Notes on Comparing Japan's Compliance Check SaaS — Information Source Design and Pricing Traps

We already run a counterparty compliance check SaaS in-house, but with the market reshuffling I wanted to revisit the choice from scratch, so I compared the six major Japanese products as of 2026-07-21 based on public information: RoboRobo Compliance Check, RISK EYES, SP RISK SEARCH Supported by RoboRobo (SPRS from here), Nikkei Risk & Compliance, RiskAnalyze, and Risk Monster. I also touch on adjacent options that come up in the shortlist: Sansan’s anti-social check option, Alarmbox Power Search, and ComCheck.

Some background for readers outside Japan: “hansha check” (反社チェック) is the standard Japanese due-diligence practice of screening counterparties, officers, and shareholders for ties to organized crime and other anti-social forces. Government guidance and Tokyo Stock Exchange listing criteria effectively require it, and it sits alongside AML/KYC as part of the broader counterparty risk management picture.

Filling in a feature-matrix comparison ends up looking similar no matter who does it, so I’ve deliberately tilted this article toward “what the table hides.” That means four areas: information source design, the real billing unit (per-name, not per-company), audit trail and approval flow, and what overseas coverage actually means.

Market Reshuffling in 2025-2026

The market went through significant restructuring in 2025-2026, and reading older comparison articles will lead you astray if you don’t factor this in.

The biggest change is that RoboRobo, provided by Open Corporation (TSE code 6572), and SP RISK SEARCH, provided by crisis-management specialist SP Network (SPN), signed a business partnership agreement on 2025-09-09, announced the launch of their integrated platform SPRS on 2026-06-01, and began accepting applications from July 2026. Historically the two were positioned as competitors (web-public-information sweep type RoboRobo versus anti-social specialized DB type SP RISK SEARCH), but at least at the information-source layer they are moving toward integration. The official announcement is at SP RISK SEARCH Supported by RoboRobo opens applications from July.

The other big change is that RISK EYES went through a full renewal on 2025-08-08, adding approval workflow and AI noise filtering as standard. It used to be positioned as “search-focused, noise-reduction type,” but it now reaches into internal-control territory, and its character has shifted noticeably.

Across the board, AI summarization, priority scoring, automatic evidence capture, and monitoring have become standard. The overall picture in 2026 is a shift from “search tools” toward “counterparty review platforms.”

How the Six Services Line Up

After going through them, the honest way to characterize the six is not by counting features but by their information-source design philosophy. Individual feature lists have converged; the meaningful differences live in how each vendor builds its data.

RoboRobo Compliance Check (Open Corporation): a sweep-and-efficiency product that lets you search Web, newspaper, overseas, SPN anti-social DB, credit, and corporate registry in one click. But the core web search is essentially Google search plus result aggregation automated, and RoboRobo itself has no in-house anti-social judgment DB. The specialized DBs come from the SPN partnership and LSEG (formerly Refinitiv) World-Check. Over 10,000 customers when all contract types are combined.
RISK EYES (SocialWire): a precision product that pre-filters noise using its proprietary public-information DB and anti-social DB. The August 2025 renewal added AI noise judgment and approval workflow. 779 deployments, 59 post-adoption IPOs.
SP RISK SEARCH Supported by RoboRobo (SPN): a members-only platform built around SPN’s QSS, the largest private anti-social specialized DB in Japan (about 600,000 entries), curated by humans since 2004 from over 100 national and regional newspapers going back to the 1960s. The single most differentiating property is that it preserves real-name records for past incidents that have since been deleted or anonymized on the web and in newspaper archives. Requires SP Club membership.
Nikkei Risk & Compliance (Nikkei Inc.): a reporting-plus-global product combining the Nikkei Telecom archive (50+ publications, 1975 onward, over 10 million adverse news items) with Dow Jones Risk & Compliance global risk data (200+ countries, 60+ languages, over 4 million risk data points, PEPs across 200 countries and 22 categories). TPRM and DDR are also on offer.
RiskAnalyze (KYC Consulting): a curated-DB product that ingests over 1,000 Japanese media sources every hour, with AI filtering and structuring the results into a judgment-ready DB. One search in 0.4s, a 1,000-record CSV in about 1 minute, 7 years of history retention, and free API. A different animal from products that live-crawl arbitrary web pages.
Risk Monster (Risk Monster Corp.): a credit-integrated product that combines the anti-social check heatmap, individual search, e-credit navi, and API, letting you do compliance checks and credit judgment through the same vendor.

