E-commerce search budgeting: Where to invest first?

E-commerce search budgeting: Where to invest first?

24 Jan 20228 min read

E-commerce search is an extensive domain, and the computing efforts to on-site search tuning can be rather intensive. But do all the search functions require the same amount of resources to implement? Where should you start when improving the search experience on your website, and what will bring you the most value at different steps?

At Wise Engineering, we've been working with E-commerce site search implementation and tuning for 15 years, and here's a general roadmap we recommend to our customers. It will help you approach search budgeting more strategically, see where you're now, and where you're going in the long run.

How to approach search budgeting strategically

As E-commerce companies begin to realize the benefits of improved search experience to their customers, it is now always clear how to start the optimization process and how much resources will be needed for the configuration of each particular feature or selected optimization.

We’ve divided the whole process of enterprise search tuning into levels depending on the engineering efforts they require. Though we do not provide the exact numbers, you can clearly see what will cost you the most. The higher on the pyramid you go, the more advanced the task is, and the more budget you will need to allocate.

For us, the pyramid also illustrates the logic roadmap of the improved E-commerce search. You can easily start with the first levels, measure the results and move up the pyramid towards the advanced search experience for your E-commerce platform.

This roadmap is applicable to all existing search engine solutions you might currently use – Elastic search, Algolia, Solr, and others. Some features require a few configurations of the search engine and they are on the initial layers of our pyramid. Some search capabilities will be engineered specifically for your business and are located on the top.

E-commerce search budgeting

Let's analyze each of the levels in detail.

Level 1 – Basic search

The first level usually starts with UX improvements and functionality available out-of-the-box by most search engines and requires a few simple configurations that don’t take much developer’s time. Before starting any significant and time-consuming configurations, try enabling the autocomplete feature, compose separate thesaurus specific to your E-commerce segment, configure a fuzzy search and stopwords filtering. Measure the performance and decide on further investments.

  • UX improvements. They can, for example, include: creating a visible search box, changing the search box location, highlighting search results, using microcopy for a search bar, showing the number of results, keeping recent user’s queries, and more.
  • Autocomplete. Do not consider this a simple feature that reduces typing efforts for customers. Despite the simplicity in the configuration, autocomplete allows your customers to express their search intent aligned with the query understanding of your online store. This helps you increase the search quality with minimum investments.
  • Synonyms. For every industry you’re selling products or services to, there is a specific synonym range your on-site search should be able to interpret. You can compose a thesaurus specific to your business and add it to the search engine configurations.
  • Fuzzy search. This configuration helps to ignore typos, plural forms, case forms and minimize the failed search results.
  • Stopwords filtering. Stopwords are the common words that do not bring any valuable context in most search cases and only expand the search results. At first, the stopwords list can include articles, prepositions, auxiliary verbs. Then it can be filled with the words you would define as non-important as you analyze the queries further.

Level 2 – Search refinement

When the basic search configurations are made, you can proceed with improving how users browse the search results and refine their searches. People hate any extra work, and your customers are no exception. It is always more pleasurable to click on refinement options than browse through multiple paginated search results.

  • Faceted search helps users refine the search results once they have a listing in response to their query. This could be done based on numerous additional parameters like size, color, brand, price, and more. With facets configured, search results change between the searches depending on the selected parameters.
  • Filters do not work this flexibly and help only to eliminate the search results based on initial criteria. For example, users can select either browsing shoes for adults or children, but they can not select both and add more parameters to get the search results changed.
  • Filter suggestion feature enabled allows users to search directly within a selected category. This also helps searchers avoid information overload and find what they are looking for faster.
  • Sorting helps to order search records based on specific attributes like price (from lowest to highest or vice versa), popularity, and other parameters.

If it is relevant to your business, at this step you can also configure a location-based search and provide results relevant to your customers from different locations.

Level 3 – Search ranking

We approached the level that would require some more implementation time, and, thus, resources. We included fine-tuning textual relevance, user-based and attribute-based ranking in this step. All of them are related to the search ranking and aim to show at the top those results that address your business goals.

At this point, the search engineering team should closely collaborate with the product or business development teams. The sample questions they would define at this step are the following:

  1. What results should be shown at the top?
  2. Does merchants' rating influence the relevance?
  3. What role does the price play? Do we show the least expensive goods first?
  4. What fields are included in the searching process and what should be added or removed?
  5. What fields in a product are the most important?
  6. How to evaluate the relevance of each product for different customer personas?

Working with your development team at this level, you can set up rules that drive search relevance based on specific parameters.

