When you upload a contract to Smallprint, you get a Trust Score between 0 and 100 in under 30 seconds. But what happens in those 30 seconds? How does a system evaluate something as nuanced as a legal contract, something that took lawyers years of training to analyze?
This post pulls back the curtain on the scoring engine behind Smallprint: the five weighted dimensions, the jurisdiction detection layer, the document-type specialization, and the severity classification system that powers every analysis.
Not a simple keyword search
Most contract review tools work by matching keywords: they look for words like "penalty", "termination", or "liability" and flag them. That approach misses context entirely. A termination clause that gives you 30 days notice is perfectly reasonable. One that locks you in for 3 years with a 6-month cancellation window and automatic renewal. That's a trap.
Smallprint doesn't search for words. It reads the entire document, understands the relationships between clauses, and evaluates each provision against the applicable legal framework.
The five dimensions of the Trust Score
Every contract is scored across five weighted dimensions. Each dimension captures a different aspect of how fair, transparent, and balanced the agreement is.
Transparency (20 points) measures how clearly the contract communicates its terms. Are obligations stated in plain language? Are costs, limits, and deadlines explicitly listed? Or are they buried in cross-references and legalese designed to obscure the real impact?
Balance (25 points) is the heaviest dimension, and intentionally so. It evaluates whether both parties carry proportional obligations. Can the provider change terms unilaterally while you're locked in? Can they terminate without notice while you face penalties for early exit? An imbalanced contract isn't just unfair; in many jurisdictions, it contains clauses that are legally void.
Legal Compliance (20 points) checks the contract against the applicable regulatory framework. In France, that means the Code de la consommation, the Code des assurances, and relevant EU directives. In the UK, the Consumer Rights Act 2015. In Germany, the BGB. The system identifies which framework applies based on the contract's language, governing law clause, and the user's location, then flags provisions that may violate specific articles.
Financial Risk (20 points) quantifies the monetary exposure the contract creates. Hidden fees, uncapped liability, aggressive price indexation clauses, steep early termination penalties, deductibles that exceed industry norms. All of these contribute to a higher financial risk score.
Exit Freedom (15 points) evaluates how easy it is to leave. Contracts with reasonable cancellation periods, clear notice requirements, and no lock-in traps score well. Those with automatic renewal, rolling commitments, and barriers to switching score poorly.
The weighted sum produces a raw score. The final Trust Score maps to four categories: Safe (80–100), Fair (60–79), Shady (40–59), and Toxic (0–39).
Jurisdiction-aware analysis
A score is meaningless without legal context. A 12-month non-compete in an employment contract is perfectly standard in the UK but may violate labor law in France if it lacks financial compensation. A 14-day cooling-off period is a legal minimum in the EU but doesn't exist in most US states.
Smallprint's analysis engine operates in three steps when it encounters a new document:
First, it detects the contract's jurisdiction from the governing law clause, the language of the document, and entity names mentioned in the text. Second, it identifies the user's country from their account settings. Third, it applies the relevant local consumer protection framework. If the contract's jurisdiction differs from the user's country, it raises a flag immediately.
This means that a French user uploading an American SaaS agreement will see warnings about regulatory gaps: provisions that would be unenforceable in France, protections they'd normally have under EU law but that the contract doesn't provide.
Document-type specialization
A lease agreement and an auto insurance policy have almost nothing in common, except that both can contain traps. Scoring them with the same criteria would produce meaningless results.
Smallprint identifies the document type from its content (not the filename) and activates specialized analysis modules:
For insurance contracts, the engine focuses on deductible stacking rules, waiting periods, exclusions that contradict the coverage name, claims filing deadlines, premium increase triggers, and subrogation clauses.
For lease agreements, it examines rent indexation formulas against legal caps, security deposit amounts versus legal maximums, maintenance obligation allocation, and end-of-lease condition requirements.
For employment contracts, it scrutinizes non-compete scope and compensation, intellectual property assignment breadth, probation period length versus legal limits, and garden leave provisions.
For terms of service and subscriptions, it flags data collection scope, content ownership transfers, unilateral modification rights, automatic renewal mechanics, and dispute resolution clauses that waive class action rights.
Each document type has its own set of fixed analysis criteria. On top of these, the system can detect and flag additional issues specific to the individual document: provisions that don't fit any standard category but still deserve attention.
Severity classification
Not all issues are equal. A missing signature line is not the same as a clause that waives your right to legal recourse. Smallprint classifies every detected issue into three severity levels:
Critical issues represent provisions that may be legally void, that create significant financial exposure, or that strip fundamental consumer rights. These are scored heavily and displayed prominently.
Suspect issues are provisions that are technically legal but create notable imbalance or exceed industry norms. A 60-day cancellation notice isn't illegal, but it's aggressive compared to the standard 30 days.
Minor issues are informational: provisions worth knowing about but unlikely to cause harm. They provide context rather than alarm.
Structured data extraction
Beyond the score and the flagged clauses, Smallprint extracts machine-readable structured data from every contract: duration, auto-renewal terms, cancellation period, fees, liability caps, claims deadlines, price change clauses, and more.
This structured extraction serves two purposes. First, it gives you a clear, at-a-glance summary of the contract's key terms without reading 40 pages. Second, it powers Smallprint's comparison engine: when you upload two competing insurance quotes, the structured data is aligned side by side so you can see exactly where they differ.
The company database
Every analysis feeds into Smallprint's growing database of companies and their contract practices. When multiple users scan contracts from the same provider, the system aggregates the scores to build a longitudinal picture: is this company getting better or worse over time? How do their terms compare to competitors in the same sector?
This data flywheel means Smallprint gets smarter with every scan. Not because the engine retrains, but because the comparative context deepens. A score of 62 means more when you can see that the industry average is 71 and this company scored 68 last year.
What we don't do
Smallprint is not a law firm. We don't provide legal advice, and we're explicit about that. The Trust Score is an analytical tool, not a legal opinion. It's designed to surface issues that deserve attention, issues you might then want to discuss with a qualified legal professional.
We also don't guarantee that every problematic clause will be caught. Legal language is inherently ambiguous, and no automated system achieves 100% recall. What we do guarantee is that you'll know more after scanning than before, and that the analysis is grounded in real regulatory frameworks, not generic heuristics.
Try it yourself
Upload a contract at small-print.ai and see your Trust Score in 30 seconds. Your first analysis is free.
Smallprint is built by Kunley SASU. We analyze contracts so you don't have to read between the lines.