How HC Search Works: Our Recommendation Algorithm

When you search for "eor" on Human Cloud, you expect Deel, Foxhire, and other EOR providers to show up as top matches. When you search "talent marketplace," you expect Toptal, MarketerHire, and Wripple. Sounds obvious — but most marketplace search engines get this wrong.

Traditional systems blend relevance with quality signals like reviews, badges, and customer counts into one score. The result? A perfect-match EOR company with zero reviews shows 45% match, while a well-reviewed staffing firm that barely does EOR shows 72%. That's backwards. We fixed it.

Three Independent Axes

HC Search separates three signals that most platforms mash together:

Signal	What It Answers	What Drives It
Match %	"How relevant is this to your search?"	Capability match, category match, semantic similarity
HC Score	"How strong is this solution objectively?"	Verified customers, reviews, kudos, badges, global rank
My Score	"How well does this fit my specific rubric?"	Your saved preference weights

None of these influence each other. A solution can be 99% match with a low HC Score (perfect fit, new to the platform). Or 60% match with a high HC Score (great company, not what you searched for). Both signals are honest. You decide which matters more by choosing how to sort.

Tier 1: Core Relevance

When you type a query, we first ask the simplest question: does this solution actually do the thing you're looking for?

If a solution has the exact capability you searched for — like "eor" as a tagged, verified capability — it gets 99% match. No hedging, no blending with other signals. It IS an EOR. That's what you asked for. 99%.

If it matches the category (e.g., "Talent Marketplace") but not a specific capability, it gets 95%. If neither matches, we fall back to semantic similarity — our AI embedding model compares the meaning of your query against everything we know about the solution: name, description, categories, features, capabilities, and industries. This produces a 40–94% match based on conceptual relevance.

The principle: structured data always beats embeddings. If we know a solution has EOR as a capability, we don't need AI to tell us it matches "eor." AI is the fallback for fuzzy, novel, or complex queries.

Anti-Gaming: Why You Can't Tag-Stuff Your Way to the Top

Here's the problem with most marketplace search: solutions are incentivized to add as many capability tags as possible. If "eor" is a tag and it boosts your ranking for EOR searches, why not add it — even if EOR is your 12th priority?

We weight by priority, not presence.

Every solution's capabilities are priority-ordered. The score you get for a match depends on where that capability sits in your priority list:

Capability Position	Match Score
Top 3 (core capabilities)	99%
Position 4–6	93%
Position 7+	88%

What this means in practice:

A focused EOR company with 3 capabilities — EOR, Payroll, Compliance — scores 99% for an "eor" search. Every one of those terms is in their top 3.

A company that tagged 15 capabilities and listed EOR as their 10th? They score 88% for the same search. They have EOR, sure — but it's clearly not their focus.

We reward focus, not breadth-for-the-sake-of-breadth. If EOR is truly what you do, make it your top capability. If it's a secondary offering, that's fine — you'll still appear in results, just below the specialists.

Tier 2: Contextual Refinement

Once you have your initial results, you can refine. We offer six refinement dimensions:

Level 2:

• Solution Type — Source (finding talent), Comply (evaluating risk), or Manage (onboarding & payment)
• Focus — What the solution specializes in beyond its core capability
• Industry — Healthcare, Technology, Finance, etc.
• Region — Where they source talent and where they serve customers

Level 3 (Advanced):

• Compliance — IC verification, indemnification, insurance coverage
• Governance — VMS integration, HRIS, SSO, payroll systems

Each refinement checks how well a solution fits your criteria. The formula:

adjusted match = tier1 score x (0.7 + 0.3 x tier2 overlap)

If you search "eor" and then refine by Industry: Healthcare:

• Foxhire (EOR + serves healthcare) stays at 99%
• Deel (EOR but not healthcare-focused) drops to ~69%

Tier 2 can reduce your match score but never inflate it. A solution that isn't relevant doesn't become relevant by having good compliance. Core relevance is always dominant.

Solution Type: Source, Comply, Manage

Every workforce solution on Human Cloud maps to one or more of three buckets:

Bucket	What It Means	Examples
Source	Finding and sourcing talent	Talent platforms, marketplaces, staffing, recruiting
Comply	Evaluating and managing worker risk	EOR, AOR, IC classification, worker verification
Manage	Onboarding, payment, owning the risk	Payroll, benefits, FMS, VMS, managed services

This bucketing is inferred automatically from each solution's capabilities and categories — no manual tagging required. A solution with both "talent marketplace" and "payroll" capabilities shows up in Source AND Manage.

This is what makes queries like "I want an FMS that isn't an EOR" actually work. You select Solution Type: Manage, and pure Comply solutions naturally deprioritize.

Sorting: You Decide What Matters

After searching, you have three sort options:

• Relevance (default) — Match % descending, HC Score as tiebreaker. Among four 99% EOR matches, the one with the highest HC Score appears first.
• HC Score — Objective quality descending. Useful once you've narrowed to 5+ results that all match well — now show the most credible ones first.
• My Score — Your personalized rubric. If you've configured what matters to you (verified reviews weighted heavily, enterprise customers preferred), this sorts by your priorities.

Sorting is independent from scoring. Switching to "HC Score" doesn't change anyone's match %. It just reorders who you see first.

Why This Matters

Most marketplace search engines have a fundamental conflict: they want relevant results, but they also want to reward their best customers (the solutions with the most reviews, the biggest logos, the most badges). So they blend everything together into one opaque score.

The result is predictable: new solutions that perfectly fit your search get buried. Established solutions that vaguely match float to the top. You lose trust in the search, and the marketplace loses its utility.

We chose transparency over opacity:

• Match % tells you relevance. 99% means this solution does exactly what you searched for. Period.
• HC Score tells you credibility. A high score means verified customers, real reviews, industry recognition.
• My Score tells you personal fit. Your rubric, your weights, your priorities.

Three honest signals. You decide how to weigh them. That's how search should work.