Subscribe
Share
Share
Embed
WFM forecasting is the blending of science and art: It needs to be fluid as things are constantly changing within the contact center. The key to accurate forecast lies within historical data, skilled forecasters, and the willingness to adapt and change. In this episode, Dave and Matt reflect on finding the perfect formula for more accurate WFM forecasting.
Calabrio Shorts is a fun-sized podcast that covers all sorts of topics around the contact center industry. No topic is off-limits as we cover frequently asked questions, industry trends and definitions, and yes, we will have fun doing it.
Tips for WFM Forecast Accuracy
===
[00:00:00] Dave Hoekstra: Hello everyone and welcome to the next episode of Calabrio Shorts. My name is Dave Hoekstra, Calabrio's Product Evangelist, and we are going to be talking about forecast accuracy today. It's one of those topics that comes up quite a bit, but we don't really do it a lot of justice a lot of times when we talk about it.
[00:00:17] Dave Hoekstra: So my guest today is Matt Osgood. Matt is a solutions engineer here at Calabrio. Who has pretty extensive experience dealing with forecasting and kind of how that works. So Matt thanks so much for being here with us today, and tell us a little bit about how you kind of arrived at this journey.
[00:00:34] Dave Hoekstra: I always love to hear how our WFM gurus landed where they are today. So, tell us a little bit about
[00:00:39] Dave Hoekstra: yourself.
[00:00:40] Matt Osgood: Thanks, Dave. Yeah my journey was a little unique. I have been working around contact centers for over 20 years now, going back to high school, and was challenged by an executive at the first contact center I worked at he'd send me kind of riddles in, email as I was working as a team lead at the time, kind of showing interest in some of the scheduling practices that they were doing.
[00:01:06] Matt Osgood: So he started sending me kind of what I didn't know at the time were workforce management. Kind of, riddles or questions and I started having discussions with it with him and he ended up asking me to start scheduling for the organization that I was working with then. And, it was kind of just a natural growth from, for me from that point, cuz I was, I had a history and applied mathematics as schooling and stuff.
[00:01:30] Matt Osgood: And so it just really, it really interested me, like the deeper I got into scheduling and forecasting, the more I wanted to learn. So I grew with that call center and then joined some other organizations doing capacity planning globally. Working with different tools, installing different tools before ultimately landing here at Calabrio.
[00:01:47] Dave Hoekstra: That's awesome. Although I have to give, tell you nothing gives my heart palpitations more than hearing the phrase applied mathematics. I was not a math guy in school at all. And there, the reason I love WFM is because you get to work with people, not because you get to work with math.
[00:02:02] Dave Hoekstra: Yeah. But I suppose there are two kinds of people, right?
[00:02:05] Matt Osgood: Yeah. For me, I mean, I the people came later for me. The, my initial love was the math.
[00:02:10] Dave Hoekstra: The world needs both of us, let's just put it that way. Right? Yeah. So let's talk a little bit about forecasting. Right? And, I'm going to assume that the people listening to this particular episode at least have a pretty good grasp of kinda what the point of forecasting is.
[00:02:23] Dave Hoekstra: Hopefully they're listening to this because it's an interesting topic. And if it's an interesting topic, we probably don't need to explain to them what that is. But one of the questions that comes up quite often is, especially in, in a ear early phases of someone trying to find a workforce management solution, usually get questions like how accurate are your forecasts?
[00:02:42] Dave Hoekstra: Tell me how, what's your, your plus or minus differential. And I know that the answer depends very heavily, but if you were to talk to somebody, about ways to start to improve forecast accuracy. What where would you start? What are some of the things, some of the tips and tricks you might be able to give to somebody to say this can have a pretty decent effect on your accuracy percentage?
[00:03:06] Matt Osgood: Yeah. I mean, a lot is dependent on where you're at currently, right? Cause moving from, a 90% forecast accuracy to a 95% compared to going from a 75% to a 90%, there's different challenges that you're gonna face at those different areas of your progression. But getting started, it's really important to understand your business.
