‘Takt Time’ for Apple’s App Store and for Healthcare
Sunday, I read this New York Times article about Apple's “App Store” for the iPhone and iPod Touch (I've been a pretty happy iPhone user for the past three months after switching over from BlackBerry).
I'm going to try to use this example to teach about two concepts that can be used in virtually any process – “takt time” and cycle time — including some questions for healthcare.
Some details of Apple operations came out through filings made in response to the controversy over Apple not carrying the “Google Voice” app and an FCC investigation. The article unveils:
For example, Apple receives 8,500 new applications and updates to applications each week. The company employs a team of 40 full-time trained reviewers, and each application is independently evaluated by two separate reviewers before getting a green light.
Post continues after ad...
Takt is a Lean concept that comes from a German word. If you're a musician, you recognize the “tt = 120” designation to mean that “takt time is 120 beats per minute.” You can think of takt as the tempo of a process. And the definition starts with the customer – what is the pace of customer demand?
Starting with takt in mind (again, focusing on the customer) is the first step in making sure that your processes can meet your customer's needs. This concept can be applied in automobile manufacturing (that's the rate of customer demand for Priuses?) or even in medical settings (what's the demand for “blood draws” to collect testing specimens from patients?)
The basic formula is this:
Using the data from the article, I am assuming the following:
- I will assume the 40 FTEs work a full day (7.5 hours of work after lunch and breaks)
- If they have 8,500 apps each week and there are two independent reviews, that demand is actually 17,000 inspections per week = 3400 inspections per day assuming a 5-day work week.
7.94 seconds? How in the world can they do an inspection of an application that quickly?
Well, they don't have to work that quickly. I'll explain.
“Cycle time” is a term that means how long it takes to do a repetitive job each time. If you are writing three blog posts in a row, the cycle time is the time required to write one of them. (More on the term “cycle time” here).
For any process to meet demand, the total cycle time must be faster than the takt time. This means a smaller number: a cycle time of 10 minutes is OK if takt is 15 minutes, for example.
If Apple has 40 FTEs working on these application inspections, then each doesn't have to get the work done in 7.94 seconds. Each one can take 317.6 seconds, at most, as shown below. If they take longer, then the work won't get done and inventory of un-inspected (and un-approved) applications will grow forever.
In the physician example, if takt time = 5 minutes (a patient arrives every five minutes), and one physician requires a cycle time of 10 minutes, then you need 2 physicians to keep up with customer demand. This keeps takt and cycle in balance. We can get our work done.
Do you use an approach like this? One potential problem is that this approach assumes customer demand is equal and constant across a period time. We know that's not typically true in healthcare. If you have, on average, a patient arriving every 5 minutes, but four arrive at once, you are bound to have some patient waiting. That's one reason why it is important to “level load” demand where you can (a subject for another post, I guess).
In the laboratory testing examples, if we have to draw blood from 180 patients for testing before morning rounds, and we have three hours to get this done (10,800 seconds), then the take time for collecting specimens (and for testing specimens) is 60 seconds. This is an example where we *can* level out the work over the three hours (we don't have patients arriving randomly).
If drawing blood from one patient (and walking to the next) takes, on average, five minutes –> how many phlebotomists do you need at minimum?
I haven't seen too many labs (or other healthcare settings) use math like this to even estimate staffing levels. Have you? Is this a better method than using benchmarks or budgets? Is this method practical, even given some real world variation? Even if you say “no,” does the explanation make sense?