advanced analytics machine learning by all the big analyst firms like Gartner, Forrester and others, for I guess, eight or nine years in a row now. So, we are also number one in machine learning as well as predictive analytics. Been around for 45 years on the same owner Doctor Goodnight has been leading us, phenomenal place to work and work in global headquarters, right where we have about 6000 employees. Globally, we have about 12,000 employees or more, and being in a global permit position to get to travel a lot, around the world and I get to meet a lot of really important people, executive levels at big companies like Miller, Coors, and Nestle and Kellogg, and many, many others, Hanes brands and others.
Michael LeBlanc 04:19
Right. So, a tier one list of customers, I was actually a customer when I was at Rogers media of your product, I used it for the Shopping Channel to do our modeling and analytics. And it really was to help us do some really breakthrough things. So again, I'm excited about having the opportunity to discuss where that sits today. But, let's talk about your new book consumption based forecasting and planning, predicting changing demand patterns in the new digital economy. It's, it's an, it's an interesting read for sure. Tell me a little bit about who you wrote the book for and the central premise that really retailers switch from supply centric to demand centric approaches.
Charles Chase 05:03
Yeah, the book is targeted toward executives but also on director levels. So, folks that are Vice Presidents, supply chain, the CFO, CEO, the company, Chief Marketing Officers, actually too, this book talks a lot about moving demand forecasting and planning, from the supply chain operations planning, where it is traditionally, to marketing, sales and marketing. I always reported into sales and marketing when I was a senior director of demand forecasting and planning. And, I also oversaw marketing analytics. And, when I was at Coke, I was part of the global market research organization, primarily overseeing all the syndicated scanner data, and POS data and all the SaaS programmers. But we did a lot of analytical work like around frequent shopper data show wallet, Brian, things like that, and long range, medium range forecasting, marketing mix modeling.
Charles Chase 05:57
So, I applied all that to one of my days when I was already applying that information. In my early days, when I was working at J&J Consumer Products, where we, I reported into the Vice President of Marketing as the Director of Demand Forecasting and Planning. So, I was already positioned downstream, closer to the consumer, I had access to POS data, also to syndicated scanner data, Nielsen, and IRI. So, we were able to integrate that information in, with the shipping data. Most companies today are forecasting from supply to the consumer there. And, the reason we're doing that is because their demand planners are too far removed from the consumer upstream in the operate-, and supply chain and operations planning. And they're forecasting shipments.
Charles Chase 06:41
So, shipments is actually the supply signal. And a lot of companies over the years got moved to sales orders thinking that was true demand, actually sales orders as a replenishment. The true demand for both the retailer and consumer packaged goods company, is the POS data and the syndicated scanner data. That's true demand, if you're too far removed upstream from the consumer in the supply chain, you don't have access to that data or a really good understanding of those things that sales and marketing folks do the influence demand. Like in store merchandising,
Michael LeBlanc 07:16
Right.
Charles Chase 07:16
displays, features, those things. And then you need to have a more of an advanced analytics background. Most demand planners, they are really managers of information and data there very little analytics background. When I was a director, I hired right out of universities, graduate students that had master's degrees in statistics, econometrics, so they were demand analysts, they already knew how to apply the analytics and I taught them the business, which only took about 30 days, and they were building models and adding value almost immediately.
Michael LeBlanc 07:48
It's true, as I read through the book that I saw it at both levels, because some parts are very, very good book to be read by executives who oversee or who are influenced or who influence demand and supply and marketing and others. There's some practitioners, because there's a few, there's a few models, a few formulas in here that if you're in that business, you're gonna love.
Michael LeBlanc 08:09
Let's talk about this. You call it signals, let's talk about, let's talk about the noise to signal ratio that we're currently experiencing. Right, so, and you touch on it in your book, when you're talking about a wakeup call, you know, how do you handle the abnormal historical data after it's over, I mean, certainly, edge cases and unusual things are happening in the before time, pre COVID, but, it feels those outliers, you know, and this is where I'd like you to chime in the fields, like those outliers would kind of be ignored by the models because they were seen as outliers and, and you didn't want to chase them down too far. How do you know, over time, what's the outlier and what's a trend that you need to pay attention to and then re, you know, rejig your logs to figure that out?
