Gain a 360⁰ View of Your Customers

Data Enrichment is the enhancing of customer-provided data (first-party data) with data from an outside source (third-party data). Another name for Data Enrichment is Data Appending. The two names are interchangeable.

When data is enriched, the first-party data in one dataset is compared with the third-party data in a separate dataset. Third-party data is collected by a company with no direct connection to the customer. Sources of third party data can include websites, social media networks, surveys, subscriptions and more. The third-party dataset always has many more records and much more information about those records than the first-party dataset. The enrichment process looks for records in the two datasets that match. 

Matches don’t have to be based solely on customer name, which can be entered inconsistently in databases. Matching can be based on categories like physical address, email address, phone number and more. When a match is found, data from the third-party dataset that is not included in the matching record is added to the first-party dataset. This creates a much fuller picture of that customer. 

Every category of customer data can be enriched. A look at three of the more popular types of data enrichment makes clear how much information it can add and valuable it can be.

Demographic data enrichment

Demographics is one of the primary methods of customer segmentation, but first-party data often leaves out important information about the customer that can be filled in by third-party data, including:

• Age

• Marital status

• Income level

• Highest level of education completed

• Type of vehicle

• Credit worthiness

The options go on and on. 

Geographic data enrichment

Geography is another important area of segmentation and enrichment. Geographic and demographic segmentation are often categorized as one type of segmentation known as geodemographic segmentation. With geographic data enrichment you can fill in quite a few bits of information, including:

• Change of address

• County

• Census tract

• Latitude & longitude

• Zip code or zip plus four

With your geographic and demographic data enriched for these details, you’d be able to target a direct mail audience with great precision. For example, suppose you wanted to reach credit-worthy female home owners between the ages of 40 and 65 who drive a domestic automobile, make at least $80,000 per year, live within a 20-mile radius of your store and are already your customer. 

Targeting that specifically might be impossible with first-party data alone, so you’d have to mail to a much larger audience just to be sure your actual target was included. If each mail piece costs $1.25 and you have to send out 10,000 — instead of the 2,000 that you need to reach your entire target audience — you waste $10,000.

Semantic data enrichment

Semantic data works by putting words in the correct context. It is an invaluable part of search, especially now that voice search is expected to reach 50 percent of all searches in 2020 (according to comScore). Semantic search interprets the context of different words in combination, so the search is narrowed to the intended request, rather than searching for all of the words in a long tail search query. 

For example, Semantic data enrichment searches text entered by customers for clues about attitudes and preferences. It then tags the customer record accordingly. So, if a customer has entered in one of their online profiles that they like doughnuts, the next time they visit your website you might recommend a post about doughnuts in your blog they likely would not have found any other way.

Two ways data enrichment is processed

Data enrichment can be processed in two ways, in real-time and by batch. 

Real-time data enrichment happens as data enters your system. So, if a customer’s level of interest in a particular product or service is reflected by online activity, their customer record is updated immediately. If it is a prospect, this new activity can automatically increase their lead score. The increased lead score could automatically generate an email or a sales call.  

Batch data enrichment is for information you don’t need urgently. Updates can be scheduled to occur at regular intervals. 

Don’t go overboard 

It’s important to limit each data enrichment project to the information you need for a specific purpose, especially when you consider that the number of ways data can be enriched is limited only by the data you have. Burdening datasets with unneeded data wastes storage space and slows down processing.

It’s also important to have confidence in the third-party data you rely on.

When you use third-party data to enrich first-party data, the two types of data have to match up. When customer records don’t match, no changes or additions are made to the data. 

But whether or not the two data sources match up is not always a question of black and white, match or no match. If black is matching and white is not matching, then the matching of each customer record can be thought of as a shade of gray. Technically, each shade represents a level of confidence in the match, with zero percent being a definite no match and 100 percent a perfect match. As the enrichment is run, each record is scored by percentage of confidence in the match.

As part of the setup, criteria need to be established that specify how different levels of confidence will be handled by the enrichment tool. Here is one common approach to handling confidence levels.

• High level of confidence | If the confidence score is 80 percent or higher, the match is automatically accepted.

• Middle level of confidence | If the confidence score is between 60 percent and 80 percent, matches are called out for manual analysis to determine whether or not they are acceptable.

• Low level of confidence | If the confidence score is 60 percent or below, the match is automatically refused

This technique is not exact, but it goes a long way to ensuring that data is not changed unless it should be. One way to limit the number of low-confidence matches is to cleanse the data before enriching it.

As noted in our article, Tune Your Data for Peak Performance, 69 percent of all companies believe that inaccurate data undermines their customer experience efforts. Missing data and duplicate data lower the effectiveness of marketing campaigns, which wastes money. 

The most common data quality problems are:

• Duplicate records. If you have multiple entries for the same customer caused by minor variations in the name, the data enrichment process won’t know which record to match with the third-party data. 

• Missing data. There likely will be empty columns, where there is no data. You need to decide how you are going to deal with them. 

• Typos. Simple mistakes can cause big problems, especially when you spell a good customer’s name or address incorrectly in the communication.

The process of cleaning and then enriching data may need to be repeated several times. The enrichment tool will call out errors in the data. You can then choose to fix, ignore or delete the record. A process of cleanse > enrich > repeat will help you reap the best results from your enrichment project.