Our flagship practice: we test your live search box against real multi-constraint shopper requests, quantify the sales you miss, and engineer the hybrid retrieval and catalog fix.
Field Evidence · Anonymized Store Audits
On October 31, 2026 we tested two US furniture retailers against their own in-stock catalogs. Furniture retailer A (Fast Simon) failed 10 of 22 requests (9 returning zero results). Furniture retailer B (Algolia) failed 7 of 24, with 9 more only partly right.
What We Test & Engineer
A standard Search Audit tests 30 frozen shopper intents (60–80 in Extended) across size/fit, price ceilings, materials, use-cases, seating/capacity, negation, and multi-constraint combinations. Read Method A in the Test Standard.
| Constraint | Example intent | A correct answer must respect |
|---|---|---|
| Size or fit | tv stand for 75 inch tv | Stand width at or above the TV’s width |
| Price | oak desk under $500 | Price at or below $500 on the test date |
| Material | solid wood dining table | Material attribute or stated construction |
| Use or audience | couch good for kids and dogs | Performance or pet-friendly fabric |
| Capacity | dining table that seats 8 | Stated seating, or a length that fits 8 |
| Negation | sofa that isn’t leather | Excludes leather while keeping matching sofas |
| Combination | grey sofa under 80 inches for a dog | All constraints satisfied simultaneously |
From Audit to Production Fix
Every result is verified against your catalog truth table, and 100% of your Advisory Audit fee is credited toward your Tier 2 build within 60 days.
In your search box on desktop and mobile, logged out, in a clean session, from a US IP address. We record results, autocomplete, and any “did you mean”.
A timestamped full-page screenshot, the result list as text (product, price, position) and the URL, named by intent and run time.
A reviewer marks each of the top 10 results against your catalog truth table. A second reviewer checks every failure and a 20% sample of passes.
Good: fitting product in top 3. Partly right: fitting items below #3 or one constraint ignored. Failed: zero results or no match in top 10 while stocked.
We deliver a vendor-neutral engineering spec—or implement the fix directly via in-stack tuning or our Talos hybrid semantic search layer.
Audit Deliverables & Live Reference Engine
Every failure includes timestamped evidence, the exact in-stock SKUs that should have matched, and a side-by-side comparison against our Talos Semantic Engine (live at usafurnish.com). See the sample audit report.
Tier 1 & Tier 2 Search Pricing
Want AI shopping assistants tested on the same intent set? Combined Search & AI Findability Audit covers both for $5,000 / $10,000. $2,500 design-partner rate available for the first 3 Tier 1 Search Audits in exchange for named before/after case study rights.
Flagship
Tier 1 Diagnostic · 100% credited to Tier 2 within 60 days
Tier 2 Fixed-Scope Search Builds
In Advisory mode, 100% of your Tier 1 Audit fee is credited toward any Tier 2 engineering build signed within 60 days of the readout.
Full 4-tier price list · Event-triggered re-test $1,500
An estimate for on-site search only, with every assumption shown. 35% is our cautious default failure rate on natural-language searches; a Search Audit measures yours.
Estimated revenue at risk a year
$163Ktoday (likely)
Recoverable if we close a quarter of today’s gap: $41K a year, about $3,400 a month.
How it’s calculated. Online revenue × 44% from shoppers who search × the share of searches written in natural language (20% today) × the failure rate on those searches (35%, a cautious default, below the 45% our furniture study measured) × 53% of shoppers who leave after a failed search. Growing and AI-era use the same failure rate. Recoverable is 25% of the Today case, our planning assumption until a fix is measured.
Sources: 44%, Constructor, 609M searches on 113 sites, Q4 2024 (vendor data); 53%, Google Cloud and Harris Poll, 2022 (stated behaviour). The 45% failure rate is from our furniture study of 22 natural-language requests and applies only to natural-language searches. The natural-language shares, the 35% default and the 25% recovery are our assumptions; the loss grows as more shoppers search in natural language and through AI assistants. These are estimates, not forecasts.
How is the Search Findability Audit different from the free 60-second Findability Score? The free Findability Score runs 8 automated requests and gives an immediate 0–100 diagnostic across 4 engineering sub-scores. The Search Findability Audit ($3,500 Standard / $7,500 Extended) runs 30 to 80 real shopper intents on desktop and mobile, checks the top 10 results of each by hand against your catalog truth table, analyzes your search logs, and delivers a root-cause engineering spec—with 100% of the fee credited toward a Tier 2 build within 60 days.
Do we have to replace our current search engine? No. In Path A ($6,000–$12,000), we tune and re-index your existing engine (Algolia, Klevu, Constructor, Fast Simon, or Shopify Search) and enrich your catalog attributes ($6,000–$18,000). In Path B ($12,000–$35,000 build (+ $1,500–$3,500/mo hosted)), we deploy our turnkey Talos Semantic Engine hybrid retrieval and reranking layer on top of your catalog.
Is there a design-partner discount for the first Search Audits? Yes. $2,500 design-partner rate available for the first 3 Tier 1 Search Audits in exchange for named before/after case study rights.
Can we see your semantic search engine running live? Yes. Our Talos Semantic Engine is live across 10,000+ furniture SKUs at usafurnish.com, demonstrating sub-120ms hybrid BM25 + vector retrieval, numeric/size constraint parsing, and cross-encoder reranking.
Next Step · Engineering Architecture & Diagnostics
A 30-minute engineering call: your store, your search & recsys stack, and whether a Tier 1 Audit, Tier 2 Build, or Resident ML Engineer fits best. No slides, no obligation.