WEBINAR: MAKING EVERY TRADE PROMOTION EFFICIENT IN THE ERA OF AI

hello everyone who has joined and thank you for joining um our webinar on how to make every trade promotion efficient in the era of ai i will begin with a brief introduction of the hosts and the company my name is vlad lauda i'm the head of international growth at lithio and i'm co-hosting this webinar with my colleague helen shaponic who is a product director of inventory optimization hello helen hi hi thank you very much for having me pleasure pleasure to always have you on the events so we are continuing and uh it's important to say that this webinar is brought by leafeo and uh lithium is a global vendor of cloud native retail supply chain automation and optimization solutions and um

everything that you'll hear today comes from the background of more than 10 years in the supply chain automation and optimization in various types of retail grocery supermarket specialty and during this period we've delivered over 160 projects in over 15 countries with a team of 200 people so we were able to learn the key um principles that our potion retail companies to optimize and the majority of these optimization initiatives that we see they have similar patterns and fall into one of the three buckets either it's the need to improve operational efficiency it is the need to support the company's expansion when when the number of locations is growing or they need to meet the sales and revenue goals that are set by the board

so for these reasons companies are adopting automation tools and that will help them handle the task so lithium provides these these solutions and as a company we strive to bring innovations closer to our customers by applying the latest knowledge and research in combination with the cutting edge technologies and of course all of this is driven by high passion of our team for our customers success so the retail companies can meet their sales and margin goals they can manage growth with confident and operate at the dramatic efficiency so this it becomes available with a leafio ai platform that features inventory optimization store and shelf space optimization and promotion forecasting now we're moving on to

um to the topic of of of today's event and we're actually have dedicated events for each and every side of the supply chain operations either inventories or the uh store and store floor and shelves or it's trade promotions as we have today and why this problem or why we are why we're talking today about the trade promotions there is a number of reasons first of all the market is the demanding a good solution for managing trade promotions and we've we've been getting signals from a variety of companies including our customers and other retail companies present in the in the market that this is something that needs improvement so when our team started looking closer at this problem we've realized that their initial request that we that companies approach

us with which is um trade promotion forecasting is just the tip of the iceberg and we um we noticed that for the majority of uh businesses trade promotions are very significant portion of their operational load for a very simple reason there's been an emerging trend of um of increasing or or or of the growth uh of trade promotions in the overall revenue of retail companies depending on the vertical it there is between 20 to 70 percent but it continues to grow every year and the process of trade promotion management is a very complex and cross-functional and requires both internal and external coordination for it to uh to have to to happen and i'd like to take a closer look at

actually what's what the trade promotion process um uh comprises of when we compare the regular sales process that is typically on the annual cycle and comprises of the following stages first the assortment has to be selected then the negotiations with the suppliers have to take place then the merchandising commercial team has to prepare everything prepare the shelf space prepare the the signs the notices the purchasing has to make sure that the volume is available for sale and then the logistics has to deliver of course there is financial infrastructure behind this and and more this means that the process is fairly complex when we when we analyze the trade promotion processes they look almost identical so the assortment selecting the right excuse which is in it in itself a very

complex task to select which products need to go on the promotion of course negotiations if the trade promotion is not initiated by the supplier then the merchandising the purchasing logistics finance in the same manner after comparing this we of course see that they are pretty similar but the difference is that they're drastically different in the frequency so the the first one is the annual the second one is the weekly or bi-weekly and this is at at a comparable rate of 20 to 70 percent of course the trade promotion is much more aggressive and and creates a lot more stress for a retail organization so the the promotion strategy in itself is divided into more more stages which is the planning that takes place initially then the preparation basically everything that

we've looked at at the previous slide the execution and the exit which includes the analysis um of the efficiency so every stage in itself is um is critical because if if one stage is not at the top of the performance it means that entire cycle is in danger and maybe with very low efficiency and the very few retail companies have mastered this process in a very very good way in a in itself every stage is further divided into into uh particular steps which is working with the with skus and the the suppliers that include different departments internally from planners marketing category managers manufacturing suppliers to creating promo mechanics of the merchandising team forecasting which which uh

