# Traffic from 67M web pages: how Brazilian company lookup site Sr. Watson climbed the charts · Web

> Source: https://stringers.io/en/web/web_srwatson.io.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> After a redesign this spring, it used public tax-authority data to generate a web page for each of 67M companies and their shareholders. Google indexed them over the summer, and it picked up long-tail traffic from searches for company names and people's names.

[Web](https://stringers.io/en/web/index.md) · 11-6 written

Sr. Watson · srwatson.io

# Traffic from 67M web pages: how Brazilian company lookup site Sr. Watson climbed the charts

**What it is** A free Brazilian company tax ID (CNPJ) lookup site: enter a tax ID and see the company's shareholders, address, contacts and relationship network, backed by a paid investigation platform for the same company.

**Why it blew up** This spring it was redesigned to generate a web page for each of 67M companies and their shareholders from public tax-office data; after Google indexed them in summer, it picked up long-tail traffic from searches for company names and people's names.

**C** **How you could follow it** Don't try this in Brazil; there are already a dozen-plus similar sites. Take the playbook ("full build of public company registry data + interlinked companies and shareholders + paid deep reports") to a country where the data is public but local lookup sites are still weak.

C grade: the playbook is clear and learnable, but this niche in Brazil is already a red ocean, its traffic has started falling since late October, and it carries two risks: Google cleaning up bulk pages and privacy complaints. If you move to another country, there are probably established sites there too. Few people know about it yet

Global rank 109k · 90 days ago, top 1M+

A Brazilian influencer who teaches online investigation turned his free CNPJ lookup tool into a page-generating machine: one page per company, one per shareholder, 67M pages in total. After Google indexed them, its global ranking climbed from outside 90M to around 100k within 1 days. But Brazilian CNPJ lookup sites were already crowded, and it came in late. What's worth learning is the playbook; whether it's worth copying depends on whether you have a country that's still empty.

### A video with only 127 views

In Sep 2026 24, a Brazilian user posted a short video on TikTok with the caption: "Have you heard of the site Sr. Watson? If you've had or currently have a CNPJ (Brazilian company tax ID), go Google your name right now. This site puts millions of entrepreneurs' personal data on display for anyone to see." The second half showed how to request removal under the law.

The video got only 127 views. But it touched on something real: starting in summer 2026, searching a small Brazilian business owner's name on Google increasingly returned pages starting with srwatson.io/pessoa/. Click through and you get a list of every company under that person's name, plus an age bracket.

### What it is

The srwatson.io homepage has just one input box: "Enter the company's CNPJ to start". Below it: "+67M companies · +1M lookups · official Federal Revenue data". Enter one and within seconds you get a profile with shareholders, industry codes, tax regime, address, phone, email, and the relationships between shareholders and between companies (homepage, read 11-07).

It says its target users are lawyers, accountants, compliance staff, lenders, journalists and investigators. Basic lookups are free; a PRO tier is priced at R$90 (about $16, monthly vs. annual billing isn't stated on the page). Every company page has a row of paid entry points underneath: company profile, lawsuits, public tenders, debts, notary protest records, and "AI deep search". All of these lead to the same company's paid investigation platform, Sherlocker; the site's code lists "Sherlocker" as an alias of Sr. Watson.

The person behind it is Bruno Fraga, a Brazilian teacher in open-source intelligence (OSINT) and digital investigation, who has written a book on intrusion techniques and has 389k subscribers on his main YouTube channel. In 2024 he posted an intro video, "We built a company investigation tool", which has 23k views so far.

### Why now: redesigned in spring, indexed in summer

Internet Archive snapshots show the timeline. In Mar 23 2026, the homepage title was just "Company investigation"; by May 19 it had become "Sr. Watson — complete company investigation by CNPJ", with "Become PRO" and "Log in" added.

