| If you are planning to implement a new electronic | | | | The solution is advanced planning. |
| record or billing system now or some time in the | | | | As Stephen R. Covey teaches, it is important to |
| near future, you may be concerned about or | | | | "begin with the end in mind". For a data |
| interested in moving data from your current | | | | conversion this means to identify what you |
| (legacy) system into your new system. The | | | | expect from the conversion and to document |
| purpose of this article is to share some of the | | | | HOW you will know your expectations have been |
| things we have learned that may help you avoid | | | | met. |
| pitfalls that are common to data conversions. | | | | Balancing financial transactions in the new system |
| Most data conversion related problems can be | | | | against the old system is an obvious and |
| avoided with informed, pre-conversion planning. | | | | important step but is frequently not well planned |
| There are also some questions, which if answered | | | | for. There are often balancing differences as a |
| early in the process, will be a great help in the | | | | result of the conversion (remember the Buick |
| overall project to implement the new software. | | | | engine going into the Oldsmobile) so one must |
| Data conversions create problems just by their | | | | identify them and decide how they will be |
| nature of moving data items in one format to a | | | | resolved. For example, a system that does not |
| database that uses a different format. Think of | | | | contain transactions for late penalties in the |
| taking the engine from a Buick and installing it in | | | | history file but does include those fees in the |
| an Oldsmobile. It can be done, but even though | | | | individual account balances, creates a balancing |
| the cars are quite similar in design and are used | | | | nightmare for the unsuspecting data conversion |
| for the same functions, this is far from a simple | | | | team. One solution is to decide ahead of time that |
| project and some things just won't "convert" well. | | | | all late penalty fees from the old system will be |
| Let's discuss just a few of the potential problems. | | | | lumped into one reconciling transaction per |
| First, data migration, or the movement of data | | | | customer. The process is to determine what that |
| from one system to another, whether in a | | | | amount will be for each customer and a total for |
| one-time data conversion or an ongoing database | | | | all accounts. Then determine and document the |
| interface, must begin with a plan to maintain data | | | | resulting error in the history file and how it will be |
| integrity. The new system probably has better | | | | handled. When it comes time to balance the |
| data integrity rules built into it than the legacy | | | | conversion, this information will save your |
| system that you have been using for years. This | | | | conversion team hours of needless work and |
| will lead to some conversion issues such as | | | | dramatically increase confidence in the accuracy |
| duplicate Social Security Numbers (SSN) that are | | | | of the conversion. |
| allowed in the old system but not the new one. A | | | | Some questions you might get answers to in |
| new system is only as good as the data that | | | | preparation for your conversion fall into one of |
| gets moved into its files. If the data from the | | | | two categories, technical and data. |
| legacy system has errors or is less than accurate | | | | Technical questions: |
| (based on control or edit features in the new | | | | What are the legacy system specifications |
| system such as not allowing duplicate Social | | | | including the release number and date? |
| Security Numbers), the data will not get better | | | | What is the Database engine? |
| when it is moved into the new system. It may in | | | | Can a report be produced with relevant data and |
| fact, get worse. Frequently, vendors who sell and | | | | saved to a file rather than paper? |
| install software do not know the right questions | | | | Is the legacy vendor available to assist with the |
| to ask that will help their clients identify and | | | | data extract? |
| resolve data integrity problems in their legacy | | | | Are there multiple source files and/or databases? |
| system. It is therefore incumbent upon the | | | | Data questions: |
| purchaser to understand and develop a plan to | | | | How many primary records are in the legacy |
| resolve data integrity issues before the | | | | system? There are many files and many records |
| conversion begins. | | | | in each file. We look for a common record type |
| One solution is to make corrections in the legacy | | | | such as "customer". The number of customer |
| system and bring all data into a consistent level of | | | | records provides a relative size of the conversion |
| completeness and accuracy, thus beginning with | | | | because the number of records in most, if not all, |
| the best possible set of data items. This means | | | | of the other record types are correlated with the |
| fixing all known data problems that exist in the | | | | customer records. |
| legacy system. One challenge with this approach is | | | | Are there data items you do not want |
| identifying the data problems. This usually occurs | | | | converted? |
| when someone in the office has developed a way | | | | Are there records you do not want converted? |
| to work around their problems rather than fix | | | | Do you have a statistical report for use in |
| them, so they won't necessarily think of them as | | | | validating the conversion? Statistics could include |
| problems. However, the new system will not have | | | | record counts by type of record, financial |
| the same facility to handle a "work around" as the | | | | balancing information, and others of your choosing. |
| old system. An example is the missing SSN to be | | | | What is the expected time line for going live? |
| discussed shortly. | | | | How long can you maintain paper records of data |
| Another example is a situation where deleted | | | | that must be entered into the record system |
| records are not actually purged from the | | | | while the final data conversion is being completed |
| database. We recently converted a billing system | | | | and validated? |
| that was supposed to have 6,000 customer | | | | These represent a sample of the questions |
| records. When we reported to the utility agency | | | | whose answers will provide good information to |
| that the conversion file actually had 23,000 | | | | set the stage for a successful conversion. |
| customer records, it was a real surprise. After a | | | | Other questions and examples could be cited but |
| brief discussion they realized the database had | | | | the point is clear. Planning and gathering |
| never been purged of old customer records. We | | | | information will go a long way in preparing for a |
| encouraged them to purge old customer records | | | | conversion that will go smoother and take less |
| and related history records based on a date of | | | | time. |
| their choosing. The cost to purge was minimal. | | | | For those who are concerned about taking extra |
| The end result was a converted database with | | | | time for this planning and preparation work, let |
| 12,000 customer records instead of 6,000. This | | | | me reference research done on this very subject |
| gave the organization an excellent converted file | | | | some time ago. This study reviewed the cost of |
| with which to begin using their new system. The | | | | making a change in the design of a computer |
| file included information on customers they | | | | system before or after the actual programming |
| unknowingly would have lost had we converted | | | | work had begun. The results of the study clearly |
| only the 6,000 "active" accounts. | | | | demonstrated that when a change is made in the |
| Other problems occur when the user fails to | | | | design after programming has started, the cost |
| communicate necessary information to the | | | | will be several times greater than for changes |
| conversion team. Frequently this happens because | | | | that are made during the design phase and before |
| information that is obvious to the user and | | | | programming begins. Similarly, the design work for |
| therefore taken for granted, is not obvious to an | | | | conversions will solve problems at a fraction of |
| outsider. An example might be one of the | | | | the cost of finding and fixing problems during the |
| following: there is no city name entered because | | | | conversion itself. |
| all entries in the database are in the same city. | | | | Plan, prepare, and know the end before you |
| Another example is the substitution of | | | | begin, set expectations and consider outsourcing |
| 999-99-9999 for a Social Security Number when | | | | your data conversion project. |
| the actual SSN is unknown. | | | | |