Leading market research firm, Gartner, suggests that 40% of all business initiatives lose value because of incorrectly linked, or messy data. Data deduplication and record linkage are two sides of the same coin. While the applications of both vary widely, the underlying techniques used to identify matching records for both data cleansing/deduplication and record linkage are the same.

Whether you want to identify duplicates before migrating to a new CRM, or want to build a Single Customer View under an enterprise-wide digital transformation initiative, you will have to perform ‘data matching’: the ability to identify all records that point to the same entity within and across data sources. Easier said than done though.

In this blog, we will take an in-depth look at fuzzy matching, the go-to approach for data deduplication and record linkage. We will cover:

  • Data Matching: Deterministic and Probabilistic Matching
  • What is Fuzzy Matching?
  • How to Minimize False Positives and Negatives
  • Why Do Businesses Need Fuzzy Matching?
  • Fuzzy Matching Techniques
  • Example of Real-World Fuzzy Matching Scenario

Data Matching: Deterministic and Probabilistic Matching

We know we need to match records to identify duplicates and link records for entity resolution. But how exactly do we go about identifying matching records? What properties should we focus on?

Let’s start with ‘unique identifiers’. These are properties in the records you want to match that are unlikely to change over time, Customer Name for instance. You can assign weights to each property to improve your matching process. Think about it; if you are migrating customer data from one system to another and need to check for duplicates pre- and post-migration, you could, for instance, choose Name as the one unique identifier and phone number as the second. Now it’s just a matter of running a search for matching Customer IDs and phone numbers and you have all potential matches identified.

That method is known as ‘deterministic data matching’. Although effective in theory, the method is rarely used because of its inflexibility: The approach assumes that all entries are free of mistakes and standardized across systems – which is almost never the case in real-world linkage scenarios.

In our previous example, if some phone numbers have country code in the ‘+1’ format and the rest start with ‘001’, the matching would go awry. That’s just one instance; there could potentially be dozens of different ways data could be entered. The names might be misspelled, acronyms used, middle name included, etc. In one system, a customer’s name could be ‘William Warner’ while another might have ‘Williaam Warner’ —  it’s obvious that there’s been a small typo and both are in fact the same customer — but the method only allows discrete outcomes, that is, all or nothing.

How do you go about determining a match when so many variations exist?

By performing probabilistic data matching, that’s how. More commonly known as fuzzy matching’, this approach permits the user to account for variations like spelling errors, nicknames, punctuation differences, and many more by combining a variety of algorithms.

What is Fuzzy Matching?

Rather than flagging records as a ‘match’ or ‘non-match’, fuzzy matching identifies the likelihood that two records are a true match based on whether they agree or disagree on the various identifiers.

The identifiers or parameters you choose here and the weight you assign forms the basis of fuzzy matching. If the parameters are too broad, you will find more matches, true, but you will also invariably increase the chances of ‘false positives’. These are pairs that are identified by your algorithm or fuzzy matching software of choice as a match, but upon manual review, you will find that your approach identified a false positive.

Consider the strings “Kent” and “10th”. While there is clearly no match here, popular fuzzy matching algorithms still rate these two strings nearly 50% similar, based as character count and phonetic match. Check for yourself

False positives are one of the biggest issues with fuzzy matching. The more efficient the system you’re using, the fewer the false positives.  An efficient system will identify:

  • Acronyms
  • name reversal
  • name variations
  • phonetic spellings
  • deliberate misspellings
  • inadvertent misspellings
  • abbreviations e.g. ‘Ltd’ instead of ‘Limited’
  • insertion/removal of punctuation, spaces, special characters
  • different spelling of names e.g. ‘Elisabeth’ or ‘Elizabeth’, ‘Jon’ instead of ‘John’
  • shortened names e.g. ‘Elizabeth’ matches with ‘Betty’, ‘Beth’, ‘Elisa’, ‘Elsa’, ‘Beth’ etc.

And many other variations.

How to Minimize False Positives and Negatives

We have discussed false positives in the previous section briefly. While they make matching more difficult by adding manual review time to the process, they’re not a genuine risk to the business because the system will flag false positives based on the overall match score.

Let’s take a look at ‘false negatives’ now. This refers to matches that are missed altogether by the system: not just a low match score, but an absence of match score. This leads to a serious risk for the business as false negatives are never reviewed because no one knows they exist. Factors that commonly lead to false negatives include:

  • Lack of relevant data
  • Significant errors in data entry
  • System limitations
  • Match criterion is too narrow
  • Inappropriate level of fuzzy matching

The most effective method to minimize both false positives and negatives is to profile and clean the data sources separately before you conduct matching. Leading data matching solution providers typically bundle a data profiler that quickly provides enough metadata to construct a cogent profile analysis of data quality, as in missing values, lack of standardization, any other discrepancies in your data. By profiling your data, you can quickly quantify the scope and depth of the primary project, whether it’s Master Data Management,  matching, cleansing, deduplication, cleansing, or standardization.