[00:03:29] Matt Osgood: The, whoever is building that forecast really needs to be in lock step with product teams or with finance teams that know the direction the business is going. In addition, you wanna be very familiar with your historical data. Make sure that historical data, Is accurate. It's scrubbed for any anomalies.
[00:03:47] Matt Osgood: You have good, smooth data. And so when you're trying to, bring those together, create that marriage of your data your historical data and your business knowledge, knowing where you're going, you wanna bring those together. Obviously, using a good algorithm. If you're using Calabrio, you got the, it's gonna do the heavy lifting of that math for you.
[00:04:08] Matt Osgood: But we need to understand where the business has been and where it's going, and then layer those together in order to qualify the quantitative metrics that the a forecast algorithm is gonna spit out to us.
[00:04:21] Dave Hoekstra: So gimme an example of how, the WFM team can work with another department in the organization to understand that where we came from and more importantly where we're headed.
[00:04:33] Matt Osgood: Yeah. So example from my past I was working at a ride sharing company. We are doing yearly projections, so we want to get a rough idea of what 2020 is gonna look like from a staffing perspective. We come into that 2019 and we have a very robust financial team that is
[00:04:53] Matt Osgood: building out the projections for, what our bookings look like for the next year. And based on that, we can, we're looking at this past year, how many actual bookings we had, compare that versus our contact volumes for all of our different lines of business. To get an idea of what that contact rate.
[00:05:13] Matt Osgood: We're able to apply that against those bookings for next year to get a baseline for an idea, a ballpark of where our forecast might be. Now I'll add, layered onto that. We're also, we're gonna talk to our product team who's, working to reduce those contact rates, reduce the customer friction in different areas.
[00:05:32] Matt Osgood: So where, we got two contacts to every thousand users that on a specific issue last year, we're expecting that to go down to 1.5 maybe next year. So we're able to layer in those product improvements, those into the overall forecast. But basing it on where we were last year and then building in where we're gonna go next year.
[00:05:55] Matt Osgood: And just out of morbid curiosity what as a ride share company, what was the singular event that drove that drove your bookings?
[00:06:03] Matt Osgood: Singular event?
[00:06:04] Dave Hoekstra: Yeah. What was the one thing? Was it the Super Bowl? Was it was. What was that one thing that just struck fear in everyone's heart?
[00:06:11] Matt Osgood: New Year's Eve, is probably, and Halloween actually is another, really big one. Interesting there, there's actually some social media noise that can cause massive changes in your needs overnight. So, that ability to be able to, flex and adjust your expectations very quickly is important.
[00:06:31] Matt Osgood: And yeah.
[00:06:33] Dave Hoekstra: Oh, I was gonna say, I think. That's such a great example how you mentioned working with the product team, right? This particular last year was a problem because we didn't have this particular feature in our product. Now we've addressed it and now it's available as kind of a self-service option, right?
[00:06:47] Dave Hoekstra: So we're expecting significantly fewer things. I think that's a great example. Some of the other things that I've seen maybe on a more basic level are make sure you're communicating with the marketing team. Make sure we know when that email that's gonna go out that says, "Hey, free shipping on orders over $40".
[00:07:03] Dave Hoekstra: If you don't know about that in advance, it's going to just absolutely ruin your forecast. These are the little tiny little things that making sure you're communicating well in advance with other areas of your your contact center and outside your contact center into the organization. These are things that can just wreak havoc on your forecast accuracy.
[00:07:23] Dave Hoekstra: Now, the other thing you said. Matt that I want to have you dig in a little bit deeper. You mentioned kind of cleaning up the anomalies for, let's go to a little bit of a basic level, because I know that forecasting on the one hand is one thing, but cleaning up the anomalies is another thing.
[00:07:38] Dave Hoekstra: What's the process that you kind of, what a typical WFM user would go through to kind of make sure those anomalies? Let's talk, let's start there. Why would anybody care about cleaning up the anomalies?
[00:07:47] Matt Osgood: If you've worked in a contact center, we've all been there when something bad an outage has happened, or we're, we get thousands of calls in a minute where we're getting, where we're normally getting, 10 or 20 calls.
