> New services cannibalize time from old services. If I come up with a new way to use worker’s time, by necessity it must subtract from the time they spend doing something else
This is not universally true. It is possible for technology to allow one service worker to service more people simultaneously. Consider the increased productivity of tech. support workers due to the introduction of asynchronous communication like chat as opposed to voice.
The model assumes that time spent on services is perfectly efficient, or rolls this into the service quality measure.
In reality, a potential service period can be divided into utilized and unutilized segments. The model of a CPU's run queue and time slots may be useful here.
If you're under-provisioning services, you'll stack up customers (processes) in the run queue. If yo're overprovisioning services, you'll have high idle times. Most likely you'll see a bit of both. If you have a perfect allocation system, the queue length is always one (the job running) and there's no idle time.
Technology allows you to approach that ideal. Waiting customers can be paired with idle servers. But it's not perfect, especially with services that have a physical component and travel. And you're left with the hard limit of there being only one hour per hour.
Scheduling and dispatch systems are also already aggressively persued. See Lyft.
What their present model shows, though, is the increased efficiency of high-bandwidth signalling between customer (rider) and provider (driver).
Taxis had previously relied on radio dispatch, which was better than on-street hailing, but essentially had a single blocking communications stream between drivers and central dispatch, with high wait times for channel space, relatively high error rates in transmission, and (compared to digital dispatch) low total bandwidth and precision.
The Uber/Lyft model is famous for bypassing medallions (though I suspect those will re-emerge). But it also replaced an obsolete dispatch channel which incumbents had failed to innovate on (reasons for which might be worth investigating).
I don't know whether the dispatch was the bottleneck. How would we figure that out?
(But having to talk to a human dispatcher over the phone would eg put me off a bit. More than checking how much the uber is on my smartphone. And once I'm in the app to compare prices to my other options, ordering is only a tap away.)
Absolutely. The model is far too simplified to have any real explanatory power. The most basic premise is flawed. Goods and services aren't perfectly complementary. I might want a burger from a restaurant, but it's too expensive so I cook one at home. I might want to take a taxi, but it's too expensive so I buy a bike. I might want a lawyer, but it's too expensive so instead I buy a self-help law book. And software increasingly blurs the line between goods and services for a grower percentage of the overall economy.
This is not universally true. It is possible for technology to allow one service worker to service more people simultaneously. Consider the increased productivity of tech. support workers due to the introduction of asynchronous communication like chat as opposed to voice.