76: Forecasting in Support with Craig Stoss

Charlotte Ward: 0:12

Hello and welcome to episode 76 of the Customer Support Leaders Podcast. I'm Charlotte Ward. The theme for this week is forecasting, so stay tuned for five leaders talking about that very topic. I'd like to welcome back to the podcast this week, Craig Stoss. Craig, lovely to have you back again. I would like to talk to you this week about forecasting in support. But let's talk about understanding the load and how we we figure out how we staff for that load.

Craig Stoss: 0:47

Yeah, absolutely. You need to service your core market with the core set of services you want to provide. And so I generally start with uh finding out the load, usually by hour. I try to break it into maybe hours initially by day. So for example, what is load at 7 p.m. on a Monday compare to load at 3 a.m. on a Wednesday? And you can start to see a set of trends. And you know, you see a gradual increase from 8 a.m. in the morning on a weekday until it peaks around 11, 11:30, and then it goes back down over the lunch hour. And so I start with looking at that and I start by understanding what is the level of service we want to provide around SLAs and at what times of those days are we beating SLAs or or are we missing SLAs? And then you can start to determine well, my glow this time justifies 10 people covering uh phones and two or three people covering chats. And you start to kind of build uh a vision of uh to achieve this level of service, I need to staff it with this level of people. I chunk the day out in, you know, two to four hour chunks, depending on the type of support we're providing. And I staff those chunks accordingly uh to allow uh people to have a break from taking real-time uh cases and focus on other things. So I always start with that type of model in order to try to make sure that I'm confident the core service we want to provide to our core market is uh is available.

Charlotte Ward: 2:13

SLAs are something that we often forget about, and particularly in certain segments, like the high-tech enterprise industries that I've I'm often involved with, uh SLAs can be as short as first response within 15 minutes, right? So, how do you map the people with the SLAs in that particular load segment in that particular hour or whatever it may be?

Craig Stoss: 2:35

Well, I I I try not to map it specifically to people. I map it to what the metrics are telling me. So to your example, if I see that um at 11 a.m. every day, we tend to miss a higher percentage of SLAs than at say 10 a.m., I would I would argue that means we're under uh not under staff, especially, but we're undercovered at 11 a.m. So that's that's really the base. I mean, there's there's lots of other things in there. Like for example, uh you can also say, when do the most number of frowny face uh responses on your results come in? And if you see a definite pattern that at you know two in the afternoon you get the most frowny faces, you can start to assume that the service level is declining uh at 2 p.m. for some reason. Maybe that is that you're breaching SLAs, maybe it's um you know the staff is is overworked for some reason at that time, or there's a team meeting at that time. And and you just have to decide what is that base level of service that I want to provide. And if you aren't providing it at a period of time, you then change the staffing model to meet that period of time.

Charlotte Ward: 3:38

Um, I hadn't really thought about frowny faces being relevant to forecasting, but there we go. Um you learn something every day. And of course, this model of breaking down uh a coverage period into hours or two or four-hour chunks or whatever it may be, and understanding how the load maps out over a working day extends quite nicely as you go into other territories, doesn't it? So this really just almost extends naturally into a 24 by seven model, and all you're doing is really extending the hours and extending the, you know, your hiring locations or your shift patterns accordingly, right? So there's very little extra work in actually building that into a bigger coverage model.

Craig Stoss: 4:16

But yeah, I I worked at a company that had existed for about five or six years, and and you if when I mapped exactly what I just said, number of caseload by hour, by day, and then I added another dimension of year on top of it, you could see some clear patterns of when we started to sell more into Europe.

Charlotte Ward: 4:33

Sales into new customer bases is one thing if you if you are talking about something as simple as a number of customers all using essentially the same product set, because that's fairly predictable. Final piece I think here that is more difficult to forecast for is when product changes significantly. And that has a less defined, less well-defined potential load on the support team, doesn't it? You know, a product upgrade or a product feature rollout, any of those things not only increase the complexity of your product, but potentially also the likelihood of failure of the product. How do you forecast for any of that where when there is so little data as as a precedent?

Craig Stoss: 5:18

I feel like if you and I could solve that problem, we'd we'd be very wealthy people. You know, I part of part of me when I think about forecasting for for uh new market expansions or or new releases is is um the preparedness of support. It's really about okay, well what what do I need to forecast as far as uh training time where my my existing team is taken out to learn this stuff? Uh knowledge-based article creation, um, you know, helping to ramp up new hires that we may need uh is part of that too. And and how do we determine that new new hire number? It's it's really hard. And you don't want to be reactive. I definitely start with talking to uh you know the product marketing team usually. I talk about the markets they're going to attack, but it is a really fuzzy number. I I always say um in support, we need to we always need to be fiscally responsible. Where I see a new feature release or when I hear about a new market being attacked, uh, I do try to focus on well, what can we do to maintain the self-service side of this? What do I do to make sure my team is prepared for it and be as proactive as with the resources we have, um, you know, um, and focus on the hiring as as maybe a secondary uh level of that. It really is so varied that I don't think there is a great math equation. And and me as a stats guy, that that's a hard sentence to say.

Charlotte Ward: 6:42

That's it for today. Go to customer supportleaders.com forward slash seventy-six for the show notes, and I'll see you next time.