[00:08:00] Matt Osgood: And those types of events can significantly skew individual intervals when they're so far out of line to your normal. Where Wednesday at one o'clock. One day we got a thousand calls, but our average for a six month period on that same interval is say 10 or 20 calls.
[00:08:20] Matt Osgood: If we build a forecast and we do not remove, remove or smooth over that one anomalous day, we're gonna end up expecting, 20, 30, 40% more volume on that interval than we should. So, as a user of a WFM system, we typically, first I, identify the historical data that, that is meaningful to us for the forecast that we're doing, whether that's multiple years of data or just Q1 data, we'll pull that up and we'll specifically look for individual days.
[00:08:56] Matt Osgood: A platform like Calabrio is going to isolate and highlight days that are deviating from the average by more than 10% or 20% depending on how you set up the system. So you can quickly identify days that are out of alignment with your normal averages as far as handle time or contact volume and how you choose to handle that
[00:09:20] Matt Osgood: whether you manually over. That day to the average for the period or something of that nature. There's different ways to handle it, but you just wanna make sure that you don't proceed with that forecast with any major any days or intervals that are majorly out of alignment with those typical averages.
[00:09:39] Matt Osgood: And cuz you can't, we, we can't plan for an outage, right? We don't know when that's gonna happen. We we want to plan for a normal Tuesday, a normal Wednesday, and then have processes in place to, in case something does happen in order to adjust on the fly.
[00:09:56] Dave Hoekstra: And so I, I think a good example of that, just to completely take it out, call center terms.
[00:10:02] Dave Hoekstra: There's a great little comic going around the internet right now that it says that, did you know the average person eats on average two spiders a year or something like that? And the guy says, Yeah, most people eat zero, but that guy's eating 40 spiders a day. Right? And that's a great example of why you want to clean the anomalies up.
[00:10:21] Dave Hoekstra: Right? If we remove the anomaly, our averages are actually much cleaner and that's, it can really pull. And so, anomalies can be an outage, like you said, but they can also be a scheduled predicted event, just like we were talking about earlier. New Year's Eve, right? When, New Year's Eve is going to be insane.
[00:10:39] Dave Hoekstra: You don't want New Year's Eve to affect your normal Tuesday distribution or your normal weekly distribution, right? Those are definitely, that's why we do it. So the the analogy that I use a lot is that forecasting is like a garden, right? You. You can't just throw some seeds down and walk away and expect to come back to these beautiful plants, right?
[00:11:01] Dave Hoekstra: You have to treat the soil. You have to, water it, you have to weed, you have to put down your organic pesticides, You have to prune. You really do have to put in a fair amount of work and as good as computers have gotten these days about recognizing these kind of things, the computer still doesn't know that.
[00:11:19] Dave Hoekstra: The reason you got way more calls last Tuesday is because some knucklehead in marketing sent out an email that they shouldn't have sent out.
[00:11:26] Matt Osgood: Right, Right. It just knows that you got more calls.
[00:11:29] Dave Hoekstra: Exactly right. It's I got more calls and all I'm doing is creating an average here and that's what's gonna happen.
[00:11:34] Dave Hoekstra: So it's very important to go back and make sure that we really spend the time. Cleaning out those anomalies that that we go through. Now, one of the other situations that comes up for a lot of people is creating forecasts from scratch, right? So this happened to me a lot. I used to work for a BPO and we would get an email that says, Hey, how many people do we need for a thousand calls a day?
[00:11:58] Dave Hoekstra: And I'm like, Whoa. Not even close to enough information there. Right? So when we're talking about kind of creating a forecast and having it accurate from scratch, Like what are some tips or some recommendations you might have in that area?
[00:12:12] Matt Osgood: So this is, Yeah. Something I ran into a lot in my, my last role, doing kind of global capacity planning and going back prior to that I was working with a company, a global company that we operated in 14 different languages and while I was there
[00:12:27] Matt Osgood: we were consistently opening up, it felt like a new language line every two months, or so as we were ramping up and depending on the type of, the line of business, usually there is some, a corollary within your organization that you might be able to reference, whether it's another queue or potentially some, Through a website, you might be able to track the amount of traffic that's coming through a specific help hub that's going to be opened up for customers to call in.
