Recent articles seem to abound with phrases like “data is the new currency.”
Although it’s fairly clear to the reader that what this and similar phrasing is referring to is the value of data, it can be taken another way too—for both currency and data, value is tied to trust.
When we begin to question the data we are using—particularly when analysis is being done to support strategic recommendations—doubt starts to creep in.
When doubt creeps in, the potential exists for decision-makers to revert to old habits and established ways of doing things.
When we doubt data, we turn to what’s comfortable.
That can lead to missed opportunities.
Using the right data is critical. We’ve all heard the phrase “garbage in, garbage out.”
So, what can communicators do to create more trust in their data?
Verifying and cleaning data
One of the first and most important steps you can take is to make sure your data is as clean as possible. This doesn’t mean going line by line to catch every mistake (unless that’s feasible—in most cases it won’t be an efficient use of your time).
What you can do is a spot check of a subset of your data. Separate off and review a chunk and see if things look correct. If there are errors, you’ll typically be able to see them fairly quickly.
Things like truncated or incorrect email addresses, content in the wrong fields, numbers where there should be names—any glaring errors will point to an area that might merit closer review.
This will point you to areas where you might need to verify and clean up your data set.
These are steps that you would take for smaller data sets—things like your media monitoring results, for example.
Undertaking this process is particularly important if there are subjective elements, like tone or sentiment, that you are relying on to make communications decisions within your organization or for clients.
Reviewing any subjective assessments present in your data is important whether you are using human review or automated sentiment analysis. Human assessment of sentiment can be inconsistent because of individual differences in interpretation. Automated sentiment can be inconsistent because sarcasm, humor, and a variety of other factors.
Verifying that any subjective determinations have been applied to your data with a degree of consistency is important.
For much larger data sets, you’ll need to undertake a more robust version of this process, called exploratory data analysis.
Christopher Penn details this process in a recent video on The Measurement Standard, which I recommend highly to anyone in PR or a communications function who wants to learn more about this aspect of data analysis.
In brief, exploratory data analysis is a subset of data science that assesses new data sets. Sets are scanned for missing fields, corrupted data, and anomaly detection, allowing the recipients of the data set to determine the overall quality of the data in the data set.
After conducting this analysis, an end user can assign a level of confidence to the data: if it’s really clean with few missing/corrupted fields and very few anomalies, you can feel good about the data that you’re using.
Keeping data updated and maintained is another critical component of instilling trust in your data.
This means everything from doing spot-checks and QA on databases that are routinely added to, to making certain that newly acquired databases are relatively free from errors before they are added to any of your systems.
For newly acquired data, it helps to have an assessment plan in place that is deployed any time a new data set is considered.
Using the data verification and cleaning steps outlined in the first part of this article, assess data on a routine basis.
Data maintenance is an area that is ripe for the use of Artificial Intelligence, and I expect that in the near future, the data sets that communicators use—particularly things like media data, sentiment, and message identification and assessment—will be parsed and assessed by AI with increasing frequency, freeing up human analysts to do the comparative analytical work that leads to insight and strategy recommendations.
The better shape data is in, the more trust you’ll be able to place in your analysis.
This is particularly important because sometimes data can upend long-established beliefs. To get the buy-in that you’ll need from executives and even other departments, having trust in your data will allow you to be more confident in the recommendations you are making based on that data.