2

Walk Before You Run – Social Media Data as the MDM Silver Bullet

The Silver Campaign Bullet

Recently I had a number of meetings with clients who are looking at social media to enrich master data in their legacy data warehouse to improve their batch and real-time campaign effectiveness. They are primarily looking at Facebook as their marketing salvation to create as close to one-on-one offers as possible but there are multiple ways to skin this cat.

Surprisingly, I have seen this thinking primarily in emerging markets. It reminded me how some countries jumped from having a poor fixed line phone infrastructure catapulting themselves to 3G wireless for all citizens within a decade. The decision makers who were part of this overnight communication transformation assume that master data management (MDM) works the same way. 

As such, it appears to me that some IT and marketing leaders are conspiring to push the technology envelope of what is definitely doable, yet financially and practically questionable given the expense and value derived but more importantly, given the current state of their master data. They are thinking about phase two or three instead of phase one in their MDM initiative.

As marketing executives ask for more data to get more out of their campaigns their IT liaison is happy to oblige. Let’s remember, MDM is an organizational journey with a number of sprints in between not a single 100-meter dash.  First, clean the data you have in phase one before you throw “good” social data on top of “bad” legacy data. If it is not done right a firm may not only have wasted precious resources or soured the corporate MDM experience for years to come but it may also face the wrath of its consumers who now “dislike” your campaigns within a second.

Firms could use just a customer name and email to scour existing structured sources outside the confines of their company, which they may be using (often without knowing) including census, tax or professional affiliations databases. They could also tap structured Facebook profile data by scraping. Instead, they want to actively mine unstructured posts and crawl the web for customers’ updates. They look to widely available scraping apps to create a more complete view of consumers, social relationships and hopefully also a more timely view of life event changes, such as marriages, moves or new connections to individuals with the same last name (presumably family members). But getting this type of information from free-text posts and possibly inferring sentiment is costly and requires expertise. If this type of access would be so easy, Facebook would not be hogging it and be valued at billions of dollars.

Some would consider the rather static profile attributes, such as name, locale, birth date, age, employer, marital status, photo, groups or friend list rather stale. Is it not challenge enough to maneuver through a changing Facebook privacy policy and default settings to get to these characteristics so you don’t have to rely on unaware users having lax public profile or message settings? Why does a bank or retailer want to be in the advanced analytics software business before they standardize and de-duplicate their existing records?

It is a significant effort to collect, index, filter, search, extract and correlate meaningful “advanced” context every day or hour from an entry such as “congrats on your vows last week buddy” posted by “Jeff” on “Mike’s” wall given an existing female “friend’s” (now spouse or brides mate) post a week ago about “nuptials” to deduct who got married to who, when and who was attending. This sentence alone made my head spin. If you don’t do this every hour or day of the year, you may forgo the very reason for this exercise – to find out about people’s lives right when commercially interesting things happen.

The other question here is what types of users will actively manage their Facebook accounts to make this pay for you and are these users of commercial interest to the company given their spending power. My common sense tells me that Generation Y and Millennials will let their Facebook friends know about life changing events rather quickly, but Generation X and Baby Boomers, due to habit, IT savvy or conservative privacy beliefs, will likely update their profile very infrequently if at all. I am sure there is research out there on this and I do not presume to be a social media marketer so I encourage you only to ask these prudent questions.

I am sure stringing together multiple different semantic technologies, like text mining, glossary based algorithms, web crawlers, sentiment engines, etc. will do all this, but does your company have the stomach to really do this to get a 0.2% uptick in response rate? Why not wait (and pay) for timeline apps to track more structured attributes like “You lost an important person”, “You had a surgery”, “Your pet is XYZ”, “Your room mate is XYZ”, “Bought a new home” or “Obtained a license”? Why not use some of this new information to also deduplicate and enrich your internal prospect or customer database? In the end, you better be sure what you want to do with this data and what you want to achieve with it before embarking on the “more data is better data” express.

Given all this, I proposed an alternative approach to investigate – maybe naively. I asked how the company is fact checking its new hire applicants today. They either told me that their recruiters (inside or outside) pay a thirdparty vetting service for a “proceed/don’t proceed” flag.

Public databases already in use for background checks

This company then spends a few minutes or hours on scouring publicly or commercially aggregated data on: birth date, name, aliases, financial health, criminal record, marital status, prior employers, past residences, etc. Often these “core” databases are accessible and maintained by public agencies (see your US city’s real estate tax site) capturing everyone’s main life events: when you were born, got married, had a child, bought a house, did not pay a bill or went to prison. They are maintained by credit bureaus, department of motor vehicles, social welfare or tax agencies and the like. Most importantly, this data or resulting analytics are available, often for sale, relatively timely and structured so some simple SQL may do the trick after all. Whole companies make their living off aggregating and providing this “basic” information, e.g. Lexis Nexis.

Once you mastered the data quality “walk”, you can start running towards your cutting edge Facebook extensions.

And then there is data privacy, something very highly prized in some geographies. Legislation would typically dictate that you would have to ask every customer and prospect for their permission to use their Facebook data.  This would either happen through an email tick box embedded in a pre-Facebook campaign but most often via terms and conditions on the company’s web-based order, customer service system, credit card application or Facebook app (often location-based). It is important to gauge not what a firm can get away with but what its’ customers are willing to bear in terms of intrusiveness given benefits derived by the users.

Do you agree? Given all this, what would your approach be? How is your organization using or planning to use social media data?

FacebookTwitterLinkedInEmailPrintShare
This entry was posted in Big Data, CIO, Customer Acquisition & Retention, Customer Services, Customers, Data Aggregation, Data Governance, Data Privacy, Governance, Risk and Compliance, Master Data Management and tagged , , , , . Bookmark the permalink.

2 Responses to Walk Before You Run – Social Media Data as the MDM Silver Bullet

  1. Pingback: Social Media, MDM, and Silver Bullets | Data Daily | DATAVERSITY

  2. Peter Perera says:

    On a social MDM-related note, may want to read: “MDM’s Blind Spot: Social Networks”. Lengthy article penned last Oct, but check it out at http://bit.ly/vECuTu

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>