Estimating ArcFM Data Migration Effort: Its not all about Record Counts

From time to time we’re asked to provide a ballpark estimate for a data migration effort. Occasionally the request may be phrased like this: “We have X thousand widgets, Y thousand doo-dads and Z thousand whatchamacallits. Can you give us an estimate to migrate this?” Generally the answer will be “No.” Not because we can’t do the work or provide an estimate of effort, but because record counts alone are usually not a good predictor of what’s involved in migrating data from one electronic format to another[1].

The Big Chunks

Every data migration effort is different and we could provide much more detail, however in terms of the typical “big chunks” of effort an electronic data migration will include at least these:

  • Design the migration process
  • Create a quality assurance plan
  • Develop the migration process
  • Prototype the process, at least twice, following these steps:
    • “Turn the crank” for the prototype
    • QA the prototype results according to your QA plan
  • “Turn the crank” for the final migration
  • QA the final migration according to your QA plan
  • Address any follow-up issues

Of these, the steps in which a feature count matter are where we “turn the crank” – meaning we execute the automated migration process. And it usually matters little whether there are 10,000, 100,000 or 1,000,000 widgets, the effort required to turn the crank (be it several hours or several days) will not vary all that much in terms of the overall scope of the effort.

So, What Does Count?

So, if record counts don’t matter so much for estimating effort, what does? Again, it the factors will vary from case to case but these are normally always important:

Every data migration effort is different and we could provide much more detail, however in terms of the typical “big chunks” of effort an electronic data migration will include at least these:

  • Design the migration process
  • Create a quality assurance plan
  • Develop the migration process
  • Prototype the process, at least twice, following these steps:
    • “Turn the crank” for the prototype
    • QA the prototype results according to your QA plan
  • “Turn the crank” for the final migration
  • QA the final migration according to your QA plan
  • Address any follow-up issues

Of these, the steps in which a feature count matter are where we “turn the crank” – meaning we execute the automated migration process. And it usually matters little whether there are 10,000, 100,000 or 1,000,000 widgets, the effort required to turn the crank (be it several hours or several days) will not vary all that much in terms of the overall scope of the effort.

So, if record counts don’t matter so much for estimating effort, what does? Again, it the factors will vary from case to case but these are normally always important:

 

[1] If we were talking about manual or even semi-automated data conversion then this would be a very different story.