Hubo Enjoys Growing Results From Site Search With AI Automation And Scalable Improvements

About Hubo

Hubo is a Belgian home improvement and DIY retail chain with more than 150 stores all around the country. It is a well-known brand in Belgium, taking the first place in the DIY market. Within this market, Hubo shows a continually increasing turnover and an actively growing market share. Hubo always seeks for innovation and offers an extensive and high-quality range of branded products.

All of the Hubo physical stores have ideal accessibility: stores are located outside the center, along with significant access roads, and have a large parking lot. This, plus the full range, the excellent price/quality ratio, and the clear shop layout make Hubo a favorite attraction for the do-it-yourselfer. Hubo.be online store is following the same rules of proximity and price/quality ratio. Online customers can choose between home delivery and picking up their goods in physical stores.

The results

Hubo Experienced Boost Of Conversions From Site Search Immediately After Implementing SearchNode

Right after SearchNode’s smooth and rapid implementation, Hubo saw that their conversion rates from site search shoot up by 7% compared to their previous solution. Since then, the results have only continued to grow. One year after the SearchNode was launched live, conversions increased 4% more, and site search usage itself grew by almost 10%, as well as CTR, which grew by 28%. Hubo enjoys the fruits of sophisticated site search and sees the growing results made possible by SearchNode’s continuous process of search improvements.

Challenges

Poor Search Relevance
Hubo site search failed to provide users with accurate search results. Its search engine struggled with spell check, comprehending queries with attributes, description information, and long search queries.

No Natural Language Processing
The company needed a sophisticated search solution, which could deal with intricate language specifics of French and Dutch; unfortunately, the previous search solution was failing miserably.

No Automation
The Hubo search team was doing a lot of hardcoding by manually trying to fix such primitive issues, as putting redirects, adding synonyms, etc. However, no matter how many hours they spent, the search was not improved at the core. Even more, these manual fixes created loads of search-debts.

Looking for solution

Before they implemented SearchNode, Hubo was using Adobe Search & Promote tool. But with the growth of the DIY and home improvement industry in Belgium, they needed a more effective solution, which would deal with their search challenges.

They started their discovery process by reviewing and talking with various SaaS providers (Fact-Finder, Algolia, Attraqt (former Fredhoper), and a hybrid between SaaS and consulting partner – SearchNode. They also considered building their search in-house, using Elasticsearch or Solr. Eventually, after careful investigations, and product demonstrations, Hubo shortlisted only to options: Fact-Finder and SearchNode.

Proof of concept

The company asked the two finalists to build a demo for their current online store, as a proof of concept. As soon as the company provided its data export, we started doing precisely that. SearchNode search engineers began with initial advanced data processing. After that, we collected feedback from Hubo, which allowed us to do a second iteration, with updated data relevance algorithms.

And that was it

Advanced SearchNode product won over the competitor, as the company saw how much their overall site search performance improved, just after initial demo implementation. They realized the power of scalable search improvements immediately, as they saw that even though they asked us to fix only a few specific problematic cases, our search relevance algorithms automatically fixed all analogical cases.

” We wanted to give our clients a good search experience on our eCommerce website, and SearchNode was the best option for the task. The implementation of the tool was very easy and fast. Search quality and relevancy based on AI and NLP creates an added value to our website. We saw CTR and CR have augmented since the implementation. ”

Bart Peeters
Webmaster at Hubo.be

The implementation of SearchNode

As with all our customers, the integration of SearchNode has been designed so that Hubo requires allocating the least development effort possible.

The Hubo.be search team’s key role was to communicate and review site search objectives and functionality with SearchNode to ensure it matches business goals and expectations, and to provide feedback during the implementation process. From a technical perspective, they only needed to provide data export API according to our requirements. SearchNode did all the rest – from data analysis and optimization, tracking scripts to search UI creation and custom AI algorithms.

The full advanced search solution was released in seven weeks from the kick-off workshop. Our search experts keep collaborating with the company as an internal part of their search team, working as though the organization has in-house search engineers to guarantee growing results.

Solutions created for Hubo

Solutions created for Hubo

Improved search relevance

Data Relevance Algorithms

We provide our clients with AI Search Platform, which, at its essence, already has advanced features and ensures great search experiences, but we want our clients to have the best possible search experiences and relevance. Thus, we build tailor-made data relevance algorithms unique to each client.

For Hubo, just like any other client of ours, we created custom algorithms, based on their requirements and specifics to support their overall business goals with the site search. The process started with dedicated SearchNode search developers analyzing Hubo.be product data and users’ behavior.

Based on this data, our search engineers built additional layer of algorithms.

But the process didn’t stop there, as SearchNode search experts keep working together with Hubo on continuous search improvements. Our team keeps on analyzing the data and releasing new Hubo-specific algorithms, ensuring that the site search is always being improved in a scalable way. Combining the SearchNode AI search platform with custom algorithms and continuous improvements ensures that Hubo has the highest possible search relevance.

