Add Witness Formula with airSlate SignNow

Add witness formula

good afternoon from the UK and good morning to those of you in the in the rest of the world this is another in the series of the beyond modeling webinars from the Lana group my name is John Bates more I'll be presenting the session Tuesday entitled common modeling challenges today's session will be approximately 30 minutes long hopefully you will be able to hear me clearly if you have any problems or you have any questions please use the chat function on WebEx to put a question through and we'll see if we can deal with it whilst continuing with the presentation today's session of common modeling challenges we're going to cover in these sessions areas that are often raised by witness users as questions to Lana's support desk issues that we often face as Lana consultants using witness on project assignments and try and show you some of the typical approaches for these challenges based on our own experience of doing this numerous times using witness and hopefully explaining some of the flexible use of witness elements different ways to use some of the elements in witness just because a machine is called a machine in the designer elements doesn't necessarily mean you need to use it as a machine we can use it in other ways and hopefully we'll show some of that today obviously common modeling challenges is a broad subject area and other aspects of will be covered in in future webinars for today's session and what am I going to cover well the advantage of this session being recorded and then available to you on YouTube means I can try and cram quite a bit in in in half an hour so some of it may go quite quickly there's a fair bit to to include but you'll get the chance to review that on the the YouTube videos that will be published shortly afterwards today I'm going to show you how we can use witness machines to represent labor so a different use of the the standard machine element in witness use of parts to trigger actions alternative modeling of shifts and use of the system variable n to control the number of resources that we that we have in a model is instead of using the shift element itself using labor to represent a computing resource so at different use of the labor element and I'm going to build all of those together into a call handling facility example to to demonstrate some of these and then demonstrated the end how that that makes an easily scalable model from a small customer service bureau to a large call center operation with any with any modeling exercise we normally start with a process definition so here's the process definition for the model I'm going to construct for you today it's fairly straightforward in terms of a call handling process so on the left hand side a call arrives in our system it will be placed into a call queue if we have available agents they will pick the calls out of the cube answer the calls deal with those calls and the core will be completed after it that's the length of time if all the agents are busy and the calls in the core q are not answered within a specified amount of time then the there will be the option for the caller to abandon in some circumstances this may then generate another new call into the system we won't pick up that particular line today which is the dashed line you'll see on the screen but we will be reflecting the number of abandons in the in the system additionally when the call agent is is taking the call there will be an option to include for specialist software to be used by the agent so we're going to include that at the end is where we'll pick up the idea of using labor as a computing resource in the model okay I shall break out into witness now to start building this model for you I'm going to build it from scratch the only aspect that I've got prepared in advance for this is an Excel front-end with a bit of data in it I've got two sets of data here I've got my staffing profile and a call volume profile both by 15-minute period of the day just for a single day so these go down to 96 separate rows covering a full day's worth of information so to construct our model to begin with I'm going to pick up some elements from the designer elements window drag them onto the the witness layout window so I'm picking up a part so I'm picking up a buffer to represent the queue I'm gonna pick up a machine to represent our agents in a moment going into the detail for the for the part this is going to be caller so the part is the caller or the telephone call that's going to come into our system in the in the model for the moment I'll just okay that the caller is going to be pushed using the in this case the visual output rules - into the into the queue over here okay the queue itself if I go into the detail for the queue I can give that a new name or call that I'll call queue and in this case to allow for the abandoned calls I want to create a maximum delay Q ie a call can reside in this queue for a maximum length within which if the call isn't pulled out by an agent the call will abandon or disappear out of the system the maximum time I shall use I'll pick up a distribution will pick the uniform distribution and I'll say that callers will hold on to the weight to be answered for between two and five minutes so that maximum time will be sampled each time we get a caller into the queue and we'll generate their own time for how long they will wait to be answered and if they're not answered within that time they will be subject to the exit rule which in this case I'm going to say push to scrap so we'll get rid of them out of the system altogether okay so they will disappear if they're if they're call isn't answered so I also need to change the output option to any on this queue because we're generating different waiting times than any or any call could disappear out of the queue and at any point in the process the call queue itself so we're going to create the calls to be to be pulled into our agents so the machine I've created over here I'm going to rename as agent I am going to create a call handling time which will be effectively the cycle time for this machine at our agent this time I'll use a triangle distribution insert a triangle distribution and I can say that the calls will occupy or lasts for a minimum of one minutes a mode of let's say four minutes and a maximum of seven minutes I'm defining my time units to be minutes so any anywhere in the model where I'm putting entering times minutes is the unit than I'll be using so I select ok on that again selecting the agent element now selecting the input visual input rule option I can pull the calls in from the queue into our agent here and likewise the output visual output rule push to ship in this case so that's set up the flow of my model I can just stick the element flow rules on just to see the look the flows that I've created which is shown there and turn those off again to improve the display a little bit clear the display a lot of the display of the agent element here so through the display option I can update the icon for my for the agent itself to make them look a little bit more like an individual rather than a machine ok I'm also going to just reposition the part queue and labor queue so let's