Charles Chase 08:52
Well, well, let's start with the outliers. Most outliers are cleansed out of the data. Manually mod demand planners, that's one of the key things that they do, they cleanse the data into what we call base and promoted, they separate the data and then they try to piece it back together manually, an Excel spreadsheet, because their technology can't correct outliers on the fly automatically. One of the great things about SAS is we have automatic outlier detection, it automatically finds these outliers in many cases, many of the outliers are really important. And they're not really outliers in the sense of the word that it's, you know, a one-time occurrence. Even if it is, you want to capture that one-time occurrence. So, well, many times we'll, we make the outlier detection more sensitive than normal, and it'll pick up more outliers. In many cases, those outliers are positive. And, they're actually promotions, promotion lives. And we'll go back to our customer and say you did you run a promotion here, here and here? And they'll say, you know, what, how did you know that. It's lower outlier detection, picked it up. It's using advanced analytics, an optimization, and it finds those outliers and corrects for them automatically and says this is an outlier. So, I'm going to correct for it. But maybe it's not an outlier and if it's not, then we want to leave it in. So, that's another way of capturing those promotion lists without having to cleanse the data.
Charles Chase 10:16
So all this work they do that we'll have to do is if they had SAS they'd just say, turn the sensitivity up a little bit, the automatic outlet, and it would automatically find them and correct for them or you can then investigate those outliers and determine whether or not they're important and if they are, where they happen again in the future and also, if you're looking at only trend and seasonality using a very simple model, that's how most DRP solutions, enterprise resource planning solutions use very, very simple models that only look at trend and seasonality. If there's an unforeseen, unforeseen disruption, like COVID-19, those trends and seasonality are no longer valid in history. So, what do you do?
Michael LeBlanc 10:57
Right, right, right.
Charles Chase 10:59
So, all you have to do is think outside the box, and I said, what do you mean by that? Well, I mean that, what are those things that are causing this disruption, and can you use them to influence demand for your products? So, the one of the things that we do we work with a Russian, retail, an online retailer, and in the middle of the pandemic, during the lockdown, they were seeing, as much as 200% increases in some of the products that were essential, and 50% decreases in those products that were non-essential. So, those of you that were essential, we're increasing, you know, exponentially. So, we started to work with them and the first thing we did is we went we said, let's, we're going to apply machine learning, learning models, more advanced, advanced predictive analytics, we're going to bring into play their point of sale data, Google Trends data for those, those categories, epidemiological data, and regional economic data and as we started to add that information, we realized that there was a strong core, what we would call correlation are influenced by those outside factors on the products more able to and allow us to predict those increases in those products and the decreases as well.
Charles Chase 12:09
So, we were able to improve their accuracy from around 70%, prior to the COVID-19. impact to 91%. In the middle. It's not just about trend and seasonality. It's about what are those things that are causing people to change, or shift or demand patterns for my products, then you have to find the data that's associated with that, and then use that data in a predictive model to model that influence and then if you layer on top of that, machine learning improves it even further. Because then machine learning starts to learn after a few cycles, and it starts picking up those patterns based on those outside factors and improves it even further.
Michael LeBlanc 12:51
Now, is that the art and science of building SAS products with you and your, your 1000s of colleagues, do they look at, you know, it's so interesting, epidemiological data, Google data, you calibrate the importance of a gauge on how much those influenced the model baked into the product? Like how much would you say, your team continues to tweak the models that then, you know, the, the end user has to play with? And I guess where I'm going with that is, how do you guys sort out the difference between let's make sure that there's enough power in the model versus too much flexibility in the hands, it could throw it off the, could be just a bit too confusing, even for the advanced user, how do you how do you balance those two things?
Charles Chase 13:34
Well, you know, there's a lot of people think that the more data they throw in here, the better. That's not necessarily the case. In fact, if you have too much data, you're underutilizing the important factors and models versus those that are not important. So, you can overestimate the model, especially when you're using machine learning. That's classic, but you could do that with almost any great model. You know, so, it's all about finding the right factors that are causing that impact, and are influencing that, impact that, I should say.