Natural Language Processing

Hubo is a Belgian company, focused on the local market, meaning that understanding Dutch and French languages for site search is more than crucial. Their previous site search solution failed to comprehend complex grammar rules, and other nuances of both local languages correctly. As SearchNode has sophisticated Natural Language Processing (NLP), and custom stemming algorithms, fortified with AI, we solved this challenge with ease. This resulted in more relevant search results, more satisfied users, and improved conversions.

Optimized Long-Tail Queries

Before the SearchNode implementation, the Hubo.be search team used to do a lot of manual work. With no automation, they only put their search improvement efforts to the most frequent queries. However, these queries were only the minority compared to all our search terms that users were typing. The majority of queries were left alone in the long-tail. SearchNode implemented automated algorithms, powered with machine learning, which optimized all queries, not only the most frequent ones. This allowed Hubo to maximize its revenue potential for all users’ queries, no matter how rare or complex they are.

Advanced Category Pages

As Hubo.be also struggled with poorly structured and developed categories, they decided to implement SearchNode into their category pages as well. This solution came together with contextual and dynamic search filters optimized using mobile-first thinking. This helped Hubo.be in providing a faster and better search experience for its users, with smoother discovery.

Fully Automated Solution

Hubo.be stands out with complicated categories because of the variety of its goods. The company has a very sophisticated product catalog, with long descriptions, and lots of technical data. Thus after the initial launch, we focused mainly on making search relevance even better so that the search could start understanding more and more complex queries. The main goal was to make everything work automatically, by empowering tailor-made AI algorithms and making scalable search improvements.

After SearchNode was implemented Hubo search team left the days of monotonous manual work behind. From hard-coders, they became strategists, whose primary goal is to plan future improvements together with the SearchNode team. The company even had to restructure its search team by integrating our search experts into it. All this allowed the company to have best-in-class search experience, and they can be sure that with  continuous search improvements, their search will keep getting better.

” What initially impressed us with SearchNode, was its ability to fix many similar search issues with one improvement in our algorithms. Before them, we spent a lot of time repairing our search by hand, just to keep it up to date, which was costing us lots of money, time, and resources. SearchNode provided us with automation, relevance, and growing results – precisely what we want in site search. ”

Marc Henkens
Ecommerce Manager at Hubo.be

Continuous Search Improvements

For Hubo.be, we have a continuous cycle of search improvements running smoothly to ensure increased site search results. SearchNode ‘s dedicated search experts work with the Hubo search team to plan, examine, introduce, and test new enhancements to the search. The fundamental goal of this collaboration is to improve site search of Hubo.be in a scalable way, by adding data-driven adjustments to core search algorithms. With Hubo, we organize a continuous process of search improvements in monthly sprints, which can be divided into two parts, by type.

1st Part: Ongoing Search Maintenance

We allocate some development time in each sprint to ensure everything is running impeccably. SearchNode ‘s dedicated search engineers are committed to ensuring that all the bugs that have arisen are resolved without delay and that all of the store’s data is quickly updated. Also, some fraction of the time is devoted to the implementation of UI changes. Finally, this time is allocated to addressing the unforeseen or critical tasks as well. By doing this, we take steps to ensure that all aspects of the Hubo.be site search function seamlessly, and can respond rapidly to any urgent task.

TYPES OF FIX

• • Synonyms

• • Quick fixes

• • Minor UI changes

• • Promotions, landing pages, redirections

• • Product priority adjustments

ROLES

• • Both: Hubo team or SearchNode team

GOAL

• • To quickly fix individual cases

IMPACT TO SEARCH METRICS

• • No significant improvement

2nd Part: Scalable Search Improvements

The second fraction of time in each sprint is devoted to the development and implementation of new features and capabilities. Hubo and SearchNode teams put together a hypotheses list of search improvements at the start of each sprint. Next, a Hubo product owner looks through the entire list, prioritizes search improvements, and decides what will be evaluated during the sprint. Following this, dedicated SearchNodes’ engineers design tests for these improvements, to examine the performance of different variants and take necessary action, if needed.

TYPES OF FIX

• • Relevance algorithm improvements (fixing a pattern of issues)

• • Ranking strategy improvements

• • Merchandising automation

• • New search features

• • User behavior analysis

• • A/B testing

ROLES

• • Hubo manages priorities and sets goals

• • SearchNode helps to identify nonperforming areas and raise hypotheses

• • SearchNode carries the implementation

GOAL

• • Making impactful search improvements

• • Reducing the need for manual work

IMPACT TO SEARCH METRICS

• • Search metrics rise over time

Continuous Search Improvements Preserves Growth

The continuous improvement process ensures that the search is always optimized to achieve the best possible business results for Hubo and the best search experience for its clients. As we continue working together with the company as one team, the site search is continuously improved in the most scalable and efficient way. In each new sprint, SearchNode engineers develop custom algorithms specific only to the Hubo case. Thus, we enrich our AI-based search platform with custom algorithms, which maximizes the company’s search potential and earnings from site search.