Fog Creek Software
Discussion Board




Defect rates for client-server apps

Hi everyone. I'm looking for some information regarding the incidence rate of defects for client-server apps as compared to equivalent local-only apps. I'm trying to build a model to compare some different product development costs and need to be able to (roughly) estimate the amount of effort required to repair defects with one type of app versus the other.

Any pointers or experience stories would be greatly appreciated!

Jeff Kotula
Tuesday, July 08, 2003

I can't give you exact numbers, but if I had to guess, I'd say the rate is at least 2 to 10 times higher, depending on the technology being used.

For example, if you're using CORBA or SOAP between multiple languages, I'd gamble the defect rate is at least 10 times higher on average.

If building web apps with an HTML front end and CGI/servlets/etc on the backend the, the defect rate is probably "only" 2 to 3 times higher.

Also, whether or not the service is being deployed over the Internet vs a company intranet is important.

Of course, experience and the type of application being made (complexity of data structures, etc) counts too.  The more I think about it, the harder I think it is to get a truely meaningful number from something as general as "defect rates in client/server applications".  However, it's probably as good a place to start as any.

Crimson
Tuesday, July 08, 2003

I would guess that the defect rate in web apps is higher because they are easier to upgrade than locally installed software, and thus have a less strict testing and QA phases.

Colin Evans
Tuesday, July 08, 2003

In the context you provide, the question may not be answerable to an extent that would be meaningful. 

A complex single tier system is going to suffer more errors than a simple client server system.  I would go so far as: given any systems of similar complexity you will find a similar number of errors per kloc. 

Where it gets complicated is whether an error makes it anywhere.  Are you concerned with errors that make it all the way to production or any error?  Catastrophic or incidental?  Architectural or developmental?

All errors are not created equal.

Mike Gamerland
Tuesday, July 08, 2003


I have a pet theory... defect rate is probably consistent across architecture...

However, defects will increase exponentially based on the number of third party products involved in that architecture.

No proof... just observation.

Joe AA
Wednesday, July 09, 2003

*  Recent Topics

*  Fog Creek Home