A visitor types "running trainers" in your search bar. Your catalog contains 15 pairs of running shoes. They get zero results. They close the tab.
This scenario plays out on almost every e-commerce catalog, at a scale larger than most teams realize. A significant share of zero-result pages does not signal a missing product - it signals a vocabulary gap between how your suppliers name items and how your customers look for them. The two rarely align.
The combination of these three figures explains the business case. Your best converters use search. A significant share of their queries fail due to vocabulary mismatches. Every failure costs a potential sale.
Why your catalog speaks a different language than your customers
The root cause is structural. Product titles in your catalog come from your suppliers: they reflect trade nomenclature, manufacturer references, and B2B jargon. Your customers search using the vocabulary they use in everyday life - often more informal, sometimes less precise, sometimes drawn from a different generation.
This gap is more pronounced in certain sectors. Musical instruments are a clear example: a merchant may have indexed "solid body electric guitar with humbucker pickups" while the customer types "rock guitar" or "electric guitar for beginners". DIY, fashion, and home appliances show the same patterns.
The four synonym families to configure first
Before getting into the technical mechanics, it helps to map the most common types of mismatches. Each family calls for a slightly different treatment.
Abbreviations and shortened names
Customers systematically shorten product and brand names. These abbreviations are stable over time and predictable - it is the first priority to address.
Regional and generational vocabulary differences
The same product can have multiple names depending on the buyer's generation, country, or regional dialect. These gaps are durable and specific to each catalog.
Spelling variants
Some products have multiple accepted spellings that coexist in everyday language. The search engine must treat them as equivalent.
Technical jargon vs common names
Your catalog may use precise technical terminology from manufacturers while your customers use the common name for the same product. The gap is often invisible from the inside.
How to identify your priority synonyms
The right way to build a synonym dictionary is not to brainstorm in a meeting room what customers might search for - it is to look at what they are actually searching and failing to find.
Two complementary sources:
Zero-result queries in the Vectail dashboard - directly accessible in the analytics tab. They are ranked by frequency, letting you address the highest-impact cases first. A query that appears 40 times a month and returns zero results is a synonym group to create immediately.
GA4 with site search tracking enabled - the search event captures all queries. Cross-referencing high-volume terms with those that generate no product clicks surfaces the words your customers use but your catalog does not recognize.
Manage your synonyms from the Vectail dashboard
Create synonym groups in a few clicks. They sync automatically to Vertex AI and are active immediately on your search engine.
Start free - 14 days, no credit cardThe layer most search engines don't have: automatic ML expansion
Manual synonym management has an obvious ceiling: it only covers what you thought of. Your catalog contains thousands of products, and your customers search in dozens of ways you have not anticipated.
This is where a second layer comes in, specific to Vertex AI Search for Retail technology: automatic query expansion (queryExpansionSpec: AUTO).
In practice, Vertex AI does not limit itself to applying your synonym rules. It analyzes the semantic structure of your catalog - the relationships between terms present in your titles, descriptions, and attributes - and automatically broadens queries when it detects a likely match. This happens outside your synonym dictionary, in real time, for every query.
Synonyms configured in the dashboard
You explicitly define the equivalences your search engine should apply. Full control over the cases you know well: brand abbreviations, sector-specific terms, stable regional variations. Active immediately after configuration.
ML expansion by Vertex AI
Google analyzes your catalog semantics and expands queries beyond your manual rules. Covers cases you have not anticipated. Improves progressively as search volume on your site increases.
These two layers are complementary, not competing. The manual layer provides precision: you explicitly control the critical equivalences for your sector. The automatic layer provides coverage: it handles the thousands of variants that no one could list manually.
One important technical note: automatic expansion only applies when it is enabled in the search configuration. In Vectail, this is the default. The dashboard synonym rules also require this activation to work - without it, Vertex AI silently ignores them.
What synonyms do not do
To use this feature effectively, it helps to understand its actual limits.
Synonyms apply to text fields in your product records - primarily titles and descriptions. If a product is classified in a "Sports Shoes" category but the title does not contain that term, a synonym "sports shoes = trainers" will only help if the title actually contains one of the two terms. Categories and attributes are not expanded the same way.
Synonyms in Vectail are bidirectional - if "trainers" is a synonym of "sneakers", a search for either term returns results for both, in both directions. One-way synonyms (A to B but not B to A) are not available. In most e-commerce cases, bidirectional is what you want - but it matters to be aware of this to avoid unexpected behavior with very generic terms.
Synonyms are not a substitute for spell correction. A typo ("sneakrs") is a different problem from a synonym ("sneakers" = "trainers"). Both features coexist but address distinct cases.
Configuring synonyms in Vectail
The Vectail dashboard exposes a synonym manager accessible from the main menu. Each synonym group contains a list of equivalent terms - adding at least two creates the relationship.
After creation, the group is automatically synchronized with Vertex AI Retail as a TwowaySynonymsAction Control. Synchronization takes a few seconds. If it fails (network issue, temporary error), a "Not synced" badge appears on the affected group - a button lets you force re-synchronization without recreating the group.
There is no limit on the number of groups. In practice, starting with the 20 most frequent zero-result queries, treating them one by one, and measuring the impact on the zero-result rate in the following days is the most efficient method.