Forecast & Furious
“If you build it once and it doesn’t work quite right, try it again.” — Tsvika Solan
Let me take you back to 2019…back when we could meet in person. I got together with Tsvika (our Product Manager) and Maksym (our Lead Designer) in Lisbon to talk about the future of Tymeshift and where we wanted to take it. Over the course of the week, we brainstormed.
First, we looked at pain points. What were our customers frustrated by and how could we fix it? That’s when we all started talking about scheduling and automation. The long and short of it is, a lot of our customers were feeling the pain of the scheduling process and the time it took to figure out how many people they had to schedule to staff appropriately. It was our goal to remove that pain point via automation. So we started walking it back. What would we need to do to get to that core goal?
We talked about Forecast, which was a feature in early-access for a few customers to test out. The goal of Forecast is to let teams predict future contact volume based on past data and get that information to help them staff smarter.
“Forecast version one was in early access when I joined Tymeshift. We were working with the customers trying to use it and that’s when we saw some red flags.” — Tsvika
That’s when we realized how key to the plan the Forecast feature would be, and also when we got honest with ourselves that the first iteration of Forecast that we’d put together just wasn’t going to cut it. The idea was that Forecast would be a tool to help you schedule more quickly along with getting to do some of the more strategic planning like how many people you need to hire.
After that week, we each walked away with the mission to rebuild Forecast and build it to be a strong foundation that a powerful Scheduling feature could thrive on top of. So that’s where it all started.
Making Improvements
Now that we knew what we wanted to do, we had to figure out how to do it. That’s where things really lined up. We’d been building up our product and engineering teams, and finally gotten that hard-to-achieve blend of people to pull it off. We took the early user flow sketches (when I say sketches I 100% mean sketches — this was pen and paper) and moved them to Miro. We looped in Eugene, our Data Science Lead, to help make it happen.
We took each part of those user flows and broke them down into features that needed to be built. We thought about everything we would need to have to get us to the finish line, which would be a better, automated Scheduling experience for the customer. That ended up breaking down into Forecast and several other components. Then, everyone started tackling their piece.
“When I joined, the first version of Forecast was already done. Except that we wanted better results that our customers could use to plan. Actually, a major part of what I worked on for version two was figuring out which parameters we should tune and what we could do to make the data better. One thing that helped a lot was working directly with the data and filtering out the outliers. The spikes on the data messes up the forecast because it can’t distinguish if it’s just a random spike or it’s actually a trend. So it was a great improvement when we did that. The advantage is that we have such a large data set to test on.” — Eugene
The underlying functionality of Forecast version one stuck around. We figured out what worked and what needed to be left behind. Basically, we took the best parts of the original and gutted the rest.
Flexibility in our features
One of the things we shouldn’t gloss over is the infrastructural changes that we’ve had to make across the board to make Forecast work the way that we want it to work. That was also likely what was causing version 1 not to perform as well as we wanted.
We went through and carefully updated the current structure in a way that wouldn’t impact users but would support what Forecast needed. Our CTO, Tiago, mastered juggling all of those things and making sure our current operations stayed strong as we made unseen changes to prep for this launch.
“Building Forecast was a really huge step for the entire team. This is the first feature that is going to use a totally new design, a totally new feature stack. It’s the first page of a brand new Tymeshift. It took us months and months to work on this page alongside some other workstream pages.” — Maksym
I have to shout out the design folks here. They did a great job of designing something that is really simple to use and really flexible. Because it’s so flexible it’s incredibly powerful. Personally, I measure product design by seeing if users use them in a way I wouldn’t expect because of the flexibility in the software. I really think our users are going to show us things we never would have considered because it’s so damn easy to use, it’s so damn fast (thanks to the tech team), and it’s so damn flexible. You can change things, but you don’t have to — it just works out of the box too.
We are psyched to continue to optimize this and make the process of predicting your future needs even easier. Hope you enjoy the feature and that it makes your planning a bit easier. Can’t wait to hear what you think!
Since you made it this far in the post, I’ll let you in on a secret. We’re slating Automated Scheduling for early this year — looking forward to sharing more on that soon!
