The Identity Resolution Problem
There are many vendors that offer Identity Resolution (aka matching of structured records).
And all vendors have the same goal which is to create the most accurate outcome possible.
Accuracy is always measured by the client which is to say what they believe constitutes a match is what drives accuracy.
One-size fits all matching solutions do not take the client into account and means the vendor is dictating what a match constitutes and not their customer.
At it's core all solutions are trying to achieve one central concept and that is to be able to calculate the complex balance between comparing each attribute such as Name (John Smith v Jon Smith) and then balancing the outcome across all attributes.
A short example: Imagine two records that have 5 attributes; Name, Address, SSN, DOB, Phone. It's easy to understand that if all attributes exactly match then the records are a match. And conversely if all attributes are completely different they are not a match.
But between perfect and imperfect there are 10's to 100's of thousands of variations that can occur and some of those variations are matches and the others are not.
How do you create a solution that CORRECTLY captures the nuances for each client?
(Reminder that accuracy is measured by the customer not the vendor)
For decades there were generally two approaches to this problem; deterministic and probabilistic.
Deterministic is a rules-based solution where rules are created to guide the outcomes. While a single rule can address more than one variation it's incredibly difficult to be able to create an effective set of rules that can handle the 10's-100's of thousands of variations.
Probabilistic is a statistical approach to solving the problem. This creates a better balance between solution and outcome than rules but is still vastly limited by the linear approach and cannot address all variations.
Machine Learning has the ability to learn the nuances and create a solution that addresses far more variations than the first two.
What are the 5 most important areas clients look for in an Identity Resolution Solution?
Easy to Use
Deterministic/Probabilistic: Take months to years to implement and require a large project team.
EntityWise(Machine Learning): Can be implemented in weeks.
Easy To Use
Deterministic/Probabilistic: Require expensive expert service teams to create incredibly complicated algorithms
EntityWise: Can be implemented by the client.
Deterministic/Probabilistic: Cannot make adjustments once the solution is live without significant effort.
EntityWise: Can be changed on the fly.
Deterministic/Probabilistic: Struggle to gain acceptable levels of accuracy.
EntityWise: Has the best accuracy.
Deterministic/Probabilistic: Are incredibly expensive, with implementation project costs, maintenance fee's, infrastructure costs, and upgrade expenses.
EntityWise: Is the most cost effective solution because of it's simple architecture and easy implementation.
In our 15+ years of experience we have used all 3 approaches and can definitively state that Machine Learning is by far the best approach.
Unable to Render Intelligent Decisions
(All Other Solutions)
New Technology Capable of Mirroring Human Decision Making
Schedule a 15min Call to Learn More
Why Machine Learning Is
a Better Technology
Can I trust Machine Learning Technology?
Absolutely. Just as you trust your current match platform, machine learning is a better and more accurate technology than older approaches.
As experts in match, we have experience in all 3 match technologies; deterministic, probabilistic , and machine learning and we can say that machine learning is by far the best approach to the match problem.
All traditional match platforms, IBM, Epic, Cerner, Meditech, Informatica etc... use old technology to try to solve the problem.
Will Your Approach Meet Our Match Requirements?
Absolutely, Our approach customizes it's approach to meet your needs better than any other approach on the market.
Goal: The number one goal to a match approach is its match accuracy. But how data is matched is seen differently by each customer which means a custom match approach, tailored to their needs, is an absolute requirement for the best accuracy. One-size fits all solutions will not achieve the same level of accuracy.
Is this approach reliable?
Absolutely, once trained and tested the answers will always be 100% consistent.
Machine learning is a non-linear technology that can apply match logic that mirrors a customer's requirement much closer and can do so through a simple use of data samples rather than configuring algorithms.
Other match platforms, IBM, Informatica, Epic, Cerner, Intersystems, Quadramed etc… create tasks, which are potential matches that are left by these platforms for humans to resolve. EntityWise steps in and resolves them, making the machine learning approach more accurate.
Does your approach work with other MPI/EMPI Platforms?
Absolutely, our solution can work with other technologies and resolve those matches the other platforms cannot.