Sansan’s anti-social check option, ComCheck, and Alarmbox Power Search sit in what you might call the “adjacent workflow integration” category: Sansan couples with business-card management, ComCheck with Mitsui & Co. group credit management, and Alarmbox with corporate-registry retrieval and credit reports. Choose them for the adjacency more than for the standalone anti-social check performance.

Here is a quick-reference table lining up the six services against this article’s four axes (information source design, pricing, approval flow, overseas coverage). All figures come from public information; actual totals shift with name count, media count, and options, so a vendor quote is required.

Service	Information source philosophy	Public pricing (tax excl., reference)	Approval flow	Overseas coverage
RoboRobo	Web search automation + SPN anti-social DB / World-Check partnerships	250 yen/case usage-based, 20,000 yen/mo (up to 100 cases), newspaper DB separate	Approve / reject / hold with comments (multi-stage details not public)	World-Check partnership, about 190 countries, about 5.4 million records
RISK EYES	Proprietary public-information DB + anti-social DB (precision)	300 yen/search per media source, 15,000 yen/mo minimum	Approval workflow shipped standard in the August 2025 renewal	OFAC, Japan MOF/FSA/METI, Dow Jones Watch List
SPRS	SPN anti-social-specialized DB (about 600,000 human-curated entries since the 1960s)	Contact for quote (SP Club membership required)	Human specialist support; in-product application/approval not publicly documented	World-Check partnership, PEPs over 1.4 million, over 1,400 watchlists
Nikkei Risk & Compliance	Nikkei Telecom reporting DB + Dow Jones	Contact for quote (110,000 yen/mo (tax incl.) up to 100 cases is a reference)	TPRM: apply → screen → assess → EDD → approve → monitor	200+ countries, 60+ languages, 24/365 updates, OFAC 50% Rule handling
RiskAnalyze	AI-curated structured DB (over 1,000 Japanese media sources ingested hourly)	27,500 yen/mo (600/year), 50,000 yen/mo (1,200/year)	Application/approval, full operation log, monitoring as fast as daily	Via Acuris: 240+ countries, 50 sanctions lists, 1.4 million PEPs and relatives
Risk Monster	Proprietary anti-social/compliance DB + credit information (credit-integrated)	30,000 yen enrollment + 20,000 yen/mo, heatmap 1,000 yen/case	In-member history; approval workflow not publicly documented	Not documented on the public core anti-social pages

The overseas column lumps “World-Check partnership” rows and “Dow Jones / Acuris” rows together, but they run on different upstream feeds, so comparing by country count alone is misleading. Compare by specific list names, PEP granularity, language, and update frequency (covered under Point 4).

Point 1: The Information Source Philosophies Are Fundamentally Different

Feature tables all look alike when they say “searches newspaper articles, web, and overseas sanctions lists.” But there are three fundamentally different ways to build the source data.

The first is “automate web search,” which is what RoboRobo does at its core. It runs Google in the background and aggregates results, so it inherits everything that flows from that: takedown requests against the target, SEO noise, and so on. The strengths are immediacy and breadth; the weakness is that comprehensive coverage of past information cannot be guaranteed in principle.

The second is “own a curated judgment-ready DB.” This includes SPN’s QSS (about 600,000 human-curated entries dating to the 1960s), RISK EYES’s proprietary public-information DB and anti-social DB, RiskAnalyze’s AI-structured DB, and Risk Monster’s proprietary anti-social and compliance data. Because entries are already filtered at ingestion, noise is low and deletion/anonymization has less impact. The weakness is that inclusion criteria, update cadence, and correction/deletion policies are all vendor-dependent, and anything outside the ingestion scope simply cannot be surfaced.

The third is “build around a reporting DB,” which is Nikkei’s approach. Anchored on Nikkei Telecom’s archive of over 50 publications going back to 1975, reported facts get an extremely high level of credibility, but the tradeoff is that anti-social information that was never reported cannot in principle be surfaced. For auditors and lead underwriters, Nikkei has the highest evidentiary authority.

Deciding upfront which type is primary for your use case, and which type complements it, matters more than the checkmark comparison. National anti-social depth belongs to the curated-DB type (SPRS especially), overseas AML/CFT belongs to the reporting-plus-global type (Nikkei), and speed on web-based reputation goes to the automated-web-search type (RoboRobo).

Point 2: The Real Billing Unit Is “Per Name,” Not “Per Company”

This was the biggest gotcha in the pricing comparison. If you just take the per-search unit price and multiply by your expected volume, real-world usage runs 2 to 7 times that figure.