put the part queue in front of the agent and I'm going to put the labor queue down in front which will become a parent in a in a short while and I can create a number of these agents now so perhaps we can say we've got 10 agents in our system so I've created the flow I've created the call queue I've created that the agents I just need to attend to the call arrival pattern now so if I go back into the detail for the calls this time I'm going to select and active with profile option when I do that that gives me an arrival profile tab the detail of the of the parts screen if I move across to there that gives me this grid where I can create a an arrival profile by time segment of the day now what I'm going to do is I'm just going to load this in from Excel I need a different display because I'm gonna go with simulation time and now if I select the import from Excel button and going to my previously created excel file what I want to load in now both the the length of the period and the call volume so that's these two columns here which I've previously created so we can see at the beginning of the day we don't get any calls coming in until we hit six o'clock and then we start to get calls and they build up through the day and they create a bit of a pattern until we get right down to the end of the day so my column range there is is j62 2k 101 so it's specifying the data in all of these columns I select ok drop back into witness and now that information has been loaded into my arrival profile dialog and if I scroll down we can see that data is shown in there so I can select ok so we can now if I run this forward briefly then I should be able to check that things are starting to work so if I run that I've got calls coming in they're coming into the core queue they're being answered by the agents as we can see at the moment with 10 agents I've always got 10 agents on duty throughout my 24 hours of the day so we didn't get calls coming in until 6 in the morning and it looks like I've got plenty of agents on to to answer these calls so that's all good we're starting to get a mud all working together I will just save this model whilst I'm working on this so as we don't lose the says we don't lose anything if we have a an issue so let me just call this demo to always wise to save your work as you as you move along okay so we've we've started to create our model we've using machines as agents and we've got a representation but it's not really very accurate yet in terms of what we're terms of what we want to represent and I'm not really displaying any of the statistics at the moment around there around this model if I run this model for five days I used that as my as my time period it will run the same call pattern over each of those days it'll recur 24-hour pattern will recur so now if I run this through to the end five days I can look at my statistics see what I've got see how many calls I've generated over the week I'm not scrapping anything so we're not losing any calls that are being abandoned because everything is being answered at the moment by the by the agents and if we look at the statistics on the the agents then again they are idle for about 78 percent of the time so there's plenty of capacity to answer the calls if we were to reduce these agents down to a much lower number let's say three then if we run that through then we can see the call queue moving as we go through the week but we will be starting to lose some calls now and as we get to the end of the week end of our five-day period if we look at the statistics then we've lost a significant number of call 1398 over the week so this is all just still testing the model out at the the moment I'm going to put that back to ten okay what I might want to do now is display some of those stats on the screen to help me interpret my model so if I go to the variables tab on my designer element window I'll pick up a an integer variable first of all and I'm just going to call this abandoned so I can check how many abandoned calls I've got from the the core queue so the going to the core queue and inactions on max I can put abandoned equals abandoned plus one so that's going to count the number of abandoned calls that I get in the abandoned queue that have been scrapped again now I might want to start counting the calls the total number of calls that I've got plus those that are answered within a certain threshold we might be interested in the percentage of calls answered within thirty seconds as a typical performance standard that you might see in a call center environment so let's add another integer on to the screen and I'm gonna call this one and said I'm gonna give it a quantity of two okay now in the call queue as the calls are answered and pulled out I can put the number of answered calls I can add them up here in my actions on output so answered sorry answered and I'll need to reference the first position in my answered variable so that's going to check the number of answered calls additionally if I want some sort of performance measure then I'm going to say if and I'll use the function time in in buffer so the time the call has spent in the buffer the buffer I'm interested in is the call queue and it's the call that's at the front of the queue the one that's being pulled out at the moment that's being answered that that I'm interested in and I'm going to say if that's less than naught point five ie less it's been answered within 30 seconds I'm going to update another counter this time I'm gonna use answered again but I'll use the second position in the answered variable so again I'm going to increment that counter by one okay there's one more thing I need to do to complete this but I'll just show you the the example now as this as this runs through if I run this through now I'm getting the answered account moving along nicely and we're answering we can see we've answered 3209 calls at this point and 3168 of them were answered within 30 second threshold if I calculate that value on the screen I can add a real variable and this time this will be my called performance okay and back in the core queue in my actions on output I can say call performance equals and I'm going to say the number of calls answered within my 30 seconds which is the answered to statistic and divide that by the total number of calls which is might answer to one statistic I'm going to multiply those by a hundred just to take out the decimal places and give us a true percentage value so now when I run this we can see a percentage of call was calculated within our 30-second time frame one of the aspects of the that I need to account for is the fact that in the core queue actions on output is triggered for every entity or every part that comes through the core Q whether it's being rejected through the actions on Macs or whether it's being pulled into the the agents now currently my actions on output takes a count of all all of those so I'm not getting a true if I get any abandoned calls I'm not getting a true calculation of the performance so I need to allow for that so what I'm going to do is I'm going to create an attribute over here let's just create an integer attribute I'm gonna call it scrapped create that in the core queue itself then in actions on Macs