What Parallel Systems Cost During an Extraction
How do you calculate the cost of running old and new systems together?
The total is the extraction team's monthly cost times the months they run, plus the ideal sync cost times those same months times the sync tax multiplier times the magnitude multiplier. Magnitude is actual sync maintenance effort divided by the effort a simple copy would take. When the data models match, magnitude is 1. When the new system splits fields and adds rules, magnitude climbs. In the article's billing example, a six-month freeze with low coupling and magnitude 1 costs $576K. Allowing product changes for 24 months, with the sync tax at 3.5 and magnitude at 4, costs $11.52M.
Why does one feature request during an extraction cost more than the feature?
The change adds four effects that apply to every remaining month: extra duration on the extraction team, sync cost for those extra months, higher coupling on the months still left, and higher transform complexity on the months still left. The last two multiply by the time remaining, so the same change early costs more than the same change near the end. The extraction stays an investment only while the change freeze holds. A product change on the old platform turns the temporary third implementation into malinvestment.
Related reading
Get notified when new content drops