in itself ask the question of what skus need to go on the on the uh on the promo and um based on the volume of products in uh in the in the overall assortment this yields that there are thousands or tens of thousands products on promo simultaneously at different locations and they may be different types of promo um so these promo the types of promo may vary from the initiated the the the weekly or monthly promo the promo that is initiated by the supplier or the promotion that is uh initiated to get rid uh from overstocks so you now get the picture of course you're facing with this on a daily basis but i just wanted to bring to your attention how complex this process is

and there is no simple solution to making sure that this goes smoothly at the core of the trade promotion is the forecast that defines the trajectory for the future success if it's not accurate the company may suffer and if if there are inaccuracy it means that this if the sales grow by three times or by 10 times sometimes even 50 times this put puts a tremendous uh tremendous stress on the logistics on on the operations and if there's not enough promo uh promo products in stock uh the customers are not gonna be happy because there's there they've been aware of it right um out of stocks are completely disastrous and the the majority of uh trade promotions lead to overstock if

they did not go into out of stock so we've we've haven't explored this topic we understand that there's a huge potential for improvement the uh out of stocks cost a lot in in losses because it includes the products that are need need to take more space on the um in in the warehouse they take up the space uh maybe in the store some products go to waste so this becomes a very very big dilemma from one side the companies are in a situation where when they are hooked on the promotion because their customers are already used to it and it improves the traffic it improves the the the the volume of purchases uh so it's a on one side it's it's a needed uh it's inevitable evil on the other side it's a huge pain

and it consumes just tremendous number of resources and it's in in sense um it becomes in the addiction so i'll let my colleague helen speak on how to actually solve this problem yeah thank you so in order to achieve a higher level of the operational excellence in programmer management we offer you the following formula to have the data that allows using usage of advanced forecasting methods to increase their accuracy in order to earn more and make less mistakes using [Music] the ai in this process and the process management solution it means the efficient problem the minimum list of data for

building the forecasting methods model in the ai is presented below it is the of course the uh prototypes and problem mechanics the duration of uh uh particular promo escape that are taking part in the particular from the price during and before chrono the percentage of discount concrete locations uh and some additional layout if you do have this the category here he and of course sales during the promo and before you can download the data requirement sheet by quickly scanning this qr code we usually approach the launch of promo projects as follows if the customer do not have enough data for using advanced forecasting assets um

we use the statistical methods which are based on the history of past promo campaigns and this history data is averaged if the customer do have this complex data that that is required on the previous slide we can use more complex machine learning algorithms for forecasting in this case um of uh conditionally uh stable or well-solved goods uh with regular demand and rather deep history of similar promo campaigns the statistical masses can also work well but unfortunately if we are talking about some new products or products that have never participated in promo before or some rarely sold goods their accuracy of statistical forecasting masses uh is

significantly significantly lower it is important to have a clear and strike structured process uh process uh the next slide please yeah thank you as vladimir said uh it is very important to have a clear and structured process with interaction of all departments that are involved in promo process including the operationals commercial department logistics marketing departments often in projects we are faced with the lack of tool for effective communication in this internal communication i mean for managing promo campaigns that may lead to a lot of miscommunication and as a consequence to a decrease of the effectiveness of the promo due to

errors that may arise during the preparation of promo and the promo implementation as we can say this if we consider the promo process from the moments of its initiation to the prom to the moment of uh effectiveness evaluation it can be simplified as follows first of all it is necessary to define the promo strategy uh for the company and on basis of which a calendar of promo campaigns is created usually such a calendar is created for a year without any specific details of specific sqs that will take part in uh concrete problems uh when the times of when the time of a particular promo comes an operational process begins and this operational process consists of four following phases

the first phase it is the planning phase when um we need to create the promo assortment uh we need to define the concrete mechanics and prototypes for each promo and their conditions uh must be agreed with the suppliers and the very important things the sales forecast of future promo must be defined uh the next step uh is the preparational phase uh where it is necessary to organize uh everything develop g6 to provide the stores with the promo goods to organize the storage of the increased quantity on the level of the distributional center in case the goods are being delivered through the dc to provide the promo merchandising layout to ensure promotional price tags uh one day before the prom campaigns

should start on the third stage of the process it is the process of execution it is necessary to flexibly take into account their current sales situation and uh as of course they may really differ from the forecasts from the planned things and it is important to react really quickly to possible deviations from the plan and as a result of the complexion of the drama there are several important tasks to evaluate [Music] the achievement of the goal that was set before the particular promo started to evaluate the performance indicators of promo to calculate and make all the settlements with the suppliers and of course make some decisions on what to do with possible overstocks after the problem

