Zoomin's Elasticsearch Performance and Relevance Optimiation - a Success Story - BigData Boutique
arrow leftBack to Customer Stories
Customer Story

With a growing and diversifying customer base, Zoomin's critical search technology needed to modernize and enhance into a feature-rich platform. In collaboration with the Zoomin team, we delivered a future-proof and cutting-edge search platform that gives them and their customers a competitive advantage.

google cloud & kubernetes
Search Relevance Engineering and Optimization
elasticsearch
Elasticsearch Cluster Upgrade and Performance Tuning
hebrew text analyzer
Custom Development of Data Pipelines and Services

The Client

Zoomin is an innovative start-up working in the customer documentation and experience space, driving relevant technical content to end-users. With enterprises such as Mastercard, Docusign, and Dell using their products, their service must continue to evolve to meet huge customer demand. Zoomin is a leader in technical product documentation experience, and their platform is one of their key differentiators.

zoomin company

Choosing the right partner for the job

“Search and findability is a key aspect of the Zoomin product,”

says Rafi Bryl, the Director of Product Management at Zoomin.

Underpinning the complex and powerful Zoomin product is a search engine built on Elasticsearch. The original search functionality was built on an early version of Elasticsearch, which needed upgrading.

“We realized that we couldn’t do what we had planned to do, in terms of scaling, AI, and customer experience improvement, on the existing infrastructure,”

Rafi noted.

However, owing to development changes within the tech team at Zoomin, there wasn’t the right level of expertise to carry out such a complex upgrade.

Zoomin was also keen to have an iterative approach to the upgrade, where an entirely new platform would be created using elements from the original platform, but with lots of improvements and upgrades.

“Whilst we still had some Elasticsearch skills in the company, we didn’t really have ‘search engineers’”,

notes a member of the Zoomin leadership team.

Fortunately, BigData Boutique was available and on hand. Not only did they come well referenced from multiple other Elasticsearch upgrade and development engagements, but they also gave Zoomin confidence through the flexibility they were able to provide.

“It was clear that they were a good fit, for a number of reasons,”

Rafi admitted.

Instead of presenting a fixed price, or price per day, BigData Boutique provided Zoomin with a bank of hours that they could draw upon for all work, ensuring that the project wasn’t limited in scope.

“We were confident we were bringing in experts who could help Zoomin modernize and stay best in breed”.

Defining the project

When approaching the engagement, both BigData Boutique and Zoomin had a number of tough architectural and design decisions to make.

“One of the biggest challenges we faced”

remarked a member of the Zoomin tech team

“was deciding whether we build something entirely new, or use the old platform”

The Zoomin search functionality needed modernizing on the latest version of Elasticsearch, but the team was keen to see if any original features could be carried forward from the original platform.

“With the help of Bigdata Boutique, we preserved the basic schema and underlying architecture, but we took fundamental design decisions that deviated from and revolutionized the existing platform”

states a member of Zoomin’s product team. The process of working out what to persist and what to change was, in itself, lengthy and complicated, and BigData Boutique’s experience in similar projects was key.

Collaborating with BigData Boutique, and making use of their expertise, it became clear that the existing platform could be stabilized for current customers, and the new search platform would be introduced separately. This would give Zoomin’s customers the opportunity to opt-in to the upgrade, as well as provide a useful UAT sandbox for early deployments.
Crucially, the new platform delivered would have to work for all of Zoomin’s customers, some of whom would not have moved to the upgraded offering. This would require complex integrations to address interoperability between new features on the new platform and users of the existing platform.
It became clear during the engagement that the new platform needed to handle many different natural languages in a fundamentally different way to Zoomin’s original search engine. Due to Zoomin staff changes over the previous years, the Research and Development Team was still growing and did not have the linguistic processing and searching skills required.
As linguistic processing and complex search technology are pillars of BigData Boutique, Itamar, Lior, and the team were able to provide invaluable guidance on how to best build the solution, before starting development. This approach would allow the Zoomin platform to expand into multi-lingual processing, future proofing their investment and the solution.

Stepping onto a new platform

Working tirelessly in conjunction with Zoomin, BigData Boutique was able to deliver the new platform, live in production, despite several operational setbacks.

“After they brought in the expertise to make the architectural decisions, they then wrote the code to our timeframes and exactly as we needed”

Rafi states.

Whilst the global pandemic was changing timelines across the world, it didn’t hamper BigData Boutique’s delivery of this complex engagement.

“The whole project was executed as we desired and to a very high standard, even though we never met anyone from BigData Boutique face-to-face”, notes Rafi.

The ability to deliver the entire project, within the timeline and as requested without having the traditional consultancy facetime, was critical in keeping up with Zoomin’s organizational roadmap.

The new platform has full backward compatibility with the previous version, ensuring that customers can have a seamless migration to the new engine at a time of their choosing.

“Achieving functional parity between the two platforms was one of the main focuses for us, and the way in which it was done was really quite sophisticated”, says a member of the technical team.

On top of the backward compatibility, BigData Boutique’s work was crucial in introducing several new features.

Complex Boolean Search

Whilst Zoomin already had a boolean search capability, the Elasticsearch upgrades and architectural changes made by BigData Boutique enabled the introduction of more complex search terms. For example, the new platform has a “not” operator, allowing users to exclude content based on particular keywords.

Curated Search

The refinement of the curated search functionality is a real force multiplier for the new Zoomin platform, made possible by BigData Boutique’s development work. It takes any randomness out of search results, allowing users to specify what sorts of results, in what order, they would like returned for any given query.

Autosuggest Rebuild

Whilst not yet live, BigData Boutique facilitated the rebuild of Zoomin’s existing auto-suggest function. This complex machine learning-derived feature is critical for customer experience and allows for more seamless navigation of the platform. Although small, autosuggest functionality is exceptionally difficult to engineer, and even more so with the amount of text that Zoomin’s customer base is supplying.

A Vision For The Future

“We’ll absolutely maintain a relationship with BigData Boutique,”

says Rafi.

Moving forward, BigDataBoutique and Zoomin will continue to collaborate as the new platform is rolled out to more customers. BigData Boutique will continue to play a role in new feature development at Zoomin, all while knowledge transfer takes place.

There are plans for BigData Boutique to continue consulting at Zoomin across a range of different areas. There have been obvious successes in the new platform development and launch, and this is something that both parties are keen to continue.

“We found the relationship very useful - BigData Boutique was helping us solve production issues on the old platform whilst developing the new platform simultaneously, even if it wasn’t in their remit”.
We use cookies to provide an optimized user experience and understand our traffic. To learn more, read our use of cookies; otherwise, please choose 'Accept Cookies' to continue using our website.