This content was created by Max Brun. This content was uploaded by Max Brun. This is a YouTube video titled "Localized Marketplace Metrics & KPI's" (created 2022-09-17) containing a recorded dialogue between two unidentified speakers (SPEAKER A and SPEAKER B). The discussion covers metrics and KPIs for localized/on-demand marketplaces with anecdotes from Lyft and references to Shift, highlighting service-quality KPIs such as expected pickup ETA (user tolerance ~3–10 minutes), price sensitivity and demand elasticity (large drop from small price increases), number of rides as a company North Star, on-time delivery, and cost.
Chapters
You know, on that note too, as far as, you know, localized marketplaces and, you know, not all markets are kind of created equal. So what are some metrics and kind of KPIs? And we can just maybe a little bit more generally speaking, or from your time at Lyft, if that's better, but that you use to kind of measure, you know, some of the, I guess, the markets, you know, how healthy those markets are.
I'll give you an example. I'll come back to an example of price. I'll use an anecdote. from Lyft, there were really 2 KPIs that we were, you know, indexed on. And it's the reason I start— I'm starting there. It's fairly— it's almost the same thing at Shift as well. So I'll first give you a Lyft example, which is more relatable, and then we can come back to Shift. So there were 2 KPIs that we were always monitoring, and both the KPIs were really what we called as— what we termed as our service quality KPIs or service level KPIs. Right, so the first one was your expected time of pickup. So you, as a rider, if you place a request in the Lyft, you know, with Lyft to go from point A to point B, how soon would you actually have somebody come and pick you up? So that was first. And the second was obviously price, right? Like, in the area of price, it's like, how price sensitive are you? Given a point in time, given the urgency, do you want to— are you willing to wait? So if we pitch a certain price to you, will you be willing to wait or will you just drop out? And, you know, the goal is obviously to maximize revenue by maximizing the rides, maximizing the volume. But so depending on— some people are willing to spend more, some, some people are not willing to spend more, and machine learning algorithms will actually learn that over time as they learn your behavior, etc., etc. Uh, the first one is very interesting, and I'll just quickly, uh, you know, summarize the anecdote very quickly. So the first one was interesting. What Lyft saw was that, uh, between up to about 8 minutes, you know, between 3 minutes to about 8 minutes, 8 to 10 minutes before going and getting into double digits, uh, you could have a ride come between 3 minutes to 10 minutes and people would not drop. People are willing to wait, right? So, so, so again, so 3 minutes to 10 minutes is a is almost tripling the wait time. If you think about it, going from 3 to 10 is almost more than tripling the wait time, but people are willing to wait for that much time. After 10, then it becomes a problem because now they want something more urgently. 10 minutes is a lot of lead time for an on-demand marketplace like Lyft, right? But they're willing to wait between 3 and 10. You would see— so coming back to what you said, right, hyperlocal marketplaces, Sometimes you just don't have, especially when you're starting in a new market, you just don't have enough density. You don't have enough drivers right now. So you may not be able to find a driver within 10 minutes, for example. It's possible, right? So, and you use batching and things like that to alleviate the problem, but it's possible. People are willing, people are okay with that. On the price side, if you go from $10, let's say as an example, to $10.15, which is, you know, which is, uh, you know, if you think about it, it's a 1.5 percent increase, the demand would drop off a cliff. So we would kind of joke like people don't know how to value their time because, you know, or how to price their time because you're, you're more than willing to wait 3 times, 3 times as much to find a ride. But if the price goes up by a few cents, 10, 15 cents, you're actually— your demand is dropping. So that's just how humans are programmed. They are very, very sensitive to price compared to, you know, their own, their own time. This is something we saw there. So ETA of pickup was one metric, price was the other. Of course, number of rides was obviously the company's North Star that, you know, you would report to Wall Street. It's the same thing on the ship side. There's something called on-time delivery, which is on the customer side, when you ordered something, how many of those are coming at the time when you wanted them to be, which is again sort of like an ETA problem. And then obviously cost is the other thing on that side.