The traffic then went like this:

| Timeline | What happened |
|---|---|
| 2026-03 to 05 | Site redesigned into a search-engine-friendly structure |
| 2026-06 | Chrome real-user rank in Brazil: top 500k tier |
| 2026-07 | Jumped to top 100k tier |
| 2026-08, 09 | Jumped again to top 50k tier, plateaued in September |
| 2026-09-03 | First time in global top 1M (rank 900k) |
| 2026-10-27 | Global rank peaked at #99,952 |
| 2026-11-05 | At #108,966, then a slight decline for 9 consecutive days |

Users are almost entirely in Brazil; it doesn't show up in Chrome user stats for other countries. The global rank is calculated on a 30-day rolling basis, so peaking in late October and then declining suggests traffic had stopped growing from late September to mid-October (inferred).

The founder's videos and courses are from a year or two ago and don't have much views, so they can't explain the surge from 7 to September. The only thing that can explain it is the redesign itself.

### Where the traffic comes from: a page-generating machine

Open its sitemap: 3,373 sub-sitemaps, each with 20k URLs, roughly 67M /empresa/company-name-CNPJ pages in total, which matches the homepage claim of "67M companies". The data comes from the full company registry the Brazilian Federal Revenue publishes openly; anyone can download it.

Every page is written the way search engines like: the title is "Full company name — CNPJ number | Sr. Watson", with a description and structured data (address, phone, shareholders, 7 common Q&As, pricing). Shareholders who are individuals link to /pessoa/ person pages, which link back to every company that person is tied to; company pages are also chained together by adjacent CNPJ numbers. So once Google's crawler gets in, it can follow links and crawl the whole site.

The effect shows up in search results. Searching a plain company name like "FRW SERVICOS MEDICOS LTDA cnpj" from a Brazilian IP, srwatson.io ranks #5 and #9, alongside old sites like econodata, jusbrasil and serasa. Searching a person's name also surfaces its person pages.

**The person pages are what sets it apart from most CNPJ sites** : they capture long-tail searches for both "company name / tax ID" and "person's name". The cost is also here: every person page has a "request deletion" option to deal with Brazil's personal data protection law (LGPD); the TikTok backlash at the top came from this layer.

Other channels are icing: lawyer-circle Instagram short videos like "Lawyers, screenshot this! I put together 15 tools" mention it; an OSINT tool list ranks it #160; investigative media reports have cited "found on Sr. Watson". These bring mostly trust and backlinks, not volume. I didn't see it running ads on Meta or Google.

### Where the money comes from

The publicly visible tiers are just two: free and PRO R$90. The real money is probably downstream at Sherlocker: pay-per-search or monthly deep investigations, claiming to connect 50+ data sources and able to look up individuals' tax IDs and hidden assets (inferred, no public revenue).

This is where it differs from peers. Same CNPJ profile page: econodata funnels people to buy sales lead lists, Casa dos Dados funnels to bulk company list exports, Sr. Watson funnels to "paid investigations". Its users are lawyers, compliance staff and investigators, with higher ticket sizes, and Bruno Fraga's name carries weight in exactly that crowd. **Acquisition comes from SEO pages, conversion from the founder's reputation, and the two are separate.**

### Who it's competing with

Very crowded. Every Brazilian CNPJ site uses the same Federal Revenue data, so the page content is largely similar.

| Website | Global rank (11-05) | Positioning |
|---|---|---|
| Jusbrasil | 1,276 | Legal portal that also runs company and person pages; its scale is over a hundred times bigger |
| Escavador | 14,832 | Person name + court case search, far deeper on the person dimension |
| cnpj.biz | 19,215 | Category leader, pure company profiles |
| cnpj.ws | 36,564 | July is still at 310k names, surging at the same time with a similar playbook |
| Econodata | 38,361 | Sells sales leads, also has person contact pages |
| Casa dos Dados | 60,971 | Bulk exports of company lists by filter, still growing |
| CNPJá | 66,717 | Leans toward developer APIs |
| Sr. Watson | 108,966 | Corporate investigations, funnels to paid investigations |

Behind it are a dozen or so similar sites ranked between 80k and 300k, many also with shareholder pages; plus Serasa's free lookup pages from the credit bureau giant, and the Federal Revenue's official lookup (free, but behind a captcha, with no relationship network). Sr. Watson entered late and now ranks mid-tier.