  • Attribute-based ranking helps to match shopper queries with not only product names, but also brand names, product types, descriptions, categories, colors, or several of these at the same time.
  • User-based ranking will take into account demographics, location, user preferences. All the criteria known about the particular user.
  • Textual relevance tuning is especially important for the cases when the complexity of catalogs is combined with user-generated product listings.

Level 4 – Relevance tuning

Custom ranking surely matters in search optimization and helps address some business objectives, but it is a search relevance model in its broader sense that we want to achieve.

We recommend starting this step with an analysis of zero-result queries that are usually the best source of insights. Why does your system fail to provide what users are searching for in particular cases? At this point, we help our clients compare successful and failed sessions, look more precisely at searches that return zero results, and find commonalities between them. Ideally, we would divide zero-searches into classes and find solutions to turn them into successful ones.

During the previous step, we defined the ranking rules that can be now neglected for some specific use cases. By defining query-dependent ranking factors, you can achieve dynamic ranking. For example, you have the rule to show more expensive products on top of the search results, but if a person types the word free in the search bar this rule should be neglected. At this step, with the right tech team, you can fine-tune your search engine to provide the most relevant search results specific to your business use cases.

Furthermore, to make the search system understand which results are relevant for broad or ambiguous queries, you would need to configure query expansion. This is a technique when the original user’s query is expanded by adding additional words that help to narrow the query semantics. Query expansion helps determine the searcher’s intent and route users towards a more refined search query.

Level 5 – Advanced search

We have to admit, this level is the most complex and resource-consuming, yet once you reach it, there are, really, no limits. At this step, you will have the capabilities to rethink how your search performs on the highest level and optimize the enterprise search engine with the most advanced technology and methods. Instead of focusing on the efficient ranking algorithm, you will have the ability to create an efficient query understanding mechanism, an advanced system that will learn along the way and focus on query performance as a key success metric.

  • Query refinement analysis helps to look at the process of refining (narrowing down) the initial user searching item and analyze user behavior after the results were provided. This data will help to provide narrowed down paths to results that your users await.
  • Query understanding is the process of establishing the searcher's intent before providing any results. Your search engine will be able to automatically determine when a search query is broad or ambiguous and expand it to provide relevant results.
  • AI-based search personalization allows you to serve users personalized search results based on their previous searches, activities, and purchases. The tech team will need to track and collect users' signals, interpret them, and build relevant user profiles. The search results will be modified based on these profiles but the system won't neglect the textual relevance you've built during the previous steps.

Configurations running in parallel

As you move along the steps of E-commerce site search optimization, two aspects should be considered no matter where you are in this roadmap. They are internal search monitoring and performance tuning. You can also learn more E-commerce site search best practices from our recent blog post.

Analytics for internal search

Some queries lead to great performance. Some queries deliver zero results. Without measuring this data, you can not move to optimization at any step. Search analytics provides you with important information on what your customers want, how they describe the products or services they are looking for, and gives you hints on how you can optimize your offers and overall search experience.

For example, you can track:

  • Queries users enter
  • Most popular results
  • Most-bought products after search
  • Most-used filters
  • Average position of search result click
  • How deep do users go in pagination (do they find good results on the first page?)

Though analytics is a topic that never gets underestimated and is often included in the project scope, setting it up properly is a daunting task for the tech team. Engineers will need to build a data tracking pipeline, normalize queries (taking into account synonyms, typos, word forms), make the analytics privacy compliant, and visualize the results for stakeholders. Analytics also opens a possibility to run A/B tests and evaluate introduced optimizations during and after each level we’ve described above.

Performance tuning

Because search experience is often the primary path to purchase, making the search for enterprise perform faster even for milliseconds can lead to increased revenue. From day one as you introduced search to your E-commerce store, monitor search for downtimes, speed, and vulnerabilities for the sensitive data you might collect.

When our team deals with on-site performance tuning, we start with tech resources analysis that includes analysis of the computing, memory storage, and network you use to run the search on your platform. After this analysis, we proceed with cluster configuration optimization and index configuration to help the search engine perform faster. All the additional optimizations depend on the business goals set for the internal search.

Reach us for E-commerce search consulting

Enterprise search is not something you set and forget. There is always room for investment, and now you have a clear vision of where you might start. You can hire the Wise Engineering team if you're looking for site search experts that can ensure quality, best performance, and scalability for your on-site search.

Learn more about our Elastic site search expertise and contact us to book a consultation to discuss your specific business context.

Table of contents