The Data Migration Project Plan should be broken down into phases, which mirror the overall project. The phases can be defined as Analysis, Data Mapping, Build, Transfer Reconciliation, Revision, Migration Reconciliation, and Maintenance.
The analysis phase usually starts with defining the overall project. As has already been observed data transfer projects usually come about through the introduction of a new system, hence the project manager will have many competing priorities, not the least of which will be the detailed requirements that the new system must address. This is where a critical mistake can occur. With focus directed to the new system the project team pays little attention or completely overlooks the data transfer requirements. In reality, most data transfer projects in either the OLTP or OLAP space require a separate detailed project plan, those that simply include a one-line entry on the GANTT chart “Data Migration” are kidding themselves.
The second element in the analysis phase is to create the data migration team. The challenge this usually presents is that the same people are expected to undertake a role in the system implementation project as well as be part of the migration team. This leads to an immediate concern over priorities. As there is clearly no right or wrong answer to this conundrum the project manager needs to make the right priority call as appropriate to the business need.
One further point worthy of note here is that the project team needs to be dedicated to the project. The most successful projects are those that have active executive sponsorship and 100% dedication from the project team. It should not be expected that the project team members continue to carry out their “day jobs” as well as perform as part of the project team. Clear communication of project criticality, along with priorities and responsibilities of the project team members, means that the justified gaps created in the day-to-day business need to be backfilled by other colleagues who understand why they are being asked to step up to fill the void.
Getting back on the topic of data migration, as an executive of a leading asset finance business, you will know that your business is supported by more than one system. In most data migration projects, we find ourselves faced with a myriad of different systems, or I should say more accurately data sources, in which the mix of data to be transferred resides. It is not unusual for business-critical data to be stored around the organization in separate desktop databases, spreadsheets, and sometimes in flat text files. Accuracy, visibility, control, and security issues immediately are brought home to the business, so inevitably these additional data sources need to be included in the data migration.
The next important part of the analysis phase involves getting acquainted with the actual data you plan to migrate. Remember, at this point of the project, you have only the minimum of information on which you are developing the data migration plan. To get a better sense of the task it is important to get some visibility as to the quality and quantity of the data to be migrated. Obtaining an extract of data gives a good guide as the extent of likely data cleansing and the quantity of data.
So far we have not considered what the options might be for actually migrating the data. In the case of an OLAP data migration, and particularly where there is a need for regular data refreshes, in for example refreshing the content in a data warehouse, there is little option but to follow an automated migration strategy. However, in the case of an OLTP migration, which is typically a one-time exercise the option of a manual approach might be worth considering. Generally, the quantity of data would be the guiding indicator. In our industry, a data transfer of more than 1,000 agreements points towards an automated transfer, with perhaps a small number of anomalies being handled manually. Less than 1,000 is typically a risk cost assessment and as a result, you may find that the overall cost of electronic migration is prohibitive relative to the quantity of data that needs to be transferred.
A further piece of advice that is worthy of note is to consider phasing the migration project. Large volume data migrations can take several months to complete and therefore carry a significant cost. This often leads to anxiety and risk pressures building up, as well the ongoing expense for seemingly no visible result. A suggestion to overcome this would be to consider building your data migration plan based on a number of smaller discreet transfers. This would demonstrate progress, justify the cost, and hugely de-risk the “all-in-one” transfer approach. You may also find this significantly relieves the anxiety that your CIO feels!.
2. Data Mapping
After you have decided upon the legacy data sources, the next step is to define what data needs to be migrated; then those data elements are mapped into the fields of the new system. This involves going through the list of data elements from each and every data source, and deciding whether to migrate and compare to those in the new system.
During this exercise, a further extract sample of the data will be necessary to help explain any terminology anomalies. As an outcome of this exercise, you will be able to determine the extent to which the data is missing or needs to be cleansed.
The mapping phase is not intended to thoroughly identify the transformation rules by which historical data will be migrated to the new system; rather, it is essentially the act of making a checklist of the data elements that must be migrated.
The build phase is where the experience, knowledge, and skill of your migration partner, NetSol in this case, really comes to the fore. A partner with industry experience from numerous and varied data migrations will deliver project reassurance with a blend of business and technical expertise that is needed to accomplish this activity.
To provide a quick start to this process NetSol has built a data migration transformation tool into which source data and appropriate mappings temporarily reside. This is in essence a staging database, which is used to verify the data mappings and supports the cleansing and transformation of data to be carried out, all without impact to either the donor or recipient databases.
Whilst a transformation tool is not the only method available, it is by far the most efficient and effective. The alternative tends to be spreadsheets, which invariably fail when you need to map one source data element to one or more target elements.
Spreadsheets typically prevent more than one person from making modifications at one time, resulting in a great deal of unnecessary administrative overheads. Spreadsheets are two-dimensional tools, and mapping is without question multi-dimensional.
The staging post data repository becomes particularly critical when undertaking a data warehouse type migration, where the need to transform and rationalize data into new tables will be a requirement to provide rapid access to online analytics in real-time.
4. Transfer / Reconciliation
The next two phases of transfer/reconciliation and revision are iterative.
Typically this would involve a number of sub-set data and full data extracts of client month-end data pulled through the staging database with scripts to transform and map as necessary to the recipient database.
This phase would answer the following sort of questions;
- How many records did we expect this script to create?
- Did the correct number of records get created?
- Has the data been loaded into the correct fields?
- Has the data been formatted correctly?
- Does it reconcile?
Invariably this phrase highlights to users the historical data elements that must be migrated that were not apparent to them during the analysis sessions.
The truth is that data mapping is not easy to get the right the first time. Most people do not realize that a data element is missing until they realize it is not there anymore. For this reason, the revising phase must follow to correct these anomalies and the transfer/reconciliation phase must be run through again.
It is important to reach this point in the migration project as soon as possible, as it will highlight any issues that may require that the recipient systems\’ data model be tweaked to accommodate any specific client-derived requirements.
The reconciliation aspect of this phase is critical to a successful migration. In conjunction with your migration partner, you must define the criteria and processes needed to verify a successful reconciliation. Remember this is your business, your data and ultimately your responsibility to sign -off on the migration reconciliation results.
The extent of the revising phase is entirely dependent upon the outcome of the transfer/reconciliation phase previously described.
Once the transfer/reconciliation and revision phases have been satisfactorily completed it is time to implement.
6. Migration / Reconciliation
As a first step in the migration/reconciliation phase, it is recommended that a full migration “dry-run” is carried out. The dry run must take place on the final code set of the new system.
The maintenance phase only really applies in the case of performing a data warehouse (OLAP) system migration. In the contract management system (OLTP) system migration, you are working in a one-time migration environment.
In the data warehouse (OLAP) migration, you will most likely be reloading the new system at timely intervals.
To conclude, data migration is, more often than not, a necessary step.