Few people have noticed it yet: no press coverage, no public revenue, no discussion on X, only spreading within Brazilian lawyer and OSINT circles. But the playbook it uses is already a red ocean in Brazil. It doesn't need anyone to copy it; it is itself a late follower.

### If it were you

**Verdict** : following it in Brazil isn't worth it; this playbook could be tried in a country where nobody has saturated it yet. What it proves is an old truth: government-published company registry data, with full-coverage page generation and internal links between companies and people, can still pull Google traffic today, even when you enter late.

**Who it suits** : a team of 1–2 people with technical SEO skills, able to handle tens of millions of rows of data, ideally familiar with a country that publishes company and shareholder data openly but where local-language lookup sites are still weak, and who have a group of professional users willing to pay (lawyers, accountants, lenders, recruitment background checks).

**Who it doesn't suit** : anyone trying to force their way into markets like Brazil, the UK, France or Germany that already have established incumbents (Brazil has more than a dozen similar sites; the UK and France are long since saturated by company information sites); anyone worried about privacy complaints or unwilling to handle deletion requests; anyone expecting the pages themselves to generate revenue (the free profile pages are just the entry point).

**How to do it** : first confirm that the target country's company registry data can be legally bulk-obtained and used commercially. One page per company, with the company name and registration number in the title; decide whether and how to build shareholder pages according to local privacy law, and launch the deletion entry point on day one; link pages together by shareholder, address and industry, with complete structured data. Monetization goes off-page: deep reports, bulk exports, monitoring alerts, APIs.

**How to acquire users** : acquisition is Google indexing. Submit sitemap shards, and first target uncontested terms like "company name + registration number". On the conversion side, find a person or community with reputation in the target professional circles, the way Bruno Fraga is for Brazilian lawyers and investigators; if there isn't one, at least get listed in those industries' tool lists.

**How long until you know** : at its pace, expect indexing and rankings to start climbing 2 to 3 months after the redesign. If there is still no meaningful organic traffic 3 months after launch, stop.

**What would show you misjudged it** : ① Sr. Watson's global ranking falls back below 300k before December, or sites like cnpj.ws that rose at the same time drop together: that means Google is cleaning up mass-generated pages like this and the playbook is failing; ② Brazil sees LGPD penalties or concentrated complaints targeting person pages: the risk of person pages is bigger than it looks; ③ Conversely, if it returns to within 100k by early 2027 and keeps crawling, that shows late entrants still have a shot in the red ocean, and a local follow-on in Brazil could be worth reassessing.

A closing note for product builders: traffic doesn't always require a new idea. A public dataset anyone can download, plus the patience to fill 67M pages with it, can also bring traffic. The hard part is when your turn comes, how much room is left in that market.

[Website ↗](https://srwatson.io)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| cnpj.biz | Web | Years ago | Tranco 19,215; CrUX br top 1k | Category leader, pure CNPJ profiles, no funnel from investigations / paid deep reports |
| Escavador | Web | around 2015 years ago | Tranco 14,832; CrUX br top 1k | Leader in name and court-case search, deeper on the person dimension |
| Jusbrasil(consulta CNPJ / nome) | Web | Years ago | Tranco 1,276 | Legal portal that also runs CNPJ and name pages; its scale is over a hundred times bigger |
| Econodata | Web | Years ago | Tranco 38,361; CrUX br top 5k | Mainly sells B2B sales lead lists |
| Casa dos Dados | Web | Years ago | Tranco 60,971 | Focused on bulk filtered company list exports |
| CNPJá / cnpj.ws | Web | Years ago | Tranco 66,717 / 36,564 | Leans toward developer APIs; cnpj.ws surged at the same time |
| Serasa Experian free company lookup | Big-company features | Years ago | Not verified | Free SEO pages from the credit bureau giant, leading to its credit products |
| advdinamico / maiscnpj / monitorcnpj / consultas.plus / cnpjcheck / consultasocio and others | Web | Mixed | Tranco 80k to 300k | Homogeneous page farms, many also with shareholder pages |
| Receita Federal official CNPJ lookup | Other | Official | Not verified | Free but behind a captcha, with no relationship network or person aggregation |

