Your ecommerce search is bleeding revenue and you probably can’t see it clearly. A shopper types “black running shoes under $100” and gets results sorted by margin instead of intent. Someone searches “couch” and gets throw pillows. A zero-results page shows up for a query that has six matching SKUs because the product data is inconsistent. These aren’t hypothetical scenarios. They happen every day at scale, and most site search tools are too dumb to fix them. That’s the environment this Constructor review is operating in, and Constructor’s pitch is that it’s built something meaningfully different.
Constructor is an AI-native product discovery platform for ecommerce. It handles search, browse, recommendations, retail media, and increasingly, agentic shopping experiences. The core idea is that every shopper touchpoint, what you see when you search, what shows up when you browse a category, what gets recommended below a product, should be driven by a single connected intelligence layer that learns from real shopper behavior. Not static rules. Not keyword matching. Actual behavioral signals feeding a model that improves continuously.
Sephora, REI, Gap, Under Armour, Bath and Body Works, Petco, Foot Locker. The customer list is real and it’s not padded with logos of companies that signed up for a free trial. These are retailers with millions of SKUs and complex merchandising requirements, which tells you something about where this sits in the market. It’s not a tool for an 800-product Shopify store. It never was.
Features
The core search and autosuggest engine is where Constructor earns most of its reputation. Unlike keyword-based search that matches query terms to product attributes, Constructor’s engine interprets intent. It looks at what shoppers do after a search, what they click, add to cart, and buy, and uses that signal to rerank results in real time. The difference in practice is noticeable: queries that would return irrelevant or zero results on a conventional search engine get handled gracefully because the model has learned what shoppers actually mean, not just what they typed.
Browse personalization extends the same logic to category pages. When a shopper lands on “Women’s Coats,” Constructor reorders the product grid based on that individual’s behavior history, contextual signals like weather and location, and catalog attributes. Two shoppers see the same category page but with different products surfaced. At scale, this moves conversion numbers. Petco reported a 13% lift in ecommerce conversions after implementation. That’s the kind of number that gets a VP of Commerce a budget for the next initiative.
The newer agentic features are worth noting, even if they’re still maturing. The AI Shopping Agent lets shoppers interact conversationally with the catalog in natural language. The Product Insights Agent generates and answers product questions in context. The Merchant Intelligence Agent helps merchandisers understand why certain results are ranking and where to intervene. These are genuinely forward-looking capabilities, but some are still in beta and shouldn’t be the primary reason you sign a contract today.
Attribute Enrichment uses GenAI to clean and expand product data automatically. This solves a real pain point: catalogues with inconsistent naming, missing attributes, or poor tagging degrade search quality regardless of how good the ranking model is. Constructor addresses this upstream, which is the right place to fix it. But it also adds implementation complexity, because you’re now connecting a data enrichment pipeline in addition to the search layer itself.
How to Use
You don’t just sign up. There’s no free trial, no self-serve onboarding, and definitely no credit card form on a pricing page. You request a demo, talk to sales, go through a scoping process, and then begin an implementation that Constructor says takes eight weeks or less on average. For their client profile, that timeline is acceptable. For anyone expecting to test the product before committing a meaningful budget, it’s a real barrier.
The implementation is API-first. Constructor connects to your product catalog, ingests behavioral data from your site, and replaces or augments your existing search layer. Your tech team will be involved. This isn’t a plug-in. It’s an infrastructure change, and it requires engineering resources on your end regardless of how smooth the Constructor onboarding team makes the process. The “eight weeks” promise assumes your data is reasonably clean and your team is responsive. Both of those are optimistic assumptions for many retailers.
Once live, the merchandiser dashboard is the main interface for non-technical teams. You can set boosting rules, run A/B tests, view analytics on search performance, and monitor zero-result queries. The interface is described as glassbox, meaning the AI’s decisions are explainable rather than black-box. That transparency is genuinely useful for merchandising teams who need to justify decisions to stakeholders or override the algorithm for seasonal or promotional reasons. It’s one of the things clients consistently mention as a differentiator from legacy search vendors.
