[0:02] Lisa Maloy:
Okay, thank you for joining us this morning. I’m first going to talk through a little bit about Hershey. We’re over 70 countries in the world. And you can see a few about her achievement that Michele Buck was our first female CEO. Milton Hershey was the founder of our company.
[0:30] Lisa Maloy:
Recently we celebrated our 120 year anniversary and not here she founded the non-commercial educational ever, the largest philanthropist.
[0:48] Lisa Maloy:
Slides here to all that we are a billion-dollar company.
[1:01] Lisa Maloy:
And a few of our products on here again a couple of years old candy we moved into another snack area we have a few additional snacks we’ve acquired recently like parents beauty and remember the other popcorn skinny.
[1:35] Angela Stewart:
Good morning. So I’m going to take you through our Cobra automation process.
[1:40] Angela Stewart:
As you can see from the slide or Sri Lanka deduction, the number of deductions are increasing across all industries. Crushy is no different. So this means for all of us that there’s a loss in productivity and efficiency. And while the number of deductions is it Increasing we’re unable to increase headcount. So, currently, we have in you’re probably all in the same boat, performance deductions are our top deduction category. And Kroger is the highest performance doctor that we have. So there are over 42,000 deductions creating the year for poker performance alone. And to break this down even further, there are about 820 deductions created a week for Brokers. There’s one deduction analyst that handles this account. We do not as I stated had the option to increase bodies, so we haven’t found another solution.
[2:48] Angela Stewart:
So for the current process, Kroger is a CPA customer so we already gather important data and it’s copy related to our HighRadius system. And using this information, critical components were the deal ID field which has windows brokers offer number or contract number. And we also use information from the comments section which is populated by the spamming type, which is Scott scanner Pat’s.
[3:25] Angela Stewart:
Okay, so once we are also in our system, we also have our trade promotion system, the information being pulled into HighRadius. So we also have the offer number and spend time on the promotion. So we worked with HighRadius on a script to create a script, so that we are able to use those fields to match the deduction with the promotion.
[3:52] Angela Stewart:
And once when there’s a perfect match, we have successful automation and the credit is processed overnight, and an auto-match is created and offset. And we are, there’s no manual touch, everything is automatic. So we have found and I want to make sure everyone understands that we do a check on this. So quarterly we have a sales analyst that reviews all of the promotions to make sure everything is working properly with automation. This is not something that we’re doing in our group at this time. But we do have a check and balance in place.
[4:33] Angela Stewart:
So we have found that we have been able to reduce the number of Kroger deductions that the manual touches by 70% and the dollar value has been reduced by 40%. This is pretty significant for our company, saving a lot of time. I want to say it saves four full days, at least a month.
[4:56] Angela Stewart:
Because it’s been so successful, we have decided to start incorporating more customers and our next group of five is going to be ready to go by the end of q1.
[5:14] Lisa Maloy:
So Bob resolution, I think we’ve all heard HighRadius it’s been into this territory with resolving from the worklist, or resolution. So we are, we will be implementing this in q1. We’re working with them right now just to fine-tune the process and the screens. But this again will allow us to select more than one reduction from the worklist and allow us to resolve it against one or more permissions.
[5:52] Lisa Maloy:
So what’s next for Hershey’s so you know, I’m very interested in the AI capabilities especially with the Trade Commission area, they’ve started reviewing some of our data to get us started and we’re looking forward to what that is going to bring us as far as those additional matches and ways to validate through the pre deduction process.
[6:36] Lisa Maloy:
And I’d like to introduce Vishal.
[6:41] Angela Stewart:
He is going to talk more to the AI feature.
Thanks a lot. First of all, good morning, everyone.
So just for the audience to meditate if you don’t know me, I am Vishal and I work as a senior product manager with a deductions cloud solution and HighRadius.
So today we’re going to be talking about a very exciting area that we are working on. In the deductions cloud space. We are working on a solution that leverages the AI and machine learning algorithms to help match create deductions or retail deductions with promotions automatically. And which will help you overall settle your trade deductions much faster?
She is partnering with us on working on the solution, they have been very helpful in providing Business Insights and data insight so that we can work on actual data to bring this solution.
Thanks for that.
