Here’s a common question: “Should we be using Linux?”

I have a standard answer: “I have no idea.”

As described last week, vendor/product decisions are social beasts, best addressed through the When/Who/How/Why/What formula. Determine when you have to make the decision, who should be involved, and how you’re going to make it. If everyone commits to “How” — that is, to the process you’ll employ in making the decision — then you’ll be able to reach consensus on why it’s important and what the organization should do.

“How” — the decision process itself — is the real work of the formula and this week’s subject. Since Linux triggered this discussion, our focus will be on platform-layer decisions. Application-layer decisions are similar, although with much heavier involvement on the part of the end-user community. It’s a seven-step process.

Step 1: List about five candidate products and vendors. It should be a diverse list, including both mainstream and wild-card candidates. (Avoid the popular practice of creating a rigged list so you can just go through the motions.)

Step 2: Determine the important features. In the platform layer, this mostly translates to applications and services. Make this list generic (“directory service”) rather than product-specific (“NDS”). Include price as a feature.

Step 3: Establish integration requirements. For database servers, you may specify ODBC compliance. For servers you may require compatibility with your data center management system. Just as new employees must be able to fit into the team, so products must be able to fit into the network.

Step 4: Specify vendor requirements. List what matters to you, not the vendor’s internal characteristics. Availability of experts and third-party enhancements matters. The potential for a product becoming an orphan matters. Quality of support matters as well. And, there’s the vendor’s announced plans for the product. The vendor’s financial strength and the product’s market presence may lead to something that matters to you but they themselves don’t.

Step 5: Establish scoring criteria. For example: When listing key applications and services you used generic terms. If support for specific applications and products matter, give higher scores to products that provide it. Built-in services may rate higher scores than third-party solutions. And so on. The key: Know how you’ll score features, integration requirements, and vendor characteristics before you do any research, not after.

Step 6: Blend requirements into a single list and assign weightings to each item. Use a three-point scale: 3 is a deal-breaker, 2 is important, 1 is useful. If the group is small, assign the weightings through group consensus. If it’s a big group, consensus is too unwieldy so you’ll just have to vote on it: Give everyone involved a fixed number of 3s, 2s and 1s to vote with and tabulate the results.

Step 7: Gather data. Read product comparisons, talk to the vendors, issue a request for proposal if you must — and use it to score the products.

Step 8: Pick the winner. Remember, though, this scoring system isn’t precise. If the top product scores 132 and the next in line scores 128, they have tied. Also remember, vendors have been known to exaggerate, so run a pilot with the top scoring candidate (or the top two if they’re close) before making a final decision.

Now … how does Linux fit into this process?

Very nicely, if you handled Step 4 properly. With an open-source product, support is entirely through third-parties (something true of many traditional products as well), the product evolves by duplicating successful innovations of commercial competitors, and the personal commitment of its advocates is part of the question of potential orphan-hood, but the vendor issues that matter to you are the same.

Product and vendor decisions easily become emotional issues. The way to combat this is for everyone involved to attach their emotions to the selection process instead. Do it by the numbers and you’ll make a good decision. Most important of all, it will be a decision everyone can support.