"...the customer remembers more about the POC than you think. They remember how pretty the screens were. They remember how seamlessly all the pieces fit together, and how quickly each task executes. What they don’t realize is that all of the data is massaged and simplified. Of course the POC could do these things! It has 3 users with 4 roles to choose from, and doesn’t include any of the “exceptions” often found in the customer’s environment. When it comes time to actually implement this feature with 30,000 users things can (and do) get quite a bit more complex."Jeff Bohren put together an eloquent post as well, pointing to his experience as the developer in the background helping out both the POC folk, as well as the SI who has to make this happen for real. It's a great read and provides insight from both ends of the spectrum.
I have a few suggestions that might make your IdM POCs a little better. So here they are:
1. Couch your immediate goals in the context of a larger Identity Management initiative that ties it to your business objectives. Jeff already hit on this point in his post, "Instead of doing a POC of who has 'The Most Magical Bullet', enterprise would be better suited to craft a long term IdM strategy and chose a vendor whose product best aligns with it." This approach voids the notion that phase 1 of the project has to cover everything under the sun. A fantastic way to do this is to engage an SI that understands the game, and can walk you through an Identity Management workshop that speaks to both your business and technical objectives.
2. From your workshop findings, carve out a Phase 1 of the project that is attainable - the best way to do that is to write up a handful of detailed use cases that boils the expected deliverables down to non-technical language.
3. Highlight the top X number of use cases to focus on for a POC. Keep it limited (but representative) of your expected project deliverables. Identify the vendors who might be able to respond, and request a technical architecture document outlining their approach to solving the use cases you have identified and have a Q&A session with them. Filter out the vendors who don't make the cut (yes, I know that's a loaded sentence), and identify the top 1 or 2.
4. Prepare a POC lab. (Another loaded sentence).
5. Bring the vendors in, but don't allow them to touch anything! OK, this suggestion point might be a bit much but I suggest that, as much as possible, have the vendor's experts sit on their hands, next to your techies while your techies drive. If the vendor whips out a canned script, your guys will know it (and document it in the findings). If the vendor has to make a nasty directory schema change, your techies will know it. If the vendor has an ugly hack that inserts pages of code into the presentation layer of the "ultra configurable identity app", your techies will know it.
6. Have a structured way to present the findings.
While I know that the approach above requires a lot of background knowledge and support, it just seems to be a much more valuable experience that actually tells you something that a demo can't. If help is needed, supplement the team with an SI that has strong experience in the space. Either that, or save everyone time and effort and just make your decision based on sales demos and references.