Before we dive into the specifics of this solution, and how it actually works, I would just like to take a step back at the back to what Angela actually said. The one limbs of trade deductions are increasing and incomes can be linearly increasing, along with deductions, and as she mentioned, the retail trade deductions are actually a major problem. And let’s just try to understand why is it a major problem? As soon as a reduction comes in along with a claim item, you need to go to every detail at a line item level from the claim and then try to figure out what is the matching promotion, from your TCMT, the Trade Commission Matching Technology. With a growing volume of deductions inline items within the claims, this becomes an increasing problem in terms of the efforts that you have to put in to find these promotions, etc.
So in order to address this problem, how do we address this problem without increasing the headcount?
Has been working with us to automate proven deductions. Has been one of the top customers was a lot of trade deductions and we were able to automate almost 70% of the trade deductions and were able to reduce manual efforts. And we are also looking at five additional customers to be added to this automation. The overall theme that we are looking at is, we want to use automation to drive overall efficiency gains because of which you can address all the increasing volume reductions without actually increasing the headcount. So, really working on global automation, we work very closely and analyze tons of data and the processes and the things that vary by broker to figure out the business rules that actually would go in. And imagine the deduction of claims with the promotions.
That’s a very major exercise because you have to scan through a lot of data in a lot of processes. And it’s a lot of manual effort. And another potential problem with that is you would have to constantly track the business set up in the system to see whether they’re performing well or not. Also, these are some of the potential issues with the auto-matching engine that we have performing very well. We feel that using, leveraging the power of AI and machine learning, we can push these boundaries further. So with that, I would just like to introduce this slide, where we will talk about, what exactly are we doing? How exactly are we leveraging machine learning and AI to help you match your deductions or claims of these mutual promotions much faster?
So what we saw actually was that on a day to day basis, the analysts are actually creating a lot of resolution. And we see those resolutions or that historic data as important where we can actually leverage machine learning algorithms to identify certain patterns from the resolutions and use them for future auto-match. So just to give an example, let’s see last month, 100 deductions or hundred claims were actually matched with promotions and out of that hundred let’s say 90 of them, for 90 of them the deal number matched with the repaid number, just giving you a specific example. So that you can get a better understanding or the date on which the claim was arranged was within the performance. So, these kinds of patterns are identified by the machine learning algorithms and captured within the models so that they can be used for future deduction. And now when it comes to benefits, it addresses the problems that I raised earlier with the current auto-matching in it. One is that you would not have to actually scan through the data and the process since manually, the AI will actually scan all the data for you within the data for you. And then you can just review whether the business rules that it has identified should be used or not. Right. So that kind of helps you expertise the whole process and go live with more auto-matching in a much faster and easier way. And the second benefit of this auto-matching, AI-based auto-matching engine is that he has said, right, so you do not have to do periodic reviews of the business rules that we’re setting up. as and when you create your resolutions as it when new data patterns emerge, these algorithms will pick those up. If a particular data pattern has become gone straight then the algorithm random rules or conditions. In this way, you would have a sustained performance of the auto-matching and it would not drop the internal and data pattern change. So we’ll just like you to focus on the outcome and not on the nitty-gritty of what’s exactly happening right. So, we will just want you to track how much automation you are getting and with the auto-matching you are getting is accurate or not, rather than spending time in the granular, data level.
So, with that overall solution architecture design, I would like to draw your attention to the main differences between the traditional auto-matching engine and the AI-based auto-matching engine.
So the current business rules that are there to automate the proper reductions of lanes. actually consider some of the attributes.
Great, what was that the algorithm was able to pick up some additional attributes or relationships from the data. So, the first major difference that we saw was that he was able to identify certain date-related relationships and also patterns.
So, just to give an example of deep related relationships, what we found out was that whenever in most of the resolutions, the posting date of the reduction was between the performance videos to the posting date would be great and then and then the commitment started. And it will be less than the commitment.
We also saw that the machines picked up certain relationships between the number and the numbers. So, there is an underlying mapping or direct mapping data being numbers are generally set and against these numbers, which it was able to pick up and math and use it for the future.
So, these are some attributes, very difficult to pick up to manual analysis, which is where we can leverage the power of AI and machine learning to help drive more automation and more accuracy.