Pros and Cons
Pros
- Behavioral AI that continuously learns from shopper actions, not static keyword rules — this is the real differentiator and it compounds over time as more data flows through
- Glassbox AI model means merchandisers can see why results are ranked the way they are, which is rare in this category and practically necessary for enterprise merchandising teams
- Named a Leader in the Forrester Wave, Gartner Magic Quadrant, and IDC MarketScape for commerce search in 2025 — analyst recognition across all three major firms is unusual
- Connected platform across search, browse, recommendations, and retail media means behavioral signals from one channel improve all the others
- 98.5% client retention rate over three years, which is a metric that’s hard to fake and suggests real, sustained value delivery
- Attribute Enrichment solves data quality issues upstream, which most search vendors leave as your problem
Cons
- No pricing anywhere on the site. None. You go through a sales cycle before you know what this costs. For a market comparison, that’s a deliberate opacity that favors the vendor
- Eight-week implementation is the optimistic estimate. Complex catalog structures, messy product data, or slow internal approval processes can push this considerably longer
- This is not a tool you can evaluate on a pilot budget. The contract sizes required to work with Constructor put it firmly in the mid-market and enterprise bracket, which excludes most ecommerce businesses by revenue threshold
- Agentic features (AI Shopping Agent, Merchant Intelligence Agent) are still in beta or early rollout. Interesting direction, but not production-ready across the board
- Heavy dependency on clean, well-structured product data. If your catalog is messy, the enrichment layer helps but doesn’t fully compensate for fundamental data quality problems
- Switching costs are real. Once you’ve integrated Constructor deeply into your search and browse infrastructure and trained the model on your data, migrating away is a significant project
Pricing
There is no public pricing. Not even a starting range. The website has a “Book a Demo” button everywhere you look and no numbers anywhere. This is a deliberate choice, not an oversight. Enterprise software vendors do this to maintain negotiating flexibility and to filter out prospects who aren’t ready for the budget conversation.
Based on market positioning, client profile, and comparable enterprise ecommerce search vendors like Algolia, Bloomreach, and Searchspring, Constructor contracts realistically start in the low six-figures annually and scale with catalog size, query volume, and which modules you license. The retailers on their client list are not paying $500 a month. If you need a number to work with for budgeting purposes, think $50,000 to $200,000+ per year depending on scale, though your actual quote will depend entirely on scope.
Is it worth it? For a retailer doing meaningful ecommerce volume with a large catalog and genuine conversion rate problems traceable to search quality, yes, the ROI case is usually buildable. Petco citing payback within a year of implementation is not a made-up testimonial. But the calculation only works at a certain revenue threshold. Below that, cheaper and simpler alternatives cover most of the need without the implementation overhead or contract size.
There’s no free tier, no trial, and no self-serve option. The evaluation process itself requires time and internal resource allocation before you see a dollar figure. That’s the honest reality of buying enterprise infrastructure, and Constructor fits that mold completely.
Who’s it for
Mid-market and enterprise retailers with large, complex catalogs are the right buyer. If you’re running a fashion brand with 50,000 SKUs, a grocery chain with perishability constraints affecting availability, or a B2B distributor with highly technical product attributes that affect purchase decisions, Constructor solves problems that simpler tools can’t. The behavioral learning layer becomes more valuable the more data you have, which naturally favors higher-volume operations.
Ecommerce technology and platform teams evaluating a Lucidworks, Bloomreach, or legacy Elasticsearch replacement will find Constructor a credible alternative. The API-first architecture, the analyst recognition, and the glassbox merchandising interface address the three most common complaints about legacy search vendors. If you’ve been fighting with a rules-heavy search system that requires manual curation for every seasonal push, the behavioral automation argument is genuinely compelling.
Small to mid-size ecommerce stores should skip this entirely for now. If your catalog is under a few thousand products and your annual ecommerce revenue is below roughly $10 to 20 million, the contract size relative to impact doesn’t make sense. Algolia, Searchspring, or even a well-configured Shopify Search with good product data will serve you adequately at a fraction of the cost and integration complexity. Constructor is a serious tool for serious scale, and pretending otherwise doesn’t help anyone make a good decision.