Michael LeBlanc 14:04
Right.
Charles Chase 14:05
And, then also, I find that if you, more than 10 or 12 factors in a model, you probably have too many. At least that's been my experience, some people put 50 factors in there and so, once you have that optimization, you're explaining, you know, 85 to 90% of what's going on that's pretty good. Okay, for a model, you're never going to predict 100% of the model and the art is not, a lot of people miss, misinterpret, what with the art is, and we put the art in science of forecasting and modeling. The art isn't about being able to use your gut feeling judgment, there's been a lot of work done out there that says, you know, rule of thumb, thumb or judgment or heuristics judgment that doesn't necessarily work so well. And, that was proven by many different research studies. It's all about understanding what are those things that are, the art comes in, is understanding what are those things that impact my product.
Charles Chase 14:05
So, for example, one of the companies I worked with prior, prior, prior to coming to SAS had a product, an Oven Cleaner product. Unfortunately, I can't give you the name of it. probably still under NDA for that I'm not sure, but just to protect everyone. Sales for this oven cleaner was declining. So, the product manager and I was in marketing. I told the product managers, before you make a judgment call, come to my team and let's see if we can model it and find the data before you do that, because that's more accurate. So they came to me and said, I think there is a correlation between self-cleaning ovens in this oven cleaning product that we sell, and it's negative. As we sell as people buy more self-cleaning ovens, they don't use this product, the cleaner oven anymore. makes sense to me, right. So, we went out and found that there was an Oven Association for I think was like $90 or $100, you can become a member and download all the data you want, we downloaded all the self-cleaning oven data and we added that to the models that one factor that we added to the model, increase the accuracy by 10%. And he was absolutely correct, it was having a negative impact on the brand. So, as more people bought self-cleaning ovens, the less they use this product now because I was in marketing, that led to some market research that they decided when they found out that most people didn't like the self-cleaning oven clean, a self-cleaning oven feature as much as they thought even though they use it and the reason for that is it takes four hours, it heats up the kitchen. And it smells really bad because it literally burns off.
Michael LeBlanc 16:37
Yeah, yeah.
Charles Chase 16:37
You know, so you have to turn it on, open your windows and go to the mall shopping. So, that led, that led to new positioning on that, these, that product and oven cleaning product. And what they did is that use our oven cleaning product in between stuff cleaning oven events, to spot clean your oven, so you don't have to use your self-cleaning oven features so much. And guess what, that turned around the brand and lead to positive growth for that, for that particular product. So, when you combine this marketing efforts, you combine predictive analytics, you go after the data that's responsible for causing these challenges and by the way, we don't have problems, right, just challenges and opportunities. So, the challenge was to find out what's having a negative impact on this product, how can we turn that product around, and that was challenging, and we were able to do that. You wouldn't, that wouldn't have happened if I was reporting upstream and in the supply chain and operations planning because all they care about is costs and reducing costs and meeting demand with the most efficient supply response and customer service and all those are good, don't get me wrong. But it wouldn't have led to all the things I just mentioned about this oven cleaning story. True story, by the way.
Michael LeBlanc 17:48
Yeah, yeah. It's interesting. And it's, it highlights the subjective versus the analytical, you know, the two need to work in concert together. You know, last, last question on this. I mean, it's been 18 months of unusual events. And I've talked to other folks who are, you know, retailers are, like, trying to figure out how their sales are versus 2019 versus 2020. Because '20, the entire year was such an outlier. How are you guys thinking about building that into the models, I mean, where it's been 18 months of outlier edge events, not single events is that something you've, again, you tweak the model, so that you kind of say maybe we need more agility in the model, maybe, again, you're talking about less, you know, less elements than more and blending last words on, on how SAS is looking at helping their customers figure this out.