So what we’re seeing over here is even within global markets, 70% of the population that they’re experiencing there is for the school board potential to push these boundaries to achieve more and-
Moving on to the second piece of information I wanted to touch on touch upon the mountain of customers. So as I said earlier, setting up this automatic engine setting of a business rule is very time consuming and tedious and I’m sure you spend a lot of time in sort of centerpieces and drag them on a regular basis.
So we just look at certain volumes from the actual data as Angela mentioned earlier, Is the top-performing customer when it comes to retail reductions and which actually accounts for 40% of the traditional customers that we are looking to automate would comprise of close to 12 1010 to 12% of reductions. Again this is based on the last six months of data.
So, this can be, however, it will give you an idea of a ballpark number. So, the main piece that I would like to highlight is doctors.
Practically not feasible and to track or do a periodic review of these rules because of the sheer volume of customers, because these rules are set up by our customers basically.
So, the sheer volume of these people will not actually allow you to not be practically feasible for you to set up auto-matching rules manually. And this area is where we see a lot of potential of using AI to do the research on your behalf, identify the underlying patterns and use them for quantum mastery. So we are expecting some automation even from the companies that will definitely add on to the consumers that we have seen in the current date.
Yeah, with that, I would like to open up for questions and answers and I’ll invite Angela and Lisa on the stage.
Would you say your name and company? That’d be great.
Hi, Jessica Butler with Attain consultant group. Quick question. When you say today with your rules where you’re matching, let’s say the offer number or the deal number from Kroger, with that on your commitment. Are your salespeople actually putting the offer number on the trade promotion system? And if so, have they always been amazing or has that changed management and you got them to do it for this?
[22:46] Angela Stewart:
With Kroger. We have a sales analyst that works with our group. There. We’re very lucky to have her on that team. So she does too and she has always provided me with the deal information. It just hasn’t always been populated in the system. Since we went live with HighRadius, they have started doing that. And we went live with a new promotion system as well. She has started doing that. But I can tell you that for the five other customers that we’re going to onboard in q1 that our sales team has partnered with us on that. So the upper management is strongly encouraging the CSCs to enter that information and we have seen some very good results.
That’s great because that’s the magic. Yes, I know that the magic key Yeah, even some people we’ve seen for those whose salespeople are not quite as accommodating people who even once it comes into the HighRadius system, there’s a lot of value in having someone on any team going getting those deal sheets and entering that information because you get so many deductions for each kind of deal. So enter it once in the system, and then her ash but-
[23:55] Angela Stewart:
Yes, it is change management for our other customers.
Wallace from Conagra brands. The quick question: Maybe two questions. One is your integration into your TPM solution if you could talk a little bit about that process and how that worked. And I assume your other five customers how it sounds like you’re going customer by customer before you get to the 48%. Is it because their portal customers the backup is easier to get I was just interested in understanding the dynamics of how you road mapped out which customers you’re going to go with, as we may be considered a similar journey or anything then.
[24:43] Lisa Maloy:
So the TPM data, we have that coming into my radius daily. So there’s a feed we have we use bi as well. There’s a feed from bi for header and material level data. So that’s how we get that rich data in there.
[25:03] Angela Stewart:
I can speak to the five additional customers. So the reason we chose the next five is their large deducting customers, but also because we have the cooperation from the CSCs to input the information into our trade system. So we’re kind of starting with the easy ones first we’re doing the most impact but actually the easiest for us to accommodate but we do have plans to continue it’s not we’re not just going to stop at six customers we are going to go further on to and where we’re using our highest deducting customers performance, adopting customers and picking from that pool. And there’s oh yes and important. They have CPA information there. They’re already using CP. So all that information is already being populated. And very important. Yeah.
Hi, this is Leslie de Leon from Shell. So how long has been AI able to self learn? What was the average learning curve for the automatic application?
Okay. So, this is just to summarize the question, your question is like, how long does it take for AI to actually get into action and start suggesting commitments? Okay. So, first of all, I would answer to that in a way that it actually depends, because for the machine learning models to learn, you will have to have for every customer you would have, have to have actually at least a critic created a critical mass of deductions from which you can actually make some deduction and you can reduce some patterns and then apply it right. So what we have typically seen is that as soon as the number of claims or deductions that crosses 500–1,000 resolutions, we are actually able to predict with good accuracy. So, I mean, it’s not mainly dependent on the time it actually depends on the volume of resolutions for every customer so as soon as you cross 500-1,000 we have seen really good accuracies within the production’s over that answers your question.
