The Challenge of Quantifying Data Accuracy
Big data is changing the world of sales and marketing. You no longer have to build generic advertising, rely on surveys to measure market share, or use a phone book to find new prospects. You have the potential to access all the information in the world at your fingertips.
Take a deep breath and really think about that. Staggering, isn’t it?
Data is one of the many tools actively changing the way businesses work. Even analytics programs, one of the most basic marketing tools, would be impossible without collecting and processing massive amounts of data.
For most companies, data is a great purchase that can transform the way your sales and marketing work, but it’s also a big investment. You have to ensure your teams are fully prepared to invest time into learning and implementing a new data platform. They’re expensive, and if you aren’t taking full advantage of your data, you’re wasting money.
It’s also important that you know what we’re doing to improve data quality on platforms like EDA and RigDig BI. Improving data quality is especially challenging for any company trying to match data to specific business entities.
Today, let’s discuss the unique problem of quantifying data accuracy for B2B data.
What is data accuracy?
Data accuracy is a term that describes how well your data reflects the real-world object it represents. For a simple example, we’ll discuss contact accuracy. A virtual contact data point purports a contact works at a specified company. If you pull 100 contacts from a data platform and 85 contacts actually work at the companies listed, you have a contact accuracy score of 85%.
That seems simple, but it can get quite complicated. Contact accuracy is difficult to measure due to a number of reasons we will explore.
Testing contact accuracy
First of all, to determine the quality of your contacts, you need a large enough sample size to have a high confidence score. Basically, a large sample size will give you a more accurate picture of accuracy. A sample size of 5 or 10 contacts, for instance, is not enough to truly ensure accuracy. You may have inadvertently chosen bad contacts.
Instead, a sample size needs to be at least 100. But to get a good idea of how big your sample size should be, you can find sample size calculators through Google search.
This will give you an appropriately sized audience with a high confidence score.
Once you have an appropriate sample size, the best way to verify contact quality is by phone. Unfortunately, this presents another set of complications.
While you may think it’s as simple as picking up the phone and making sure the contact is still there, it’s not. Contacts aren’t any easier to reach for verification than they are for a sales call, and we all know how challenging that can be. There are instances where a phone number has been disconnected or wasn’t answered. For the latter, you can leave a voicemail, but you’re still reliant on them calling you back.
That’s just one attempt.
If you’re using a call center, this can be expensive and unreliable. Call centers tend to have a wide range of capability within their staff. It’s also critical to ensure that the call centers you work with are trained to appropriately handle a phone tree in the limited time they are given. Often, due to the nature of the call center, they aren’t able to be patient and fully ensure a contact is valid.
Calculating Contact Accuracy
To quantify contact accuracy, you must define a call disposition table. This will tell you exactly what qualifies as a good or bad contact from your sample, and it helps you set a standard. For each item in the table, you define if the disposition (outcome of the verification action) counts as positive, negative, or neutral. Here is a sample list:
- Company correct, but not the individual
- Contact works directly for the company
- Name is correct (may include nickname, shortened names, and alternative spellings)
- Contact is verified by a person
- Contact is verified by a telephone directory entry
- Contact never worked for the company
- Contact has left the company
- Contact is a consultant for the company
- Contact is deceased
- Contact works for the company but we have the wrong location
- First, middle, or last name are missing or abbreviated to one letter
- Bad phone number
- No answer
- Refused to answer questions
- Hang up
- Do-not-call suppression prior to calling
- Do-not-call claimed
- Caller couldn’t navigate the phone tree
- Caller placed on hold too long
- Wrong contact job title
- Contact spelled incorrectly
- Duplicate contact
You place each contact in one of the three disposition groups, and then calculate your accuracy. For example, with the criteria above, you might make 200 calls to get the following results.
- 80 positives
- 30 negatives
- 90 neutrals
In this case, here is what the contact accuracy calculation would look like:
One problem with disposition tables, however, is they aren’t standard. Every call center (and companies that maintain one), have a different disposition table. This can lead to different scores according to the different call centers being used presently or used in the past. This is a natural result because each company has different criteria for what they deem a useful contact.
For instance, the table above doesn’t distinguish the contact’s role. For some organizations, they might only consider a contact accurate if that contact’s role is correct, such as a CEO or human resources manager, whereas the next company doesn’t care about the role.
This also leads to a complete lack of industry standards for benchmarking.
Big data is changing the world of business, and you realize that it’s necessary. You have to have more in-depth knowledge of your customers and prospects. That’s why you’ve invested the time and resources into implementing the platform and the processes necessary to succeed.
Without data, you’re moving slow compared to the competition. But that means we have to be constantly aware of the current quality of our data so that you are getting the most accurate information possible.