Where users come from

- **Verified** **SEO (programmatic, company pages + person pages)** : sitemap 3,373 shards × 20,000 URLs each, roughly 67M /empresa/ pages; complete JSON-LD; in Brave, searching company names puts srwatson.io/empresa at #5 and #9; /pessoa/ person pages can be found by name search; homepage title switched to an SEO title between 2026-03 and 05

- **Verified** **Founder Bruno Fraga's YouTube / IG / masterclasses** : @brunofragax (389k subscribers), an intro video from 2 years ago with 23k views; the Aulão channel's Sr. Watson and Sherlocker courses

- **Verified** **Lawyer / OSINT-circle KOCs and tool lists** : IG Reels "15 ferramentas", "#advocacianapratica"; ravenastar.com/xrav tool library; ainvestigacao.com coverage citing it

- **Verified** **TikTok** : @startup_cortes one video with 32k views, 2024-12; 2026-09 has short videos stoking privacy panic

- Inferred **Paid ads** : not in our Meta / Google ad libraries; the card says srwatson.co is running ads, unverified

## Sr. Watson (srwatson.io) research report

> In one line: a free CNPJ lookup tool that Brazilian OSINT influencer Bruno Fraga (YouTube 389k subscribers) built 2024 years ago. In spring 2026, the site was rebuilt into a programmatic SEO site: using the Federal Revenue's public data, it generates a separately indexable page for each of 67M companies and their shareholders (individuals). Once Google indexed them, traffic surged starting in June 2026. This growth didn't come from content distribution; it came from **going after the "consulta CNPJ" long tail in Brazil, a market that is already crowded** . The funnel's endpoint is PRO (R$90) and the paid platform for the same investigations, Sherlocker.

### 1. What it is

- Homepage: enter a CNPJ and get a profile in seconds: shareholders, CNAE, Simples Nacional, address, phone and email. The homepage reads "+67M empresas · +1M buscas · Dados oficiais da Receita Federal", targeting lawyers, accountants, compliance staff, lenders, journalists and investigators (https://srwatson.io, read 2026-11-07).

- Structured data on the page has `Organization.alternateName = "Sherlocker"`; the pricing is "Plano gratuito R$0 / Plano PRO R$90" (BRL, from the JSON-LD on /empresa/ pages; billing period not stated). At the bottom of company pages there's a row of "Investigações premium": Dossiê, Processos, Licitações, Dívidas, Cartório de Protestos, AI deep search... all paid entry points.

- **Who's behind it** : Bruno Fraga, a Brazilian OSINT / digital investigation educator who wrote "Técnicas de Invasão". His main YouTube channel @brunofragax has 389k subscribers; 2 years ago he posted "Criamos uma Ferramenta de Investigação Empresarial | Sr Watson" (23k views); the side channel "Aulão com Bruno Fraga" (38.9k subscribers) made a Sr. Watson masterclass about 1 years ago (1,920 views), and 2025-11-26 posted an intro to Sherlocker (6,739 views). Sherlocker (sherlocker.com.br) is his AI investigation paid platform, claiming 50+ data sources and the ability to look up CPF/CNPJ/hidden assets. Bruno Fraga himself has posted Sherlocker content on Instagram. Sources: YouTube pages read via r.jina.ai; Jina search results.

