| You depend on the quality of data and information | | | | you only what the "squeaky wheels" have to say, |
| to provide a stable foundation for your decision | | | | drowning out the valid and important and balancing |
| making. Decision making often involves responding | | | | views of the "well oiled wheels". Squeaky wheels, |
| to something, so you need your data to validly | | | | volunteer surveys and easiest-ones-to-measure |
| describe what you are responding to so that you | | | | are examples of data sources unlikely to give you |
| choose the right responses. | | | | accurate enough data. |
| Whether your data is quantitative (based on | | | | --> Define your population carefully, and select |
| numbers) or qualitative (based on perceptions), it's | | | | random samples to avoid bias. |
| integrity depends on 5 widely recognised qualities. | | | | Readable |
| Relevant | | | | Unless the data you collect is clearly defined, |
| Make sure the data you have selected is directly | | | | legibly presented, easy to organise for analysis, |
| appropriate to the purpose of the performance | | | | makes sense to its users and can be easily |
| measure you selected it for. Be careful of data | | | | interpreted and understood by them, it won't |
| that seems interesting: it doesn't mean it is | | | | matter how relevant, representative or reliable it |
| relevant. Trying to gather more data than you | | | | is. It just won't be usable. The numbers need to |
| really need, especially in surveys, can negatively | | | | be in a format you can use. |
| impact on the other dimensions of data integrity | | | | --> Design your data collection forms and |
| (below). | | | | questionnaires carefully to give you the data in |
| --> Be ruthless and collect only the data you have | | | | the format your analysis needs. |
| a use for in monitoring and diagnosing | | | | Realistic |
| performance. | | | | Trade off the degree to which your data is |
| Reliable | | | | relevant, representative, reliable and readable with |
| Collect enough data and collect it carefully to | | | | the level of resources you will need to invest to |
| ensure that it is precise enough (especially if it is | | | | make it so. Make sure the value you get from |
| an estimate based on a sample) and continues to | | | | using your data is greater than the effort you |
| be precise enough as you collect it over time. | | | | invested in getting it. Beware of the temptation to |
| Would you rely on one day's rainfall to draw | | | | invest in sophisticated automatic data capture |
| conclusions about annual rainfall? What about five | | | | systems (such as bar-coding and voice recognition |
| days' rainfall? How many days rainfall would you | | | | software) - if you haven't got a simple manual |
| need to get a precise enough estimate of annual | | | | system working well first, then these systems |
| rainfall? And what would this depend on? | | | | are likely to cost you much, much more than the |
| --> Design your sample sizes to give the reliability | | | | savings they appear to promise. |
| you need. Don't guess. | | | | --> Pilot test your data collection processes to be |
| Representative | | | | sure they will deliver cost-effective data. |
| It is important that the data you collect are | | | | TAKING ACTION: |
| observable events or characteristics that describe | | | | If you have a performance measure or KPI that |
| the full scope of what your performance | | | | triggers more debate about data quality than it |
| measure is supposed to be measuring. This | | | | does about performance levels, then use the 5 Rs |
| means that it is unbiased, or accurate enough. | | | | of data integrity to work out where the data |
| The last thing you need is for your data to tell | | | | collection process can be improved. |