that's why we strictly recommend using professional tools to organize the problem management process because it is really a complex thing one of which you can see on the screen on the slides you can see the ui for leafy problem management solution with the help of which retailers gain transparent promo process visible for all departments and better forecast accuracy for promo sales it is a very good question how the retail companies can measure the profit management efficiency you know different retail companies that we face with use different metrics uh to evaluate the effectiveness of promo uh campaigns unfortunately uh many companies still do

not distinguish their pro sales and their regular sales and the effectiveness of promo based on some own criteria uh most often we receive a very good question from the retailers uh about the forecast accuracy that we can provide with the help of our solutions in order to compare this accuracy with their current accuracy they do have but it is not so easy to answer such a question without understanding what methodology is used for such calculations and what is the current accuracy level to figure out uh we can how this accuracy can be calculated on which levels it could be calculated um we are having the slides so their accuracy can be calculated on the uh enterprise level uh we mean the whole retail chain level on the category level

uh on the asking level and the level of the s3 location drill drilling down from the highest level to the lowest level in our practice we can calculate the accuracy on different levels and it really depends on what goal we do have with our customer for example if accuracy is needed to for evaluation uh for example of the effectiveness of replenishment to the distributional center of course in this case we will use the level of the sq if the accuracy is needed to evaluate the replenishment of the store we of course will use the sq location level in some cases we need to control the indicator of the accuracy and it is requested by the company's management and but in this case it is really important to understand uh at what level

it is really required the accuracy can be also calculated uh with the using of four different indicators uh mean absolute error it is the most simple indicator and it it is just a simple difference between the fact of the plan uh mean absolute percentage or uh it means it measures the percentage or from the forecast in relation to the actual values um you know it this uh assumes no preference between uh what day or what products to uh was predicted better it is commonly used uh to measure forecast errors but it can be deceiving when sales are very low and reach numbers close to zero um so rarely it is really it is really used

but in not all cases uh weighted average percentage sure it weights the error by adding the total sales here and uh the metrics we love more it is the weighted mean absolute percentage or is used when the use case requires to priority in certain sales it gives weight to the prioritized item that uh basis uh their prediction errors covers it uh to uh evaluate the accuracy we use the last indicator it is the weighted mean absolute percentage order uh which to rebalance on the purchase price of each products since an error of calculation for the expensive goods cost the company uh much more than than ever for good with lower purchase price

but the accuracy indicator will not give you the very good understanding on whether you have achieved the business goals uh that were set for the promo or not as well as what consequences you have in terms of the inventory level that's why we recommend evaluating the final financial result with the help of the following indicators if the purpose of the promo is to increase the stored traffic and as a result sales it is necessary to ovulate their sales plan we see facts if there are difference and the plan sales are higher first of all you need to pay attention on the product service level during the promo and if it wasn't uh lower than 99 we should use one of the methods uh of forecast world aurora calculations that i have mentioned before

uh if the plan sales uh girl over uh we also recommend evolating the error and immediately analyze their overstocks and make some decisions about the overstocks if the purpose of chroma campaign is to increase the margin we should evaluate sales but also the margin plan this effect in terms of the purchase price and their supplier base if the result of promo sales doesn't meet business goals uh we should pay attention and deeply analyze goods availability for the first day of chroma at the very beginning of comma uh and during all promo period uh there are post-drama overstocks uh of course we should calculate uh what percentage of the purchase goods uh remains and uh some made decisions whether to keep it and uh

maybe some prolongation of promo campaign uh or something like that or make decision on returning uh sales goods to the suppliers uh and the uh last thing it an hour which should determine how many days of sales we have the current stock balance uh okay so um making a long story short here we have an example of how company can improve the results of uh promotional activities uh there are a result of our customer a grocery retail chain that consists of 90 locations and operating uh 20 000 tesco used in average uh the percentage of revenue that is being done during the promo is very high there's more than 40 percent

uh this company uses different prototypes different chromo mechanics and before the start of our project managers were calculating the promo forecasts manually at the first stage we received the data clintus and built an ai model using different uh machine learning algorithms after that we have started the piloting project uh where uh we took just a couple of product categories and handle promo for this part of esca use with the help of the ai forecasts but in real life as we got really good results we proceeded with rolling out to all promo activities of this retailer as a result uh we have the