Charles Chase 18:37
Yeah, you know, as we move, you know, from each phase of the pandemic, to finally, hopefully, to the pandemic going away or subsiding to a point where everyone would get a normal flu shot every year with the with COVID you know, supplement to it, our booster, as we say, and we would still keep that data in the models and as it declines the those factors that are influencing they will decline as well. But believe me epidemiological data is Google Trends will change and we can leave those in the model until we get to a point where we don't, we're we only need maybe three years five years of history and it's outside that historical range then it would just drop off and then we would bring in other factors so you're always bringing in new factors it's not something we do at one time and then the model is perfect and it's also not a model I love it when people say Charlie what's the, the model you use for all our products.
Charles Chase 19:36
I go there’s.
Michael LeBlanc 19:36
We build models, your product build models, right, is that an appropriate characterization?
Charles Chase 19:40
Yeah, our product, our product, our product uses first generation AI, which is, most people referred to as an expert system and it's based on a lot of different rules as well. And it automatically builds the models for you, very well, by the way, and it's interesting because I'll go in and I'll try to beat the system. Build my own models because I'm an econometrician. And If I do beat it, it's only by a few percentage points, which is pretty amazing, because
Michael LeBlanc 20:05
And a few more, a few more hours of work behind those few percentage points.
Charles Chase 20:10
Yeah, and the whole idea is to let the analytics do all the heavy lifting and only make adjustments. Based on a on, one, I call it low touch forecasting, or on an exception basis. So, in other words, instead of a demand planner touching every single product, but every single forecast inside, which is probably almost impossible using Excel, because you have millions of forecasts, but if you look at the hierarchy of a company, and we also forecast on a hierarchy, our system, which is proven to be more accurate, but you want them only to touch those other 10 to 15% of the products that you can't forecast because you either don't have enough history, or you don't have the factors are there's a lot of sparse data where a lot of zeros in the data, that's where you want them to focus you want them to focus on, on the other products that the system outperforms them almost 100% of the time with analytics. And, and that's really our, our objective. So, it's not uncommon for our customers to start out before they use as touching 70% of their products, because their analytical technology and their legacy system is really basic to going to the opposite, where they accept 70% of the products that are being forecasted by our statistical engine and only touching 30%. In many cases over time, they're all down at 10%.
Michael LeBlanc 21:33
Lighten, light, lightens the workload. You mentioned a couple times, we've mentioned a couple times. AI, ML, machine learning and you talked about it extensively in the book. You know, for everyone, let's make sure everyone understands what AI and ML are, from your perspective versus clever programming. Like what, where are we on a scale of one to 10, you know, one being really just fancy programming and 10 being you know, this is world changing, kind of, you know, mission true machine learning? How do you think of the definition between the two, because sometimes it's a little blurry when people talk about it, you know, AI is really just good programming sometimes. How do you think about it, and how much impact do you think it has today and will have in the years to come?
Charles Chase 22:19
Great question. Well, first of all, I, when I, most people think AI is going to solve world hunger, and that the machines are going to take over the world and I tell them that only happens in the movies with Arnold Schwarzenegger.
Michael LeBlanc 22:32
Judgment Day, yeah, we're just waiting for judgement day.
Charles Chase 22:35
Yeah, is that, there's always going to be collaboration between people and people, and people in machines. So, it's never going to go away, it's going to be minimized. So, the machines are going to do more of the heavy, heavy workload working. So, let me explain what I mean. So, what you need to do is first take the letters AI and flip them to IA. It's really about Intelligent Automation, supported by machine learning. So, what does that mean, that means we want to automate all the mundane non-valuated work that people do in this case, demand planners in our Excel spreadsheets like cleansing the data, manipulating the data, that can be completely automated, and then applying machine learning to make them smarter. So, one of the things that we could do at SAS, we have something called assisted demand planning that's patented, where it actually learns over time from the demand planners, adjustments, manual adjustments, and determines when they are adding value when they make an adjustment to the statistical forecast and where they're not adding value. And where the statistical forecast is always more accurate, then tells them where to make adjustments in the product hierarchy, where not to what direction either raise or lower, and stay within a range. And we found that that actually reduces the time it takes the demand planner to finalize a forecast by as much as 50% and improves their accuracy, their bet their value add by anywhere between 6.3 and 9.2%. You can see I'm an econometrician.
Michael LeBlanc 24:01
Yeah, you're very precise. Yeah.
