A vital element of any new ERP system implementation comes when you migrate data. The ability to recognise useful information from legacy systems can help ease the transition between the old and the new.

The task of migrating data should be considering early in an ERP implementation cycleIt is key to the success of your project – by leaving it too late your project is at risk of abject failure. 

While it can be hard to decide which data to retain and which to cleanse, a review of legacy data is a chance to start afresh.

What data are we planning to migrate?

Breaking down your source data into distinct categories should be one of your first tasks – usually it falls into the following categories:

  • Transactional data – regularly changing data makes it harder to migrate – open sales orders, opening balances and stock on hand. 
  • Reference data – typically static data that rarely changes over a period of time, for example; country codes, tax codes, product classes etc. 
  • Catalogue data regularly changing data makes it harder to migrate – open sales orders, opening balances and stock on hand.

Each category would be migrated at different stages of the implementation, with more changeable data migrated close to the go-live date. 

It is also necessary to question whether your legacy system’s source data is of good qualityAfter all, if your business is investing in a new ERP system, why bring over any “bad data”? 

Data in your legacy system you should consider cleansing would be old customer records and addresses – as well as consolidating any repeated information 

How are we going to migrate?

It is wrong to assume that the data migration must be done automaticallyMore time and money may be spent than is necessary building complex programmes that deal with small amounts of data that could have been manually migrated. 

It must be considered what outdated data can be extracted from the legacy system, and is it worthwhile in doing soCould the data be migrated consistently – if the data is going to be migrated it must be done accurately, otherwise valuable time will be wasted and data will have to be reviewed manually in order to validate its quality 

Manual migration of data may appear daunting, especially considering the potential number of records. The effort can be quantified through a range of criteria, including: producing an estimate of the number of entries, time dedicated to entering data, and determining who will enter the data – temporary staff might workbut they would not have knowledge of the data’s context that a permanent employee might. 

Testing your Data

On large projects at least one person will be responsible for automatic data migration. During the project some customisation work may occur, so migration schemes may need regression testing to ensure that it still functions as intended during the simulation phases. 

Detailed planning is required throughout the migration and we recommend producing a summary of the migration activities and methods. All project members and sponsors should be informed to ensure they are aware of each activity’s time commitments. 

There are three main data migration methods: 

  • Transactional data – regularly changing data makes it harder to migrate – open sales orders, opening balances and stock on hand. 
  • Database import – if using SQL Server or Oracle it is possible to use a direct insert method into the database – this is a skilled approach, but has clear performance advantages.  
  • Programmatic import – programmecan be used to import data – not only ensuring the data’s integrity, but this also results in a more stable end product. 

Data migration is a vital element in any business system implementation project, and it is critical that it is given high priority in the project planning, with the requisite resources and skillsets required to ensure the overall quality of the implementation project isn’t compromised. 

If it’s not given enough thought, or considered too late in the process then your entire ERP project, or worse, your business could be damaged.