Zipcar is a car sharing service that allows one to purchase the benefit of automotive mobility without having to actually buy the vehicle. Unlike traditional car rental services, Zipcar allows one to rent a car by the hour coupled with an annual premium, instead of by the day. This way, Zipcar allows customers to buy only the mobility they need – nothing more and nothing less.
Late returning of cars appears to be a problem and long-standing theme with Zipcar. In fact, for late returning cars, the customer is charged a substantial late fee of $50. With such a high late fee rate, one can only surmise that late returning cars is a large enough of a problem and Zipcar’s response to this is to change the behavior with a large penalty.
But, Zipcar’s response doesn’t address the root cause. Instead, high late penalty fees only addresses a symptom, but not the root cause.
Variability, Utilization, Queueing
I have never used the Zipcar service, but I am an admirer of such a unique and simple idea that adds tremendous value to people. So, I’m not speaking from experience, but I want to address the concept of variability and utilization as it applies to the Zipcar business and the problem of late returning cars and late fees and also the concept of WIP Explosion, which is a phenomenon we see in Queueing Systems.
Let’s contrast for a minute. For Zipcar, what would perfect performance mean? I submit perfect performance is the following:
- Each capable car is 100% utilized, meaning that each hour for each capable car, is “rented”.
- No car accidents.
- No car is ever late – that is, a returning car is never late so the next customer scheduled for that car does not have to wait.
- Each Zipcar car is returned with fuel levels at High and the Zipcar car is returned clean.
Now, looking at perfect performance above, how likely is it that the above happens 100% of the time? Unlikely I’m sure. But why?
How Does Variability Degrade Performance?
There are generally two types of variability: Internal and External.
Internal Variability
Taking from the list of Perfect Performance Variables above, internal variability could be the following:
- Returning a car dirty, leads to delays for the next customer. This can be considered “set up times” in the language of Lean.
- Returning a car late, leads to delays for the next customer.
- A Zipcar Car breaking down. In Lean, this would be considered “down time”.
External Variability
External variability is typically in the form of unknown, bursty, or irregular demand.
Variability and Utilization
As you can see in the above discussion on Variability, it is often not wise to schedule with 100% utilization as a goal. Doing so, will likely lead to the following:
- Wait times for the next customer of the Zipcar Car
- Customer complaints via customer service regarding the late fees
The answer of high penalty fees for late returning cars will likely not be effective in itself. This is true because Variability is a given in dynamic systems. And, if a system is scheduled to 100% utilization with no regard to variability, the customer experience will be negatively affected.
The Concept of Buffer
By Buffer, I mean “slack” of a type in case the process doesn’t perform perfectly (as described above). Given the physical law of variability, there are typically responses to it:
- Inventory Buffers: In case there are problems, there is an inventory buffer so the downstream processes can still perform.
- Capacity Buffers: In case there are problems, there is a capacity buffer so the process can still perform (an extra machine, additional people, etc.)
- Time Buffers: In case there are problems, we add slack of time, so that downstream service processes aren’t negatively impacted.
In the case of Zipcar, the countermeasure that addresses the root cause of late returning cars is to add a time buffer in the system. In other words, instead of scheduling Zipcar cars every hour, add slack in the schedule between rentals – that is, if a car is scheduled from 9:00 – 10:00, schedule the next customer at 10:15 to add a 15 minute buffer. Couple this scheduling approach with a late fee penalty, then the root cause is address through both policy and by addressing the root cause of variability.
Related Posts:
Become a Lean Six Sigma professional today!
Start your learning journey with Lean Six Sigma White Belt at NO COST
Tommy says
I notice that you take big inspiration from the book factory physics in many of your posts. If you havent read it and it is just a coinsidence you should read it.
Pete Abilla says
Hi Tommy,
Yes, I love the book Factory Physics. My background is in Operations Research and that book is essentially the bible, as it were.
Thanks for taking the time to read shmula.
Mark H. Davis says
Interesting article, Pete. I agree with the principle of natural variation — and that the tendency is to schedule out at 100% of capacity to supposedly maximize productivity — but I don’t see how the root cause of variation was actually addressed here, as you state at the conclusion. Seems that the variation was only incorporated into the scheduling. Cars are being returned “late,” so we’ll just push the schedule out so that they’re not “late” anymore. Would there also be an opportunity to understand the Pareto of late returns and address accordingly? Thanks again for an interesting topic.
John Hunter says
Good stuff. Without knowing the situation (myself) couldn’t a high late fee be the solution? The high late fee makes those renting cars very likely to return them on time to avoid the fee. It isn’t clear if you have data that the high late fee is just a penalty that doesn’t make the system perform better or not (to me reading this anyway). I agree, I think zipcar is an interesting innovation. It would seem to me real time (internet enabled communication) would help a great deal – notify of bottleneck, report cars needing service, tell user that car is late but these 5 nearby location have a car… Some of this is just trying to make the problem have a lessor negative impact.
Jerome Coignard says
It’s a very interesting article, Pete. On a similar topic, I would be interested in reading your point of view on how Netflix’s queuing concept can reduce the variability in customer loyalty for their subscription model.
Lee says
“I have never used the Zipcar service” – Next time, find out what the company’s solution actually IS, before you ASS/u/ME something else and offer substandard solutions which lack business sense. This is Zipcar’s solution:
There is no charge for cars returned 4 minutes or less. Customers who know they will be slightly more than that, caught in traffic, etc., phone zipcar for remedies which may include re-routing another customer who is waiting for the car, blocking the auto-late-fee charge, or allowing the customer to extend their rental time by 30 minutes increments. This flexible method also reduces complaints of the high late fee since customers are given the opportunity to fix their situations before they get out of control.
Adding an automatic buffer of 15 minute around rentals is a substandard idea since it would typically cost Zipcar 6 hours of rental time over hourly slots each day.
Pete Abilla says
Hi Lee,
I never professed to be an expert on the Zipcar service. I admire the company and what they’re doing.
My suggestion is based simply on Queueing principles, not what makes sense for the business, which I indicate in the post.
Your comment is interesting. I wonder how much time over the course of a rental day, all those exceptions you mention “add up to” in terms of rental hours. Do you know? And if you do, how does that compare to the automatic 6 hours per day?