|
How to find the best data quality approach
Data Quality approach once fixed leads to the data quality plan. This plan includes the initiatives linked to Data Quality and also the business as usual elements. For example you can have a plan , which has:
Four Data Quality initiatives:
- implementing a common mapping system between the feeding business system and the general ledger
- implementing common business sub-ledgers
- normalizing the product codes
- implementing a new JV system
Business as Usual elements:
- Ongoing training on financial data quality
- Daily monitoring of Business System
- General ledger data synch..
You have seen various considerations we should take into account while finalizing an approach for data quality program. The fact is that there is no single and straight approach for a data quality program. It is actually a permutations and combinations of the set of approaches. To keep the matters simple, here are the heuristics you can use to build your own approach for your own DQ program:
- More important data needs to be fixed first.
- For most cases the priority should be to implement prevention. For the areas, which cannot wait, the fixing of existing data may take precedence.
- If the new data influx volume is high- implement the filters to ensure that only clean data enters the system. This to be followed by cleanup of existing data and finally go into the prevention methods.
- If the new data influx volume is low- place the prevention methods (instead of spending energies on inlet point filters), followed by one time cleansing of data.
- If the data is no longer in use, and is only required for some reporting/analysis- Do one time clean-up of the data either in the production system OR your staging area/data warehouse, depending upon where it is easier to do .For example, easier to fix it in new data warehouse environment instead of working on flat files of legacy production system.
- If the data influx volume is not high, but the data gets updated at high rate- Give first shot to the prevention methods. This is because once the data starts getting updated correctly, the errors will be over-written with correct data. For example if the inventory status of part with faulty reference number is updated almost on an hourly basis, it is better to fix the system, which is feeding the consumption details to the inventory management system.
- Start quality program with a function/system having a higher level of readiness.
- Keep high aspiration and go for small & realistic goals. Display the showcase and get greater level of sponsorship, before taking a larger leap.
Finalizing Data Quality Program Approach Given that we have done the root-cause analysis, we know possible approaches, the various considerations of data quality approaches. This leads to the proposed solutions, which contains the following: - The final solution set with cost and benefits.
- Reasons for the final solution set.
- Sequence for implementation of the solutions.
This proposed approach is then taken through the stakeholders. As they are going to pay for the program, they will collectively decide upon the priorities. This is the point, when realistic quality expectations will come out.
For example – Head of Renewals received 15% return mail on renewal notices of an Insurance company. If he is presented with two options – - First option is, through building better front end capture process, one can reduce the address list faults to 5%. This option costs USD 10000, and will be done over 3 months.
- Second option is to do a field investigation, and it will reduce the errors to 1%. This will cost USD 5 million and done over next 2 years.
Chances are that first option will be selected.
PLEASE REFER Execution-MiHPractice Tool Data Quality Approach, Scoping, Planning and Tracking Work-Tool
|
