I am usually considered a big fan of custom software (also known as signature software, or bespoke software to those in the UK). I think companies today often put too much reliance (faith maybe) on and poorly assess the total cost of ownership of a packaged application and concurrently overestimate the risks associated with custom development. Having a consultancy that delivers both signature software and implements Tier I ERP products gives me an unusual vantage point and a fair amount of balance in this area, or at least I think I have my bias under control.
I recently came across a case study about the IT operations at Crutchfield Corp, the electronics retailer HQ'ed in Charlottesville, Virgina. This is a company that seems to have gone overboard in the custom development activity. According to this Computerworld case study ("Closely Held Applications") Crutchfield has 18 developers supporting 300 custom applications which consitutute 85% of the running software code inside the company. (I am assuming they are excluding office productivity applications and operating systems from this figure).
My first comment: 300 applications?! This seems like an exceedingly high number by any standard. Even Oracle and SAP's complete enterprise suites comprise fewer than 200 total modules (applications) and there is probably not a company on earth that could actually use all of them (possibly if WalMart, General Electric, and P&G merged, the combined entity could use all of the modules). Now, this is not to say that those suites deliver the totality of all possible business applications, but it seems unlikely that a company, particularly a catalog/online retailer, should have 300 business applications.
And while Crutchfield's CIO says a great deal with which I agree, for example noting the wisdom of developing custom applications for those areas which are strategic differentiators, I think he may have extended his thinking in this area too far. The article suggests that even the General Ledger is custom written. As I have often said, the GL is not a place that companies should be trying to differentiate themselves; clearly the SEC and IRS both tend to frown on too much innovation in that area.
For those that are interested, and until I get around to getting them up here, I have some brief materials that helps frame decisions between custom and packaged applications. In general, I suggest evaluating the particular application area along these seven dimensions:
- Relevance to competitive advantage
- Impact on customer experience
- Need for flexibility to meet changing requirements
- Functional or technical package immaturity
- Long-term cost of package
- Cost of changing business processes to fit the package
- Cost and effort of integrating packages
While I weight some of these differently depending on the exact situation (and some other variables), the criteria tend to suggest that while an application like a web-based ecommerce system could be a good candidate for custom software (based on its high rank in the first four criteria alone), the GL clearly could almost certainly be handled off-the-shelf.
For most organizations, the answer to the "custom or package" question is "yes." The important thing is making a sensible decision about each application and not slavishly following a default choice in one direction or the other. All things in moderation.
Comments