Michael LeBlanc 24:03
Yeah.
Charles Chase 24:04
And, we've applied this to companies very effectively. And each time we applied it, it gets more and more accurate. So, you think about, think about it that way. There's, it's really about Intelligent Automation. And, by the way, machine learning models are not, have been proven not to be 100% effective across every situation. In fact, sometimes basic predictive models are just as accurate.
Michael LeBlanc 24:27
Right, right.
Charles Chase 24:28
And however, but see people, they go after one thing instead of saying it's really a toolbox. So, you have to think about analytics as a toolbox of different types of modeling and analytics and you want to apply the right toolbox to the right situation. So, I always talk about mechanic, let's say you take your car to an auto mechanic. What is the first thing they do to it, they plug it in to your on one onboard computer to a giant diagnostics machine. Okay, and it goes around and a diagnose, diagnose is the problem and it tells them exactly where the problem is, and they know exactly what part to replace, but now they have a whole set of tools that they have to use, they don't just use one tool. So, if they had a hammer, and how would they fix that part, if they only had a hammer, they have all kinds of tools that will help them get to that part, take it out, replace it in and put everything back, they don't just use a hammer and, and they haven't diagnostics machine that diagnosis, the situation that uses predictive analytics for the most part, to tell them where the potential problem is. So, instead of replacing a very expensive part, it may only be a very inexpensive part.
Charles Chase 25:37
So, you have to look at applying analytics that way across all your challenges, all the situations that you're looking at, to apply some cases, you don't need to apply machine learning. A typical predictive analytic model that's not a machine learning model that doesn't have a learning algorithm will be just as good. And a lot of research has been done recently by a gentleman named Spiers Macrodactylus, out of Greece, and he's done his M-1 through M-6 competitions, starting in early '80s, up until about a year, two years ago, and the last one, two years ago, found that when you combine a machine learning model with a traditional predictive model, or not even a what we call a time series model, it looks at trends and seasonality. Those models, which we call ensemble models, outperform all the other models. And if you use just a machine learning algorithm by itself that usually fails. So, it's also about combining models as well, that's something new that we're learning.
Michael LeBlanc 26:38
Wow. So, this this symphony of math basically, that you put in the hands of users to, kind of, conduct and pull tools out of toolkits and all these things, when do you think or are we at the point where it all comes together, and you could get to something like predictive retail? So, I know before any demand happens, what the demand is going to be, and even taken to its farthest reach, I'm going to send that product to the consumer is based on everything I know, and the models, and AI and IA and machine learning. I know they're gonna want that item, are we, Is that a is that science fiction or is that a possible future?
Charles Chase 27:17
No, that's, that's possible. Anticipatory analytics. We call that and talk with that quite a bit in the book. Where a good example is today, we can forecast and do analytics at the edge. So, what do I mean by that? I talked about that also last chapter. So, you can be a large retailer of beacons in the store, for example, you can triangulate those beacons and you can collect, collect data at the, at that point, at the edge in the store, using something called Event Stream Processing. And combining our machine learning, you can start to uncover patterns and make decisions right there in the store, send a signal back to replenish the shelf, send the order back to the retailer warehouse and even go further only back to the manufacturer but let's, let's start from an angle. Let's go forward a little bit. So, I'm, I'm Charlie Chase I, I like European style suits, the size regular 40. I like black, blue, brown, and gray. So, I walk into a large, big box retailer that sells suits. And I opted into their program online. And I have my profile, they've been tracking my purchases, wherever I travel. I live in Raleigh, I happen to be in New York. So, I go to one of their stores on the east side. And the minute I walk in the store, my phone is, my iPhone is on and it connects, automatically, those beacons to me and says, hey, Charlie Chase is in the store and he's walking, walking toward men's suits in the men's department. So, before I even get halfway to the men's department, I'm getting pinged with promotions and we have European style cut suits here with different brands in your size 40 in blue, and black. And, we also have gray and brown on our website store. Would you like us to wrap those up for you and send into your house? Now, how cool is that, I'm going to a men's suits and it's, by the way, it's saying and by the, it's given me a promotion saying it's 30% off today.