Two reasons.

First, the minimum billing unit is not “company” but “name” (each of company name, representative name, officer name, shareholder name is one search). Screening one counterparty typically means querying the company name and the representative name simultaneously, so a single new engagement burns at least 2 name-searches. Add 3 officers and 2 major shareholders and you’re at 7. “300 yen per search” translating to 2,100 yen to screen a single company is a normal outcome.

Second, public web search and newspaper article DBs each have a two-tier “search fee + article view fee” structure. On RoboRobo, the newspaper DB usage fee is 6,000-6,600 yen per year, headline view is 5-10 yen each, full-text view is 50-150 yen each, and generative AI summarization is a separate 20,000-22,000 yen per month. On RISK EYES, the base is 300 yen per search per media source, and searching all four sources (newspaper, web, blog, sanctions list) makes it effectively 1,200 yen per search, on top of which newspaper full-text view fees pile up.

When asking a vendor for a quote, the reliable move is to have them work through a specific example in the quote itself: “cost to screen one company with 7 names (company + representative + 3 officers + 2 major shareholders), including full newspaper article views.” A rate card alone does not let you calculate cost per case accurately on your own.

Point 3: How Well Do Audit Trail and Approval Flow Actually Hold Up

If you’re anywhere near IPO preparation or internal-control audit, operational controls matter more than search features. The question isn’t whether an evidence-preservation feature exists; it’s what gets preserved, how far it goes, and at what granularity.

Six things to check.

Are search conditions (keywords, boolean expressions), execution timestamp, executor name, information source, and article URLs automatically saved in a tamper-evident form?
Do downloaded PDF captures and CSVs actually print all of the above items in the format auditors and lead underwriters expect?
Can the searcher and the approver be split into different permission roles, with self-approval prohibited (four-eyes principle)?
Can send-backs, holds, denials, comments, attachments, and operation logs be preserved?
Can scheduled monitoring and annual full re-screening be automated, including for representative changes and company-name changes?
On contract termination, can you bulk-export the saved evidence, PDFs, CSVs, comments, approval history, and operation logs?

Of these, the approval flow (segregation of duties, self-approval prohibition, send-backs and escalation) and post-termination data handling are the areas where fewer vendors have definitive public information, so they must be on your vendor questionnaire. RISK EYES, Nikkei TPRM, and RiskAnalyze document their approval-flow implementations concretely on their official pages, but for RoboRobo, SPRS, and Risk Monster, the specifics of segregation-of-duties support couldn’t be confirmed from public information alone.

Point 4: Don’t Compare Overseas Coverage by Country Count

“Covers about 190 countries” or “over 240 countries and regions” tells you nothing about what you can actually do. For overseas coverage, look at these five instead of the country count.

Specific list names (OFAC SDN, UN sanctions, EU sanctions, Japan MOF/FSA/METI, and other national regulator lists)
PEP (Politically Exposed Persons) category granularity, and whether close associates and relatives are covered
Language support (60+ languages breadth, and effective local-language search capability)
Update frequency (24/365, daily, or slower)
UBO (Ultimate Beneficial Owner) coverage and OFAC 50% Rule (OFAC 50% Rule FAQ) entities

The two vendors that publicly disclose these five in concrete terms are Nikkei (via Dow Jones: 200+ countries, 60+ languages, 24/365 updates, over 54,000 OFAC 50%-Rule entities, over 297,000 state-owned enterprises) and RiskAnalyze (via Acuris: 240+ countries, 50 sanctions lists, 1.4 million PEPs and relatives). RoboRobo and SPRS advertise overseas coverage via World-Check, but the specific lists included in the base fee, UBO handling, language support, and secondary identifier fields largely require follow-up questions.

If overseas trade or foreign officers are core to your business, Nikkei or RiskAnalyze; if you’re primarily domestic with overseas as a supplement, World-Check via RoboRobo or SPRS is usually enough.

Thinking About Small-Team Operation

For us, staffing up a large legal and general-affairs team isn’t realistic, and we’d be doing several hundred cases a year with a small team. At that scale, cost structure is driven not by the per-search price but by how much per-case review time you can cut.

RoboRobo’s design cuts benign-article review time via its 3-tier AI priority scoring and generative AI summarization. RISK EYES groups articles on the same incident automatically, cutting duplicate review. RiskAnalyze filters noise at ingestion and matches against a structured DB. Each has a different “efficiency lever,” and vendor-reported figures (“up to 95% noise reduction,” etc.) aren’t a substitute for measuring against your own real counterparty data.