increasing of the product availability during the promo uh 42 and just now it is uh 99 and the overstocks decreased by two times uh i mean the after promo over stocks uh it was a very good result for the company and vlad what can you say about the overall business values that our customers could get after the implementation yes uh thank you helen uh that was good thank you for a detailed overview uh on the measurements this is very important something you can't measure you cannot improve and then the the example yes the the the latest example that um we've had with what with one of our customers so that was just one episode overall there is a potential

uh for improvement that is uh even higher so using the tools that leverage the power of machine learning can bring up to 20 improvement in availability uh during the promo and up to 50 less over stocks so let me bring you back just to to to to the problems that we were discussing at the at the beginning the two challenges if there are no products in stock it's a huge disaster if there are too much product after it is also a disaster so the ai helps to solve for that problem additionally there's less time required during the planning stage and execution stage just because the majority of the calculations are automated the trade promotion history is centralized and this is really um a very very uh important point because when the

history lives in different places when it resides in emails with with the suppliers internal chats um some erp system and it's not accumulated in a single place it's very hard to analyze and find points for improvement so centralizing all of the history is is crucial and then uh the advantage that is bringing the adventure the advantage that is brought by the machine learning is the continuous improvement so the way the machine learning works is that it it it continues to look for for patterns for points to make the process more efficient uh with every new trade promotion so there's always when things are done differently there's always a chance to get just a little bit better and all of this is possible with with the machine learning models

i'm sorry for interrupting you there is some question from valdos uh what do you call promo type and promo mechanics okay so the promo type thank you for for question for the question uh valdes the promo type is the weekly promo or the monthly promo or the promo that was initiated by the manufacturer uh the uh those are the different types of the seasonal promo uh so the holiday season that's the that's the different types of promos that we just distinguish and then mechanics of the promo this is also very important the the ai models they uh they they they identify these patterns the combinations between types and mechanics so the mechanics would be two plus one uh buy one get one free uh is that sort of thing so there are different mechanics and that's exactly

how uh how the system there's much more of course uh but uh that's how system uh system would be able to recognize that uh do we have okay one more question um your tool is able to forecast promo sales of sku which never was in promo in the past how thank you uh helen can you help with this question uh yes we are able to forecast from sales of new esca use or squ's which have never been in promo before and it is the main difference between the statistical methods of forecasting and the machine learning approach how i i'm not sure that uh i can answer such a question just now during the webinar but uh uh yes uh we do it we we do this uh and the results are much better than

uh with the using of the statistical masses so in in a couple of words uh just simplifying everything uh we take the level of the category for the particular squ and use the promo sales of uh similar esca use for forecasting um there is this new this new item but of course not in all cases and of course it is my answer is very very very simplified thank you thank you we do have another question how you identify similar manually or automatically we can we can use use both of these approaches in some cases when we do have the information uh from for instance in the erp system that uh some squ's are similar to other sqls in this case we can use the manual

manual approach but if we do not have such information we can define the similar speeds automatically thank you we have another question from muhammad can we see the history of sku performed yes of course and it is a very important feature of uh of our tool because uh the first there are there are two uh reasons for this the first reason uh only when you can see the history and to understand the patterns of past promo you can understand the problem efficiency and make some decisions on how not to do something that leads to some problems during promo or afterwards uh and uh it is the first thing and the second one is

that only on good deep uh reach brahma history you can base the machine learning model to give you a really good result from the point of view of the forecast in accuracy okay very good thank you thank you for the yes uh answers uh thank you for the questions really really good ones thank you valdes thank you mohammed any additional questions there is one more i think in the chat what history is expected and what is minimum uh so uh we usually expect the history um let's say two or two and a half years long but in some cases uh we can do the history of one year especially after all the covet things happened with us uh unfortunately the

sailor spartans uh patterns uh really changed and the history uh more than one and a half year um could not be so relevant unfortunately but uh if the customer do have a deep history for uh promo sales we take a deep history and the minimum value is one year as i have already said thank you thank you um yes if these are all of the questions we can move on okay one more the two is able to evaluate the effect of discount advertising merchandising can so can the tool evaluate the effect of uh discount advertising merchandising uh what do you mean by the evaluating the effects so i think how how it's how these things