Michael LeBlanc 29:20
Right.
Charles Chase 29:20
For those same suits. So, I'm getting a coupon in the store in my, on my iPhone, and I don't have to do anything. All they do is say yes, I'm going to ship them to my house.
Michael LeBlanc 29:30
It's an interesting example because we used to struggle sometimes with how to detect behavior that's going to happen anyway and behavior that we need to change with an incentive. So, you describe for example, that in that instance, I'm going to get 30%. Maybe you are going there because you're getting that suit anyway and I didn't need to offer you an incentive. I just needed to speak to you in a very relevant way. Any, any thoughts on that, I mean, I've struggled with that and in different roles. I know a lot of retailers, kind of, figure out, what's the difference between motivating changing behavior and allowing organic behavior to run its course with some clever communications?
Charles Chase 30:11
Well, we have a product called CI 360 that allows you to track your customers online, on their, on their mobile device, as they look at and how they search online for products, and then you could start to remarket your, and read, and change the way you're marketing to those consumers, because you're capturing that information in real time and that's another way, that's something different than forecasting and planning.
Michael LeBlanc 30:41
Yeah, yeah.
Charles Chase 30:42
But it can complement that as well.
Michael LeBlanc 30:44
Yeah.
Charles Chase 30:45
A good example is when we launch a new product, we can use text mining and sentiment analysis, once we forecast the new product launch, let's say we're in week three or four of that launch, we can actually start to crawl out on the internet and see what people are saying about our product on Facebook, on Twitter, and other vehicles and determine whether they liked the product or they liked the promotion, did they find it on shelf, and also capture what you're saying about I wish it came in this color, or this flavor. Now, you're capturing sustainability information about how to come out with line extensions in the future with different colors and flavors. And then if they like the promotion, and it's, and your, and now you're in week six, or seven, and you start to see sales start to wane and then a little bit of a decline, you can then run that promotion again. And if they can't find that on shelf, you can reach out to your sales team and say, Hey, guys, how come this is not on shelf, you need to talk to retailers get this product on shelf. People like it. They go to marketing and say, hey, you should start looking at developing different flavors, different colors for this product, different package styles, because this is information, we're capturing in real time using digital technology.
Michael LeBlanc 31:57
Right, right. It's very interesting. All right, last question for you. What are you hearing from the people you talk to you travel the world, either real, in real airplanes, or virtually, I'm sure, talking to people, what are the top three things that retailers are asking that you'd want to share with the listeners, kind of, top three questions answered top three tips? You know, however you want to approach it, but if you had three things to say to the audience about tips, other than I guess, number one, buy this great book, what would, what would two and three be?
Charles Chase 32:29
Well, I think that, you know, all this surveys and responses I get from retailers and in large consumer packaged goods companies. Its how can we improve our ability to forecast demand in this digital economy? And also, how can we handle disruptions in the future, like COVID, and then the second question, follow up question is always, how, can we, what data do we need, because you can't really do a good analytical modeling without good data. So, one of the things we learned at SAS, you know, 45 years ago, is that in order to apply analytics, you have to have good data. So, we've always been able to provide access data, in almost any format today. So, for example, we can access a PDF data, Tara data, any kind of data in any format, pretty much we can access that data, normalize it, harmonize it. It's stored either on a data lake, on demand signal repository, enterprise data warehouse, and then run the analytics. And we can even run it in memory, in memory, in real time, using you know, like I said, at the edge. So, there's, so, that's the second question.
Charles Chase 33:39
And then the third question, I guess, would be, you know, how do I position is with my executive team, to get their buy in, because the biggest challenge that companies have, when they try to move from a non-analytics-driven organization, to an analytics-driven organization has their own corporate culture.
Michael LeBlanc 33:57
Yeah, it's not the, not the tech, right. It's not the vendor base. It's, its culture, right?
Charles Chase 34:01
Yeah. They think is culture.
Michael LeBlanc 34:03
Giving up control, basically, to, to not even control but just that perspective, right. It's, it's gut instinct versus analytics a little bit, that's probably at the extremes of both sides. But that's interesting. That's often really both the hurdle and the enabler, I guess, at the same time, right?