[00:12:56] Matt Osgood: So you can use that as a baseline. But something I've done a lot in the past is, trying to, working with the teams that are creating this. This new line of business or new product to understand how it relates to the rest of the business to identify which lines of business we have existing that might have similar customer behavioral patterns, and we can use that to initialize.
[00:13:19] Matt Osgood: Our arrival patterns. So we might know that generally when our customers are going to call might be very similar between this new line of business and an existing line of business. We can take that existing line of business, apply the indate templates that we've built for them.
[00:13:37] Matt Osgood: And apply those to this new line of business and leverage like I was saying, potentially website information as far as traffic. Pair that in different areas of our help sites to get an idea of what our contact rate might be, what our overall inbound volume might be compared to existing lines of business.
[00:13:57] Matt Osgood: Marry those two together and you're typically able to get a good working baseline from obviously the less you're less, you're able to discern about that new line of business front, the less accurate your forecasts are gonna be upfront. , but typically The people that are closest to that launch are gonna know a lot about it.
[00:14:15] Matt Osgood: And, you can work across your organization to draw insights from from those product owners.
[00:14:21] Dave Hoekstra: So, you bring about, a good point is that a lot of times when building a forecast There are organizations that are fortunate enough to be able to only use their historical interaction data, right?
[00:14:33] Dave Hoekstra: How many calls did you get? And you can go, Yeah. But there are a lot of organizations that have to bring in external factors, right? Things like website hits or new account signups or things like that. How would you bring in some of those external factors to enhance that accuracy? What are some things that somebody might be able to do to kind of figure that out?
[00:14:53] Matt Osgood: I mean, typically you're, you need, you're looking at ratios and you want to, and you're comparing versus established groups, so you know where. Maybe on a, like a claims issue, I know what my traffic is through this specific customer service contact path, and I know how many phone calls, I know how many emails I'm getting from that.
[00:15:16] Matt Osgood: So I can I have that ratio. And then with a new line of business, I might have one side of that ratio. And then I'm filling in the blanks, by comparing it to that existing area, that's the ideal is that you have, an established group that you can look at that same ratio against to apply versus a new line of business.
[00:15:37] Dave Hoekstra: So let's go back to your original kind of point you made. What if I am at sitting at 90? Percent, and I want to get to 98%. Is there a magic bullet there?
[00:15:46] Matt Osgood: No magic bullet. Like what you were saying before it's hard work and attention to detail. There's. A lot of moving parts, and you gotta be really you want to identify trends in your inaccuracy as well as your accuracy, right?
[00:16:02] Matt Osgood: Cause you're when you're getting, when you're getting that close to the line Typically, you're going to identify either changes over time in your customer's behavior that you want to stay ahead, you want to stay on top of. So where if we're looking at the times of day, the days of week in which our forecast is starting to deviate in the rates of change of that.
[00:16:25] Matt Osgood: So, maybe three weeks ago on my weekend graveyards I was within 5%. And then I've see seeing that dipping to where by, by this week, over, over that weekend I'm off by 10% and it's continually moving in the same direction, up or down and continually reevaluating. Those those arrival patterns and updating your forecasts.
[00:16:49] Matt Osgood: The one major thing also in, in regards to that level, the high level of forecast accuracy is how often are we revisiting it? . Cause some organizations, Which I've worked in the past, we'll put out a forecast for an entire year and an entire quarter is locked.
[00:17:06] Matt Osgood: So we have forecasts three months out, halfway through that quarter. We may have new information to where we know our forecasts are not gonna be good for the rest of that quarter. But we can't change them cuz we're locked into budgets with partners or with other organizations.
[00:17:24] Matt Osgood: We know that we're gonna manage to change numbers, but due to how we operate, we can't actually change the forecasts. So, the pro process is a big part in being able to get to that, 95 plus percent accuracy range. But you're never getting there. If you're not staying close to the data, you're not staying close to the business and the owners of the different areas that are gonna drive the business forward, like we were talking about.
