Of course, we all strive for quality. No one wants to do a mediocre job, right? But in the greater scheme of things, data quality isn’t that big a deal, right? I mean, it’s just numbers.
A growing number of social sector organizations are embracing data-driven decision-making and impact measurement. But I sometimes still struggle against the mistaken idea that data is an optional extra. A nice-to-have.
The truth is, data quality can mean life or death (literally) when we’re using it in the social sector. If something with the data and algorithms that Netflix uses goes a little hinky one day, you might miss out on seeing the latest episode of your favourite show. That’s no fun, but it’s not the end of the world. But if there’s a problem with data being used to run social programs or make public policy decisions, the results can be a lot more serious.
What Poor Data Quality Did to Texas Moms
Maternal mortality is a big problem in the US. (Maternal mortality is a problem anywhere, but the US is one of a few countries that is facing an increasing number of pregnancy-related deaths while numbers in many developed countries are declining.) And the state of maternal mortality in Texas was looking particularly grim when a study found that the state’s (already high) maternal mortality rate doubled between 2010 and 2012.
What was happening to pregnant and new moms?
The answer: they were being miscategorized. When the Texas Maternal Mortality and Morbidity Task Force examined the data more carefully to determine the cause of the spike, they found that, of 147 maternal deaths listed:
- 47 were clearly pregnancy-related
- 74 had no evidence at all that the women were pregnant
- 15 lacked enough detail to determine if the deaths were pregnancy-related
- 11 deaths occurred outside too far after delivery to be counted
- An additional 9 maternal deaths not included in the first count were also found
Data quality wasn’t killing these women. But the fact remains that a huge number of already scarce resources went to figuring out what was driving such a dramatic spike in maternal mortality rates, when in fact, the number hadn’t risen at all.
Where Does Bad Data Come From?
When (agencies) can’t come up with the appropriate data — or simply rely on bad data — it’s a lot like trying to drive a car with an empty gas tank or like putting salt in the gasoline.
There a number of ways that poor quality data can creep into analysis. This is true of data everywhere, but some issues are especially prevalent in government and the social sector:
- A lack of resources or data analysis expertise
- Difficulties with data sharing or collaboration
- Errors in data collection or data entry
- Technology problems (using software improperly, or relying on outdated hardware/software)
- Agency management issues
Private corporations often have big budgets to throw at data analysis that will help improve their bottom line. But nonprofits and government agencies generally lack access to those kinds of funds — or the will to spend them on data collection or analysis rather than their core mission. Faced with siloed data, lack of standardization, privacy concerns, and staff who lack data expertise — the social sector can’t compete with the salaries corporations offer expert analysts — sometimes working with bad data can feel inevitable.
Protecting Data Quality
Not only does bad data threaten the reliability of data-driven policy making, it has myriad other ill effects – from creating civil liberties concerns, to undermining trust in government, to increased costs, waste and inefficiencies.
Stefaan Verhulst, Why We Should Care About Bad Data
So how can we protect data quality? How can we keep bad data from appearing in (and being used for) our analysis? Here are a couple of easy (and inexpensive) actions you can take to ensure your data quality is high:
- Auditing and cleaning your data
- Collecting data at the appropriate level
- Creating data biographies to ensure your data is trustworthy
- Beware of digital discrimination and data deserts
- Recognize the dangers of unknown algorithms and Big Data complexity
Another important step is to develop a data culture in your organization. If your entire team sees the value of data and understands how it will enhance their efforts, they are more likely to work to protect data quality. (Not sure where to start? Check out our post on building excitement about data.)
And of course, it never hurts to ask for help. Contact the experts at Datassist for assistance in maintaining the integrity of your data.