Marketers have always claimed to listen to customers. Now that we have a chance to listen to social activity and feedback, we find that it not only overwhelms us with volumes of data, but it forces us to approach customer relationship management (CRM) and automation in a new way.
Using social sentiment data in direct marketing outreach requires an adjustment in practices. Data from social listening is fundamentally different primarily because it’s not tied to an individual, and may not even be tied to an audience profile. It’s also hard to understand and manage because it’s often generated on a web property not owned by the brand. In short, it’s hard for marketers to adapt to operating in the fluid state of social data when we are used to operating in the precise world of CRM data.
This gap is very real. If the audience member is in the database, then we know what to do with her. If she’s not in the database, then we either have to create an incomplete audience record or ignore the data because it doesn’t fit to our established CRM norms. For example, I can capture data from Twitter that includes the person’s handle and keywords from the post, but that record won’t include all the data typically used for CRM like email address, first and last name, or customer ID. This record could be a duplicate of someone already in the database. It could be a high value customer or could be a person of no value to my business.
“That feels limiting to me,” a marketing VP at a financial services company said to me the other day. “I don’t know how to make that linkage in my current environment.”
It’s easy to say, “Well, then. Change your environment.” But that is oversimplifying the reality for most enterprise marketers and retailers. Consider these some options:
1. Match the data as best you can.
Some marketers are comfortable making a match in their database when a certain number of attributes match up – usually three to five attributes. However, this is a bit risky because the email address, last name, postal address, and even customer ID/member loyalty card number could be the same for a married couple or a parent/offspring. If there’s no customer ID but other attributes match, we take the risk of making the match and then being unable to separate it later. There is nothing worse than making invalid assumptions about customers. While I believe – and data supports – that consumers are comfortable with marketers using data to create custom experiences (in fact, I think many consumers expect this level of personalization), they want us to use that data responsibly and with integrity. A mismatch makes marketers look foolish, or worse, irresponsible.
2. Keep the records separate.
The financial services vice president encountered this problem. “Individuals” who are not known by a CRM attribute – email address, customer ID, etc. – are impossible to target in traditional ways. Similarly, website visits generate visitor records that are missing personally identifiable attributes. That brings us to the next choice.
3. Develop content for different levels of “knowability.”
Segmentations can be set up to target fans, visitors, or followers by keyword, recency of activity, or by social network. That could have some value – a technology firm might want to send its latest whitepaper to people who have tweeted certain keywords. Of course, some of those people might be current customers, or competitors, so it’s not an exact science. Even with full accessibility to Facebook resources, it’s largely a manual process right now to identify and respond to those people individually.
4. Maintain non-identified data for a short period of time.
Want to test the value of social sentiment data to your CRM efforts? Think of it in short-term conversations. Instead of a traditional CRM approach of planning ongoing communications, consider if there’s value to use the CRM database to generate the right set of content based on your current propensity models. If there are 50 conversations going on now around the need for integrating social sentiment data into CRM practices, then use the CRM application to determine the best content based on response to various messages in other channels. Use the propensity model to determine if this is a conversation that often leads to nurturing or movement through the sales cycle. That way, when we do reply to those 50 conversations, it will have the highest impact.
5. Develop unified reporting.
One aspect that can be improved for any of these choices is to bring data into one database so that cross-channel reports can be run. This gives you a 360-degree view of the marketplace, if not the individual customer.
As always, marketers are looking for marketing automation and eCRM software to help absorb and make sense of social data. It’s relatively easy to load data into a database and run some models on it. That can be valuable to guide content and frequency decisions. However, the next step is for the software applications (or a series of applications working together) to connect the dots on identifying audience members and managing the publishing of content across channels. I’d love to hear your ideas on how you are making these choices, and if you’ve found some amazing technologies that help you address this issue.