You place a DoorDash order and glance at your phone a few minutes later. It says 28 minutes.

Then it jumps to 34.

A little later, it drops to 26.

If you’ve ever used DoorDash or any delivery app, you’ve probably had that moment where you’re like... what is going on here?

It feels random. And yeah, a little frustrating. But there is a reason for it. It comes down to how these systems are built to handle situations that keep changing.

What You’re Actually Seeing, and Why It's Always an Estimate

The estimated delivery time isn’t rigidly set in stone. It’s the app’s best guess based on a situation that revolves around constantly changing data.

By the way, this is mathematical modeling, taking something uncertain (in this case: a restaurant, a driver, traffic, a delivery address, and timing), and turning it into something that can be estimated and updated. 

So, back to our example, what is the estimate actually based on? Well, it’s pulling from a bunch of things at once, like how long the restaurant usually takes, how busy it is right now, where drivers are, what similar deliveries have looked like, and even current traffic.

The model pulls all of that data together and assigns a number. But, in order to actually do this, it has to simplify reality. It can’t account for each possible delay or change, so the model makes assumptions about what matters most and what can be ignored. Those decisions are part of the estimate you see.

Once your DoorDash order is in the works, things can start to shift. The restaurant falls behind. A driver drops the order, and someone else picks it up. Traffic builds somewhere along the route. Next time you look at your phone, the time is different.

That’s all it is. Things changed, so the estimate did too. And math is at work behind the scenes!

The Same Thing Happens in Your Maps App

Open Google Maps, put in an address, and it will give you the time that you should arrive.

Then you start driving… and it changes. Up. Down. Whatever. You don’t stop to think about it. You just accept it and go with the flow. 

But it’s not just reacting to traffic in that exact moment. It’s making a guess about what things will look like by the time you get further along on the route, based on what’s happening now and what usually happens in that area.

If you use Waze, this is even more obvious. If someone reports an accident or a slowdown, you’ll see your route or timing change almost immediately. (And since Google owns Waze, a lot of that real-time, user-reported data feeds into how traffic is understood more broadly!)

Google has talked about how its traffic predictions work, but basically, the arrival time updates because the system is constantly adjusting as things change. It’s not a fixed number. It never really was. It was the dynamic result of a mathematical model as it responds to new data.

Why Your Netflix Homepage Feels So Personal, Too

Now let’s check in on how Netflix uses models. Instead of estimating an arrival time, it’s predicting behavior. It looks at what you watch and for how long, and what other people with similar habits watch. Then it builds a set of recommendations.

Those recommendations are not fixed either. They change as your behavior changes. And even with all of this information, there is no perfect recommendation. The Netflix model is constantly updated based on how people respond to its recommendations.

What All of These Systems Are Doing

On the surface, all of these models look pretty simple. A delivery time. A route. A recommendation. But what’s behind them is a set of decisions and some math.

Someone has to decide what information matters. What can be ignored? How much weight to give each piece of info? And what the system should care about most. For example, would it be better to underestimate or overestimate the arrival time?

That’s why the results can shift. The model isn’t trying to give just one answer. It’s just trying to keep up.

The Connection to Real-World Math

If you’re ordering food, using a map on your phone, or scrolling through recommendations, you’re constantly interacting with all of these math models. 

You don’t need to understand all the math that's behind them to use them. But once you start paying attention, you might notice some patterns. You start to wonder why something changed. You recognize some repeating patterns in what gets recommended. You might understand why delivery estimates fluctuate the way they do.

All of these systems take a real-world situation, simplify it, and provide updates as new information comes in. It’s how math modeling works in a dynamic setting. And it’s constantly impacting things you do every day, whether you realize it or not.