Once you’ve profiled your data, you will know exactly which business rules to apply to clean and standardize your data most efficiently. You will also be able to quickly recognize and fill missing values, perhaps by purchasing 3rd party data.

Cleaner, more complete data reduces false positives and negatives significantly by increasing match accuracy because your data is now standardized. The fuzzy matching algorithms you use, the matching criteria you define, the weight you assign to different parameters, the way you combine different algorithms and assign priority – these are all important factors in minimizing false positives and negatives too. But none of these are going to help much if you haven’t profiled and cleaned your data first. See how DataMatch Enterprise has helped 4,000+ customers in over 40 countries clean, deduplicate, and link their data efficiently.

Why Do Businesses Need Fuzzy Matching

Research reveals that 94% of businesses admit to having duplicate data, and the majority of these duplicates are non-exact matches and therefore usually remain undetected. Fuzzy matching software helps you make those connections automatically using sophisticated proprietary matching logic, regardless of spelling errors, unstandardized data, or incomplete information.

But it’s not just about deduplication. From a strategic perspective, fuzzy matching comes into play when you’re conducting record linkage or entity resolution. We touched upon this briefly in the previous section too; the fuzzy matching approach is invaluable when creating a Single Source of Truth for business analytics or building a foundation for Master Data Management (MDM), helping organizations integrate data from dozens of different sources across the enterprise while ensuring accuracy and minimizing manual review. See how a major healthcare provider was able to save hundreds of man-hours annually.

Here are some ways that fuzzy matching is used to improve the bottom-line:

  • Realize a Single Customer View
  • Work with Clean Data You Can Trust
  • Prepare Data for Business Intelligence
  • Enhance the Accuracy of Your Data for Operational Efficiency
  • Enrich Data for Deeper Insights
  • Ensure Better Compliance
  • Refine Customer Segmentation
  • Improve Fraud Prevention

Learn more about the benefits of fuzzy matching

Fuzzy Matching Techniques

Now you know what fuzzy matching is and the many different ways you can use it to grow your business. Question is, how do you about implementing fuzzy matching processes in your organization?

Here’s a list of the various fuzzy matching techniques that are in use today:

  • Levenshtein Distance (or Edit Distance)
  • Damerau-Levenshtein Distance
  • Jaro-Winkler Distance
  • Keyboard Distance
  • Kullback-Leibler Distance
  • Jaccard Index
  • Metaphone 3
  • Name Variant
  • Syllable Alignment
  • Acronym

Get more information on fuzzy matching algorithms here

Example of a Real-World Fuzzy Matching Scenario

The following example shows how record linkage techniques can be used to detect fraud, waste or abuse of federal government programs. Here, two databases were merged to get information not previously available from a single database.

A database consisting of records on 40,000 airplane pilots licensed by the U.S. Federal Aviation Administration (FAA) and residing in Northern California was matched to a database consisting of individuals receiving disability payments from the Social Security Administration. Forty pilots whose records turned up on both databases were arrested.

A prosecutor in the U.S. Attorney’s Office in Fresno, California stated, according to an AP report:

“There was probably criminal wrongdoing.” The pilots were either lying to the FAA or wrongfully receiving benefits. The pilots claimed to be medically fit to fly airplanes. However, they may have been flying with debilitating illnesses that should have kept them grounded, ranging from schizophrenia and bipolar disorder to drug and alcohol addiction and heart conditions.”

At least twelve of these individuals “had commercial or airline transport licenses,” the report stated. The FAA revoked 14 pilots’ licenses. The other pilots were found to be lying about having illnesses in order to collect Social Security payments.

The quality of the linkage of the files was highly dependent on the quality of the names and addresses of the licensed pilots within both of the files being linked. The detection of the fraud was also dependent on the completeness and accuracy of the information in a particular Social Security Administration database.

See how companies in your vertical are using fuzzy matching today

Fuzzy Matching Made Easy, Fast, and Laser-Focused on Driving Business Value

Traditionally, fuzzy matching has been considered a complex, arcane art, where project costs are typically in the hundreds of thousands of dollars, taking months, if not years, to deliver tangible ROI, and even then, security, scalability, and accuracy concerns remain.

That is no longer the case with modern data quality software.

Based on decades of research and 4,000+ deployments across more than 40 countries, DataMatch Enterprise is a highly visual data cleansing application specifically designed to resolve data quality issues. The platform leverages multiple proprietary and standard algorithms to identify phonetic, fuzzy, miskeyed, abbreviated, and domain-specific variations.

Build scalable configurations for deduplication & record linkage, suppression, enhancement, extraction, and standardization of business and customer data and create a Single Source of Truth to maximize the impact of your data across the enterprise.

Download the datasheet and see how we can help your business grow!