Running Parallel test before going live


For a small company ~ 100 employees, Gas Engineering Consulting Company, How long do you recommend running a parallel test before going live?

Hi Ricka,

I have never done a parallel run before going live.

We would typically have a UAT (User Acceptance Testing) phase in which all of the business processes are tested by the Super-users and once the testing is complete do a final test data conversion, end user training and then go-live cut-over.

With modern ERP systems it’s not like your going to go live and then find the whole thing shuts down in a weeks time. It’s not like you’ll need to check the invoice values are posting to the General Ledger (unless you have been doing some serious mods.) It’s pretty safe to test everything works, fix the errors and then go live.

The other argument against a parallel run is - who actually has time nowadays to do their job twice in two different computer systems?

I totally agree with you. It can be quite difficult do make parallel systems. If two systems ERP’s work differently it can be difficult to compare results. The best way is to enforce tests before going live.

Not how long to test but how much to test?

We make sure users test all changes to the application before we update our production/live databases.

Users should test all functions of the appliaction: from creating data to management reporting.

Do they have the time to do that?


Although you say “not how long but how much”, you must be careful to try to “time-box” the testing phase of a project. If the acceptance testing is allowed to go on for too long (usually because the users have not prepared the test scripts properly or they lack the time to commit to doing the job properly) then the costs can quickly spiral out of control. You should agree a duration for the testing phase with the project team and try to stick to it.

I’ve done a parallel on a fairly large project (in fact there were a few parallels as different divisions were brought on). We found it to be extremely useful at revealing remaining issues in gaps that we thought we had filled. It let’s the “rubber hit the road” to make sure all of the requirements that were gathered are actually matching up to job functions the users and managers expected.

If you have the time and budget to do it, you definitely should.

We implemented a 2 month long parallel on each occassion.

Several factors you should consider:

  1. Resource to use for parallel period. It will be a night mare if your users if they are not “strong” enough. If the period makes more mistakes, more errors… no good.
  2. What are the standard results for a sucessful test?
  3. What is the bottom line of this test? If you are not convinced during the User Acceptance Test, this parallel would make things become wrost.

Hope this help…