In the previous articles on dFM, we covered the basic concepts and how to determine the basic factors. This posting explains a simple process for building a matrix from these factors that will allow you to "cost" a scenario. A "scenario" is an option for building and architecting a software application. We then score this scenario using a score-card (described here), to compare the various scenarios.
This is a common weighted scoring system for decision support. Its not a perfect system; the subjective weighting are controlled to a degree by a democratic voting system, but a single factor could still play an unfair pivotal role in a final score for a potential scenario. This technique is shown to spark some ideas, not as a prescriptive technique.
Step 1 - Brainstorm a set of factors that influence - cost, time, performance, security, and reliability.
Step 2 - Group similar factors under a topic heading
Many of the brainstormed factors can and should be measured by one metric. An example might be the number of web servers, and the number of database servers. This can be joined under the single topic heading "Number of Servers Required"
Step 3 - Build the "Scoring Matrix" for each group.
The complexity in this step takes an absolute measurement and maps it to a simple 1 to 5 score. All measures should be 1 being lowest "cost, effort, time, best performance, most secure, most reliable" and 5 being the opposite direction.
(These are just very simple samples. The list and scores are completely fabricated in this picture)
Step 4 - Determine the factor weightings (what is more important than what)
I have three basic approaches for this -
a) You just decide relative importance and allocate weightings subjectively! Great if you can get away with it, but the intention of Design for Manufacturing is to find the "best" design from a balanced set of design objectives. If we just designed software for ease of development, we may not fully consider the operational aspects and the cost of ownership over time.
b) Paired Options (or Pairwise comparison): Get the group of people representing different domains of expertise (operations, development, business, marketing, sales) together again and have them vote A versus B, A versus C, etc. to determine what is more "important" to that person. Take an average measure of the room (more think A is more important than B). Total up the number of votes for each factor, and determine what percentage that total is of the entire selection. % weight = (count / total pairs (15) ) * 100
c) * my preferred* Stack ranking: Get each representative from each domain specialist (operations, developers, business, storage, etc) to stack rank the factors, most important to least important. This will uncover and account for individual biases. It is my preferred method because it also allows people to say "No Impact" for a specific factor on a "perspective" of the factor. Total up the relative ascending score (assign 1 to the lowest row that was a factor, and 2 to the next one up, and so on) and determine what percentage that score is of the total points available. This will look something like this -
Consider doing this for more than one perspective. E.g. As far as Cost, how would you rank importance. As far as time to market, how would you order these factors. What you are looking for is a way to mathematically represent that one factor has a much higher impact on a desired delivery "factor". Some axis of ranking might be:
1. Capital expenditure - Upfront cost
2. Operational Expenditure - Ongoing costs
3. Time to market - How quickly can it initially be developed
4. Performance - hitting agreed service and performance levels
5. Security - what factors make a system more secure
6. Reliability and Stability - what factors make a system able to handle failure more gracefully)
Whatever method you employ, the outcome needs to be a table of the factors, and a weighting multiplier, E.g.
Number of Servers Required - 24%
Bandwidth - 20.6%
Searches / sec - 10.3%
Number of Deployment Items - 17.2%
Number of Third Party Components - 13.7%
Number of Features - 13.7%
Step 5 - Score a Scenario and Multiply By Weights
Break out excel. Use the scoring sheet from Step 3 to score a scenario. Multiply each score by the weights determined in Step 4. Total all of the weighted scores. Do this for other scenarios to compare.
In the next installment of this article series, i'll demonstrate this basic technique with some real examples and prepare a final spreadsheet template you might find useful in doing your own option analysis.