On fixed cost: an early-stage startup doing 1-10 cases a month can consider Alarmbox Power Search’s Light plan (3,000 yen/month) or RoboRobo’s usage-based plan, both of which avoid a fixed monthly floor. For steady operation above 600 cases per year, RiskAnalyze’s annual bucket makes budgeting predictable. If you can consume roughly 50+ searches per month, RISK EYES’s 15,000 yen/month floor becomes stable and includes approval flow.

What to Measure in a Trial

For items comparison tables and vendor materials can’t decide for you, the fastest path is putting the same test dataset into each vendor and measuring yourself.

Break the test cases into five types.

Known serious negatives (companies or individuals with confirmed reports of regulatory violations, organized crime, etc.): tests sensitivity (detection power)
Clean individuals sharing names with famous people (matches to celebrities or names of past serious offenders): tests noise-narrowing capability
Small/medium businesses with common company names (identical names shared by dozens of companies): tests entity resolution
Newly established clean companies: tests clarity of “no match” and false-positive rate
Foreign officers or overseas counterparties (English-notation overseas entities and foreign personal names): tests matching against sanctions lists and PEPs

Then measure these three, with the same operator running the same sequence with a stopwatch.

Time to complete (min/case): from data entry through hit review, decision status setting, and evidence PDF save
Noise ratio (%): share of hits that are unrelated same-name articles or benign articles with no compliance-risk relevance
Click count and screen transitions: UI path length from search start to evidence acquisition

Treating “high hit count” as accuracy will make you rate noisy tools highly, so the trick is to measure “missed known-related information” and “review time per case” together.

The Tool Isn’t the Finish Line

Reading across the three source reports, the point that hit hardest was the shared conclusion that no tool eliminates the human requirement for final judgment on whether to engage, or for deeper investigation of suspicious hits. Deciding whether a hit is a different same-named person, a minor past administrative sanction, or a serious risk warranting stopping the deal is a legal judgment that lives in internal governance.

The Japanese government’s guidance on preventing damage from anti-social forces and TSE listing criteria require organizational responses, coordination with outside specialist bodies, internal rules and manuals, training, exclusion clauses, and board reporting. What the tool can automate ends at information collection, first-pass screening, and evidence generation. Scope definition (counterparties, officers, major shareholders, agents, subcontractors), standardization of search conditions and judgment criteria, segregation of duties and exception approval, escalation to legal and executive leadership on high-risk hits, exclusion clauses and deal-termination procedures, coordination with police and prefectural anti-organized-crime centers (for example the Tokyo Metropolitan Center for the Elimination of Boryokudan, JA), and event-driven re-screening on personnel or ownership changes all need to be built as separate governance work.

It’s easy for internal discussion to focus on tool selection and drift into “we’ll be done once we adopt this,” but what auditors and lead underwriters actually want is governance, not tool adoption. If you don’t build the rules, operational flow, and final-judgment recordkeeping in parallel with selection, you’ll pay for the rework later. Making the operational status visible in daily standups and team meetings makes the audit conversation much easier.

Wrap-Up

Japan’s compliance-check SaaS market went through significant reshuffling in 2025-2026, and RoboRobo and SP RISK SEARCH partnered into the integrated SPRS platform. The assumptions in older comparison articles no longer apply.
The meaningful differences between the six aren’t feature counts but information-source design: automated-web-search, curated judgment-ready DB, and reporting DB are three different philosophies. Feature-table comparison spins in place until you’ve picked which is primary for your use case.
The real billing unit is per-name, not per-company. One company plus its representative, three officers, and two major shareholders is seven names. Don’t estimate annual cost from the rate card alone.
For audit trail, the question isn’t whether it saves; it’s whether search conditions, timestamp, executor, source, and article URL persist in a tamper-evident form. For approval flow, look at segregation of duties, self-approval prohibition, and escalation.
For overseas, compare specific list names, PEP granularity, languages, update frequency, and UBO/50%-Rule coverage. Not country counts.
Trials should put the same five test-case types (known negative, clean same-name, common company name, new clean, foreign) into every vendor and measure noise ratio and time per case.
What the tool automates ends at first-pass screening and evidence generation. Final judgment, rulemaking, and structural governance remain human work inside your organization.

That’s all from evaluating six counterparty compliance check SaaS products against the four axes of information-source design, billing unit, audit trail, and overseas coverage, from the Gemba.