Charles Chase 34:20
Yeah. And it still requires not only a champion, but a champion who reports into an executive sponsor, who's in the C level suite, in order for that change to occur. And what happens is, in many cases, the change does occur, and the focus has to be around changing people's skills, implementing horizontal processes, because most processes in companies are vertically aligned, not horizontally. So, they're not aligned across organizations like sales and marketing use different data. They have different performance metrics from the supply chain. They don't, hardly talk to each other. And, that's the reason why a lot of these processes like SNOP and integrate business planning fail because they're not horizontally connected and sharing performance metrics. And then it requires scalable technology and data.
Charles Chase 35:10
So, you know, and then even when they get adoption, usually what happens when they get adoption, a champion gets promoted or moves on to another project. And everyone falls back to the old way of doing things. So, it's not sustainable, didn't stick. So, in order for it to stick, you need to replace that champion. Until it does stick, it may take only six months, or may take two years, when it came to forecast, and when I was a Director of Forecasting and Planning, I was that champion, and I reported to a Senior Vice President and in one case, actually important to the CEO and President. So, I carried a lot of clout, and I was there permanently. So, I was the permanent champion. So, not only did we get adoption, but it was sustainable. So, you have to remember, you can always get adoption. But is it sustainable? Are they going to continue to do it after that champion leaves and use the technology?
Michael LeBlanc 36:06
Well, my guest is Charlie Chase. The book is Consumption Based Forecasting and Planning. The company is SAS, Charlie, how do folks get in touch with you or learn more, where do they go?
Charles Chase 36:15
I mean, well, they can, they can email me, I think you have my email and charlie.chase@sas.com. I'm also out on LinkedIn, I have over 40-, 4200 followers on LinkedIn. And I have several 100 on Twitter as well. So, I'm on those two platforms. But the easiest way is to send me an email or message me through LinkedIn.
Michael LeBlanc 36:37
Very good and is there any resources if we went to SAS comm over and above the book any, and you put up thought leadership papers up there, is anything we can go to of that sort?
Charles Chase 36:46
I also recommend to have a blog. I have a SAS blog site and author page too, so, you can, a lot of the blog articles that I write end up being chapters in books.
Michael LeBlanc 36:56
Right.
Charles Chase 36:57
And then I, I was writing a column, a column in a journal business forecast, a quarterly column, but I had to stop because they got a little busy. But now those are other options. And also, I'm going to Predictive Plan Demand. Under retail and, and consumer products on the SAS website. Once again, throughout the SAS website look for, under retail, and consumer products, goods, and then under predict and plan for demanding and find a lot of white papers. There are many of those white papers I've written or co-authored with a lot of my colleagues, but and then also other information, customer success stories, videos, and others. Yeah.
Michael LeBlanc 37:38
Well, all right, well, it's a comprehensive, good discussion, comprehensive set of data, so to speak, both that those resources which I'll put in the show notes on the book, and listen, thanks for joining me on The Voice of Retail, great discussion, fascinating work, and certainly topical each and every day. So, Charlie Once again, thanks for joining me on The Voice of Retail podcast.
Charles Chase 37:59
Thank you for inviting me. It was great having a good time; I enjoyed our conversation.
Michael LeBlanc 38:04
Thanks for tuning in to today's episode of The Voice of Retail. Be sure to follow the podcast on Apple, Spotify, or wherever you enjoy podcasts so you don't miss out on the latest episodes, industry news, and insights. If you enjoyed this episode, please consider leaving a rating and review as it really helps us grow so that we continue to get amazing guests onto the show.
Michael LeBlanc 38:23
I'm your host Michael LeBlanc, President of M.E. LeBlanc & Company Inc. and if you're looking for more content or want to chat, follow me on LinkedIn or visit my website at meleblanc.co
Michael LeBlanc 38:32
Until next time, stay safe and have a great week!
SUMMARY KEYWORDS
model, data, sas, demand, analytics, product, people, forecasting, forecast, outliers, predictive analytics, machine learning, book, retail, marketing, suits, store, retailers, planning, learning