SAP Enterprise Demand Sensing: Improving Forecast Accuracy in the Short Term Horizon

Learn how by determining daily sensed demand at the item, location, and customer segment granularity – Enterprise Demand Sensing (EDS) can improve short term forecast accuracy to predict out-of-stock situations in the near term to minimize expediting and transportation costs, while increasing customer service levels.

  • http://sapvideo.edgesuite.net/vod/2013/int/sap-enterprise-demand-sensing-improving-forecast-accuracy-in-the-short-term-horizon-demo-us.mp4
  • 512
  • 288
    • Welcome! Today we will be reviewing a demo of the Enterprise Demand Sensing solution to see how it can improve short term forecast
    • and help solve a common problem in today's supply chains.
    • The demo will review how a planner would use the tool to understand and communicate why a particular item,
    • bicycles in this case is stocking out in some locations while oversupplied in others.
    • The first thing a planner would do to solve this problem is arrive at the landing page we see here.
    • Across the top we see navigation links for supply chains where we can review the different supply chains scenarios
    • where we can create what-if scenarios and analysis where you can do things such as set activities.
    • In the bottom portion of the screen we see customisable widgets where we can put graphs, or charts or My Recent Activity.
    • The first step in solving this bicycles issue is to navigate to the supply chain with the bicycles in it.
    • Upon drilling in to the particular supply chain we are presented with the option to look at either our sense demand solution or our forecast accuracy analysis.
    • We are going to see this option for every base period or each successive run of the EDS solution.
    • To start a bicycle analysis lets first understand the current forecast accuracy.
    • Clicking on that forecast accuracy link takes us to this screen.
    • What we see here are the item location customer or what we call demand stream level information.
    • For each one of these demand streams we see a spark line representing the historic sales and historic forecast
    • that we have created for that particular item and some metrics describing the accuracy there.
    • We see CV, Bias and Demand Interval all of which are sortable for example clicking on bias
    • will first sort from negative bias to positive bias and another click will sort it from positive to negative.
    • So the first thing the demand planner in our scenario wants to do is look at just the bicycle items.
    • So we are going to use the filtering feature up here to go to the items tab, remove the check box for everything,
    • filter only for bicycles and drill down.
    • We have now filtered the results to the only five locations where bicycles are being sold. We are ready to start our analysis.
    • So, the report we got from our manager was that locations LA and DE were oversupplying
    • and had excess inventory while location KY recently had been stocking out.
    • Looking at the spark lines for each of these locations we think we understand what's happening.
    • The locations LA and DE appear to have some positive bias in the forecast meaning we are forecasting more than
    • what's actually selling. This is expected to cause an oversupply scenario.
    • The KY location however is having the inverse problem. This stock out lead it to deeper depth.
    • We have the ability to drill into these spark lines to generate a larger chart.
    • We see here on the bottom, the particular data that is building these spark lines and then at the top we see a larger graphic of that data.
    • So, looking at the big picture we've historically had some troubles forecasting for the bicycles at the KY location.
    • In recent periods where we have been stocking out we see that our forecast is not clearly high enough to match the actual sales that we were seeing.
    • This confirms what we expect and we realise that our stock outs are being driven by the forecast issued.
    • So, now we are ready to report back on this information and we can use some of the collaboration tools we have here.
    • With this comment button I have the ability to post comments on my analysis.
    • "KY location under forecasting causing supply issues" I can post this
    • and now it will be available for anyone who comes back to review this data.
    • I also have the ability to share this information through e-mail.
    • By clicking the share button I can insert a user's e-mail, I have the ability to cc myself if desired and I can build out the subject and body of the e-mail,
    • "Bicycles under forecasting causing stock outs". In this e-mail a link to this exact page where I am reviewing will be included
    • and I also have the ability to include a screenshot so that the user doesn't even actually have to navigate to this UI.
    • I have sent and now I have started that collaboration process where my management knows the scenario
    • and my supply planners can start reacting and my demand can start reacting.
    • So now that we have done the root cause analysis we would next like to understand how can EDS help in a scenario like this.
    • To do that I will navigate to the demand sensing portion by going under "Actions' – "View sense demand'.
    • Arriving on the demand sensing screen we see a very similar screen to the forecasting accuracy view.
    • We see the demand streams, the historic forecast and sales inputs and the results from the Demand sensing solution.
    • These results include a sensed demand MAPE, the original forecast MAPE and the delta between those two
    • and then on the far right we see a spark line of what the future forecasted sense demand is predicting.
    • We again want to continue this analysis of bicycles so I will filter my items for just the bicycles.
    • Filtering just for bicycles and looking at the MAPE change column on the right we see that there is ample opportunity for Demand sensing to improve our forecast accuracy.
    • In particular we are worried about the location KY where we were stocking out.
    • Lets drill into this spark line to see how demand sensing could help.
    • What we see here now is that same historic forecast and historic sales,
    • but we also see what a historical sense demand would have predicted represented by those green lines.
    • We see in fact that in the scenario where we were stocking out, demand sensing was picking up the pattern
    • that sales were increasing and more accurately predicting what our sales would be in those last periods.
    • As a Demand planner I know that this improved forecast accuracy certainly would have made my supply planner's
    • and customer logistic group's jobs significantly easier.
    • So, this was the weekly granularity, we would now like to see if there's an opportunity at the daily level to see any improvements.
    • Switching to the daily view we see that demand sensing was not only improving at the weekly level
    • but then its picking up the granular daily shipment patterns and improving the daily forecast as well.
    • Comfortable that this sense demand is improving forecast accuracy
    • we would next like to review what's the future forecast that demand sensing is predicting for KY location.
    • Moving out to this far right spark line I have the ability to see what's the sensed demand for future periods at the KY location
    • and we see that sensing demand is indeed increasing the forecast moving inventory to this location where we historically have been stocking out.
    • Feeding these outputs back to the supply planner will allow us to prevent stock outs in the near future.
    • So, that's the demonstration of how Enterprise demand sensing can help solve common supply chain inventory problems
    • and improve your short-term forecasting accuracy for better operational effectiveness. Thanks!
  • Tweet this Share on Facebook Share on LinkedIn Share on Google+ Email