I got to know Zingg while looking for solutions for the large enterprise entity resolution problem. There are several open-source solutions, but none could work with the massive dataset that I had. Zingg is, however, a very powerful tool to perform the task by smart blocking, interactive labeling by the user, and then classifying potential candidates into the same cluster. By running the jobs on Spark, Zingg is capable of handling very large datasets. I also got a substantial amount of help through direct contact with developers and that was a unique support experience for me.

I highly recommend Zingg for entity resolution tasks.

Zingg helped us take poor, inconsistently recorded data and simplify it so it can help us tell a fuller story of campaign finance in North Carolina. We’re now able to get a better picture of where political money is flowing to and from in the state, which would otherwise not have been possible without Zingg. The product that Sonal has built is very powerful and efficient, and the community she’s built enhances it’s usefulness even further. Users (often the creators themselves) will regularly jump in to troubleshoot issues very quickly, and new features are rolling out by the week.

Would strongly recommend Zingg if you’re looking to solve an entity resolution based issue.

If you have a fuzzy matching, entity resolution, or record linking type of problem, you really need to try out Zingg . . . especially before attempting to build your own solution or purchasing some expensive enterprise software (speaking from experience here).  Zingg's interactive approach to finding/soliciting training labels from data SMEs is unique in the OSS ecosystem and the underlying ML models just flat out work.  Plus, Zingg runs on Spark, so we know it will scale (obligatory plug for Databricks here!).

I've been continuously impressed by Zingg's performance across multiple verticals (e.g. Public Sector, HLS, Retail, & FinServ) and how quickly it can converge on a model, even at large scale.  

Moving forward, I'll be leveraging Zingg with Databricks for more and more use-cases, like Procurement, Customer 360, M&A, Churn-Risk, FWA, etc.

When I put some effort into figuring out Zingg, I was able to solve problems and complete work that I thought would only be possible by adding a new team member or outsourcing. Zingg brings the power of machine learning for the purpose of entity resolution to the no-coder, and when I got stuck, there was never a question that the Zingg community couldn't answer.

Hey Team! Just wanted to update on the issue. I was able to resolve it and did a POC. You guys have built an awesome thing! Already starred the repo. Great work team.

Thanks for making it open source and providing this kind of personalized support.

I just wanted to express my thanks to the Zingg team for being so responsive to my suggestions/feedback/requests! The ability to export labeled data, link multiple datasets, and output the same z-cluster id has made Zingg a very powerful tool for my e-commerce project. I hope everyone finds the improvements useful for their own projects.

Sonal has been quick to understand the obstacles/challenges I presented and took swift action to improve Zingg to resolve the challenges to help get to my desired results. For example, the feature to use pre-existing training data suggested by @Degi directly prompted me to request a feature to export labeled data from Zingg and then import it to a new model/instance using the newly created pre-existing training data feature.

Most notably, both feature requests were completed in under a week! If you have any feedback or suggestions for Zingg, don't hesitate to bring it up because it's likely the problem you are facing will help others and it's the easiest and most effective way to contribute and make Zingg better!