[00:17:51] Dave Hoekstra: I think that's the, maybe the lesson that's really important here. Is that the tool is a big chunk of it, right? The tool's going to save a lot of legwork and a lot of processing and a lot of Excel spreadsheet creation. Or I know you, you're a big fan of Google Sheets, but.
[00:18:08] Matt Osgood: They allow you to scale too.
[00:18:10] Dave Hoekstra: Yeah.
[00:18:11] Dave Hoekstra: Yeah. In scale to, the point where, so what we notice a lot of times is, hey, we implemented this WFM tool and our forecasts got more accurate, and we're spending a lot less time doing it, but it's the law of diminishing returns. We can get from 75 to 90%. By clicking a few buttons and really not thinking about it.
[00:18:29] Dave Hoekstra: But to get turn 90 to 95, you have to put the work in. You have to really understand where those little pieces are. And I think that's the thing that I want people to take most from this is that, as much as we'd love to admit, That there's a button out there that can do all of this for you.
[00:18:46] Dave Hoekstra: There's still, there's a reason that we kind of refer to forecasting as kind of a blend between art and science. The science part. There's a lot of math and a lot of fancy algorithms that are working behind the scenes to do a lot of this work for you, but, The art of this is understanding your business, understanding the organization, understanding those factors that aren't necessarily captured by a phone switch.
[00:19:09] Dave Hoekstra: Right. Those are, this is great. So we covered we covered, cleaning up the anomalies. We covered kind of putting in the right amount of work. Any other things that you that really kind of trip off the radar for you when talking about forecasting accuracy?
[00:19:24] Matt Osgood: I think I mentioned it a little bit there in the last part, but reporting and the data availability, cuz you, you you want to have a consistent and easily accessible and consistent view into how your forecast is performing.
[00:19:39] Matt Osgood: And so it's, I feel it's very key to have that kind of live streaming dashboard available to where you kind of always know where you're. Compared to what you were expecting, and you're always evaluating that. Yep.
[00:19:54] Dave Hoekstra: And not only that, but debrief sessions as well, right? Postmortems? Yes. That truly look granular at that information and can I'll let you talk about that a little bit.
[00:20:02] Matt Osgood: Yeah. Yeah. So, and that, that goes back to being on top of the, that pulse of the business we were mentioning when. Somebody sends out a product marketing email without, and work that workforce wasn't aware of and we spike way off. Those are types of things that in the past, they can actually be a good thing long term because you're identifying a gap in your processes.
[00:20:24] Matt Osgood: You're able to bring that organization into the WFM fold, so to speak. Have a postmortem, here's what went wrong. Next time this happens, here's what we want to do next time. Something. And then you can create a closed loop process around that with all your groups. So, yeah. And depending on the speed at which your organization moves something like, Biweekly or monthly cadence meeting with each one of those kind of different shareholders across the organization that you're interacting with in order to build those accurate forecasts.
[00:20:56] Matt Osgood: Very important.
[00:20:57] Dave Hoekstra: Yeah. It's just like growing tomatoes, man. It's you can have good tomatoes this year, but if you want great tomatoes next year, you gotta. You gotta learn from what you did and what the changing factors are and things like that. So yeah. Matt, this has been amazing.
[00:21:09] Dave Hoekstra: I think this is super valuable information a lot of people are gonna benefit from, so it's been really great having you be a part of the Calabrio Shorts group here. Thanks so much for joining us.
[00:21:19] Matt Osgood: Yeah, thanks for having me, Dave.
[00:21:20] Dave Hoekstra: Absolutely. And to those of you who are listening, thank you as always for spending some time with us at Calabrio.
[00:21:26] Dave Hoekstra: If you need more information or want to tap into somebody with that kind of expertise that Matt has, give us a shout. Let us know. Go to Calabrio.com and and send us a message. We'll be glad to talk with you about your forecasting and many other things contact center related. So we're super happy to have you as a listener and we'll see you again on the next episode, Calabrio Shorts.
[00:21:44] Dave Hoekstra: Thanks everybody. Have a great day.