Any more questions? Okay.
So from Nestle. What are the modules of HighRadius you guys are currently using in Hershey?
[27:53] Lisa Maloy:
We’re using the cash application. Then CPA which we’ve been on for quite a while cash application we went 2015 and then deduction last year.
Hi, good morning, Andrew Boris with Accenture. Question for you about how the conversations went with your controller or your risk and quality team about using AI to do auto-matching. Can you talk a little bit about how the journey was to get people on board with allowing you to be able to use or leverage AI?
[28:41] Lisa Maloy:
So we have not had a conversation with our controller team on the AI we’re in the process of working with HighRadius, just looking at the data and aggregating it to see what the capabilities are as far as the auto-matching piece. Everyone’s onboard without automation so Hershey is validating At a contract ID, and everyone is comfortable with the value if you know the accuracy of that and comfortable with that approach. I’m also sharing a little bit more about how we do trade settlements. So 40 Pm for us, we are settling into a rebate if you’re familiar with TPM, and so every resolution is one rebate at a time right now. So it gets down to the material and we allocate across all the materials automatically. So we’re not going in there and specifying a particular dollar amount for every material. So that speeds up the process a little bit for us. But everyone’s onboard uncomfortable with that piece of it. Okay, thank you.
This is Joy from Johnsonville. And I know we talked but one more question based on what you just said, Are you doing your rebate accruals based on item level, and but then you’re not settling mean you’re settling an item level but you just spreading it like butter.
[30:03] Angela Stewart:
So we do basic cool across materials either on the shipment or forecasted data. But we are basically when we do an allocation it is spreading across the materials to that one rebate evenly. Okay.
Okay, I think this has to be the last question.
Hi, Laura Carper from Conagra brands. So question in relation to that, how did you convince your sales team to go from Hey, I planned at maybe a promo group level which is where we’re at too, you can plan at a promo group level but we are going to spread it across.
[30:52] Lisa Maloy:
So they actually plan at what we call pack type or down to material depending on the customer. We’ve decided there’s value in that level of material data. Okay.
Okay, great. Thanks very much. I’m not going to move because I know I’m gonna sell out. Thanks so much for everybody. Thank you, Angela, Lisa.
If anyone will stick around for a little while if anyone wants to talk to us or have any more questions, okay. Okay, great.
[0:02] Lisa Maloy: Okay, thank you for joining us this morning. I'm first going to talk through a little bit about Hershey. We're over 70 countries in the world. And you can see a few about her achievement that Michele Buck was our first female CEO. Milton Hershey was the founder of our company. [0:30] Lisa Maloy: Recently we celebrated our 120 year anniversary and not here she founded the non-commercial educational ever, the largest philanthropist. [0:48] Lisa Maloy: Slides here to all that we are a billion-dollar company. [1:01] Lisa Maloy: And a few of our products on here again a couple of years old candy we moved into another snack area we have a few additional snacks we've acquired recently like parents beauty and remember the other popcorn skinny. [1:35] Angela Stewart: Good morning. So I'm going to take you through our Cobra automation process. [1:40] Angela Stewart: As you can see from the slide or Sri Lanka deduction, the number of deductions are increasing across all industries. Crushy is no different. So this means for all of us that there's a loss in productivity and efficiency. And while the number of deductions is it Increasing we're unable…
HighRadius Deductions Software acts as a powerhouse for proactive deduction management to prevent bottom-line erosion. It provides automation, process standardization, and a platform for cross-departmental and customer collaboration. It supports deduction management by providing some key features like back-up document capture which captures deduction data from customers and supplies the information required for resolution; auto-capture proofs of delivery (PODs), bills of lading (BOLs) from carrier portals & emails; structured deduction resolution, collaboration & approval workflows to streamline the communication and approval process; along with automatic deduction correspondence, and automatic data push to customer portals. The result is a proactive deduction management operation that recovers revenue normally lost to invalid deductions.