are affecting the promo forecast yes yes and as i have said in my presentation we take into consideration the percentage of the discount and it is the main feature because of because the model uh takes into consideration that uh before the before promo sales and before promo prices and uh correlated with the future promo price and based on this we can we can get the the right forecasts uh of course uh we are taking into into the consideration the type of advertising and take into consideration the merchandising layout because it really affects the promo sales okay

very good valdes thank you for the questions uh we have another one uh from muhammad uh what you encounter during the covet uh with your existing customers uh promo your existing promo customers uh promo lines um and i think that's the first question the second one is how you're going to use promo focus next year with covet buying so i think the first first question is about the the history what we've noticed during the covet and the second one is what's going to be in the future uh so i think that uh we faced at the very very beginning of um of the pandemic we faced the decreasing of the percentage of promos sales uh because of some lockdown things

that were happening uh in in our customers retailer chase some of our customers could not work during the lockdown and of course we faced a lot of problems in in business of our customers but let's say in two or three months everything everything went back and now of course we face some differences uh if if to compare uh with the situation that was before the climate uh but i think that it is now our our new reality and uh we need to adopt to this new reality nevertheless we like it or not and then next year we are focusing uh

on promo of course but also we face that different type of retail that do not use promo as the main as the main focus in their business next year uh so with these customers uh we are working uh let's say with this not so stable but more stable sales and increasing the forecast accuracy for non-promo sales and with those customers uh who are increasing their percentage of promo of course we are working on the increasing of the forecast accuracy will with different methods uh with adding new features to the ai model now with adding new history and new types of drama in the ii model and of course with some consulting consultant listings uh i mean we are helping our customers with

making the promo process more relevant to their business goals okay thank you uh i think we have uh one more question um and then we'll we'll be continuing uh after this is carved sales data um given the run forecast uh due to large sales during the covet so how how do these spikes affect the uh the actual uh forecast uh nevertheless these spikes were made by uh their covert consequences or some good promotional activities or let's say some seasonal spikes or anything else the main thing is the right uh basis of sales so it is the hardest thing for the model to understand the baseline sales after all uh we can take different factors into the consideration to make

the better forecasts but to have the quality in this baseline sales we need to uh clean the data from such spikes and if you had great sales during uh so i think that uh now we are during it and uh where when it will end only god knows it so if we have if you had great sales during i don't know some seasonal spikes or some uh good promo uh or some other activities we need to clean the data to their baseline sales first of all and after all to add some factors that are uh influencing the for

the which are influencing the the next problems thank you helen thank you uh actually we are we're going to continue uh thank you for the questions and if there are any remaining questions we're always happy to talk we're we're happy to give you the details uh feel free to reach out either the website or they're getting other contacts we're happy to talk because this is a very very large topic and it's very hard to share everything during the webinar such short event and we're moving forward uh we're moving forward and we we've stated there's going to be a an exclusive offer for all of the participants of this uh uh webinar so the offer is that the first three requests that we receive from retail companies that are looking to explore the potential of of improving

uh their uh trade promotion performance we are offering a free simulation when we will take the sales data and we will run it through our ai model and we will provide the report on what what is the potential of uh improving post promo over stocks of improving in promo availability on improving product turns and forecast accuracy so what we will provide is a comparison of what you've had and what you could have had with use of ai tools so once again the first three requests are going to have this free simulation so uh we're happy to to to share the outcomes of and advantages of such tools

um this has been everything that we wanted to share and to summarize is is um that the trade promotion is is very very significant part of their retail operations and it has significant impact a good way to go about the trade promotion is to actually uh fine-tune the regular sales process and then go to the next level with the trade promotion improvement the problem of trade promotion efficiency is very complex there is no single uh single silver silver bullet if you have a good forecast you need to have a good coordination in order to execute well and these are three separate processes data and ai of course can contribute to much better a chance of success

what's really important is actually having the data ready and what we've shared earlier the list of basic data requirements is what every retail company should prepare uh if the company is planning to use ai in the future without the data ai is useless and by using such tools there's a chance for continuous improvement in trade promotion so uh once again trade promotion is just one part of the process that we help retail companies um manage and uh we're happy to share more details feel free to send send us your requests send us your questions and uh we're always open to sharing the expertise and taking on the new challenge of new customers this has been everything helen thank you very much

for sharing your thank you for sharing the the answers to the questions and thank you for all participants for staying with us uh during this event thank you vlad the divorce is a big pleasure for me all right wish everyone a good day and uh stay healthy you