I'm gonna say I'm going to set that to scrapped equals one so if a call is scrapped it's gonna set an attribute for that call of scrapped equals one then in actions on output I will only perform this calculation if scrapped equals zero so I only do my performance calculations for calls that aren't scrapped so that will give me now correct calculation of the of the core performance okay what I want to do now is to incorporate the my staffing profile in my excel sheet I have various staffing numbers across the day they also start at 6:00 so a my first agent comes on at 6 o'clock in the in the morning and rather than have individual shifts for each person or each group of staff because typically in a in a call handling environment you get a whole of quite a wide number of variations and shifts it is easier to determine how many people you have on duty by minutes of the day and then we can use that in the model to determine the resources available to us so we can see how these numbers move and change by 15-minute period so that's what I want to reflect in the model next and if I create then something to read that in I'll select another integer variable I'll call this my staff and I'm gonna need a quantity of 96 because I've got 96 15-minute periods in the day okay I'm now going to use the Excel wizard to read that end so I go into my actions for initialize within my staff variable select the Excel option select my file which is my demo wand or XLS file going to Excel and now I want to read in the number of staff so it's just my staff volume column that I'm reading in so I specify that column down to the end of the day select ok back into witness now I can just edit out the path here to avoid me if I share this around other machines then it won't lock in my path so if I just leave in the name of the excel file and that will work okay so let's just check that that's reading something in if I just in run that forward and then stop so I then scroll down I can see it's read in those values for me into my stuff variables and that's good what I'll also do is cut that action out of there and move it into my main initialize actions okay so I'm cutting it out of there and I'm pasting it into my main initialize actions this is just so as I don't lose track of the actions I've got reading in from the spreadsheet and I also can set the order of execution of a couple other items that I'm about to set up now okay and now I need another integer variable that is going to hold the number of staff on duty at the current point in time in the simulation run so I'll create another variable I shall rename this as on duty okay just has a value of 1 in my initialize actions having read in the value values into my staff variable I can now set on duty to be equal to the first position in my staff array effectively because that's the first item of the day that's the first staffing element that we're going to be reading okay so that's initialized that now I'm going to create a part to use as a trigger so in my simulation branch I can right-click define and add a part and I'm calling this trigger creates close so this trigger is going to be creating a part every 15 minutes so I select active and I want a part to arrive every 15 minutes and every 15 minutes I want to update the number of staff on duty from my staff variable that I've read in in this case that my first arrival wants to be at 15 minutes and as this part is purely a trigger it's just a dummy part I can push to scrap to get it to get rid of it and remove it from the model itself okay so the idea of that is purely to give me this actions on create position every 15 minutes where I can update my my staff on duty variable I'm going to need a another counter to add into the the model to do this so I'm going to create another variable I'm going to call this timeslot okay and I'm gonna use that in my in my trigger I first of all need to initialize that as well in in my initialize actions so I need to say time slots or equals one as my first to initialize my first time slot in the model back in to trigger back into my actions on create for trigger and this is where I want to update my time slot so I'm gonna say time slot equals time slot plus one so that's going to increment every 15 minutes all I need to be careful of is when I get to the end of my day at slot 96 if time slot equals 97 then I want to reset it back to 1 again so I reset my time slot back to 1 and now the crucial bit I need to reset my on-duty to be equal to the staff variable brackets and then use the time slot marker to tell me which which of the time slots I'm looking at in my stuff variable so that means every 15 minutes the on-duty variable will be updated with the number of staff on duty ok I shall save that again now when I start running this model then we can see the number of staff on duty changes every 15 minutes of run this rather quickly so that's working fine and now I simply need to put in my agent here I need to say if and less than equal to on duty pull from the core queue so now when I run my model let's slow this down a little bit what's happening is at this point in time I've only got two now I've got three stuff on duty so as I go through the day now I've got four stuff on duty but it's only because I'm using the value of n within my agents and I know I've never got more than ten agents on duty in this model then I am now using the correct number of staff on duty and in fact now if I fast forward this through to the end I can see overall I've achieved an 88% performance in my model in terms of answering calls within 30 seconds I've lost 34 calls with this staff profile that I've had on but I've quite quickly generated here a very flexible call center modeling tool and I can now use that to adjust the the call volumes my staffing profile and work out how many staff I need on by a year by time of day to achieve whatever level of call performance and I might need the final piece of this was to add some resource in and we're getting fairly close to time so I shall very quickly add in a resource into this to represent the the computing element so we might have a resource we we ought to really give that a new-style so I'll just alter the display of that and make it look a little bit more like a computing resource so somewhere we have a computing resource alternate read and now simply against the agents I in the labor all I can say something like I need to if I rename Mike as something like software okay and then in my labor rule I can simply refer to that let's say if random is less than no point eight then software is the rule I'm going to use so I just put some brackets in there and now when I run this we're going to be using a software constraint which will affect the the call performance in the system okay well we've run right up against time there if I come back to just close this off the next webinar is scheduled for the week beginning the 24th of February which will be on witness Excel connectivity so hopefully you'll be able to join us for that this will be published on YouTube as a video and I will try and address any questions that we've had via the chat or alternatively you can always contact us at support at Lana comm thank you very much indeed for your time