- Only a website; no app found. All traffic is on the web.

### 2. Traffic data (our traffic database)

- Tranco: 2026-09-03 first entered the top 1M (901,225), then kept climbing: 09-24 343,133 → 10-15 129,279 → **10-27 peaked at 99,952** → 11-05 108,966, and it has slipped slightly for 9 consecutive days. Tranco is a 30-day rolling window, so it reflects real traffic from roughly late September to mid October, meaning growth had stalled in that period and may even have started to fall.

- CrUX Brazil: top 2026-05 500k → top 06 500k → **top 07 100k → top 08 50k → top 09 50k** . Globally it's in the top 07-09 1M tier. Real Chrome users moved up one tier in July, another in August, and plateaued in September.

- Users are almost all in Brazil. Of the 21 countries in CrUX, only br has it.

- The timing lines up: Wayback archives show that on 2026-03-23 the homepage title was still "Investigação Empresarial", and by **2026-05-19 it had changed to "Sr. Watson — Investigação empresarial completa por CNPJ"** , with "Seja PRO / Acessar Conta" now showing. So the redesign and SEO push happened between 2026 3 and May. Google indexed it and rankings climbed from 6 to August, and Tranco only picked it up from September on.

### 3. Growth lever: programmatic SEO (verified)

- `robots.txt` allows everything. `sitemap.xml` is **3,373 child sitemaps with 20,000 URLs each, roughly 67M /empresa/<name>-<CNPJ> pages in total** , which matches "+67M empresas". lastmod uses the company's founding date.

- Every company page has a full SEO setup: title "<razão social> — CNPJ xx.xxx.xxx/xxxx-xx | Sr. Watson", meta description, canonical, and JSON-LD (Organization with taxID, address, phone, email, member; FAQPage with 7 Q&As; SoftwareApplication with pricing; SearchAction). Shareholders link internally to **/pessoa/<name>-<middle 6 digits of CPF> person pages** , which list every company tied to that person and their age bracket. Company pages also link to each other, chained by adjacent CNPJs. Static assets are served under `/indexacao-static/`, which looks like a separate deployment built just for indexing.

- Indexing check (Brave search, country=BR): searching "FRW SERVICOS MEDICOS LTDA cnpj" puts srwatson.io/empresa/... at #5 and #9, sharing the results page with econodata, jusbrasil, serasa, solutudo and maiscnpj. A Jina search for "\"Sr. Watson\" CPF exposto" directly returns several srwatson.io/pessoa/ person pages. **Name searches are also feeding it traffic** .

- The person-page layer is what sets it apart from most CNPJ sites (advdinamico, econodata/contato and escavador have something similar), so it captures both company/CNPJ long-tail queries and person-name long-tail queries. Every page has a "solicitar-remocao" (removal request) link, presumably to deal with LGPD.

### 4. Other channels (minor)

- **The founder's own content** : Bruno Fraga has been promoting Sr. Watson on YouTube, IG and masterclasses since 2024, which is where the early seed users came from. But those videos are 1–2 years old with modest views (23k, 1,920), so they don't explain the sharp rise from 2026 7 to September.

- **Organic recommendations from professionals** : lawyer IG Reels such as "ADVOGADO(A), PREPARA O PRINT! Separei 15 ferramentas…" (DefuNHmM8QS) and "Você conhecia essa ferramenta? #advocacianapratica" (DN5rmJ3jgzV); the OSINT tool list at ravenastar.com/xrav ranks "Sr Watson — Investigação Empresarial" at #160; delegacia.org's tool list includes it; and the investigative outlet ainvestigacao.com cites a "consulta à base Sr. Watson" in its reporting.

- **TikTok** :@startup_cortes 2024-12-02 发过一条「Sr Watson site para saber se o CNPJ é verdadeiro ou falso」,3.2 万播放、1,002 赞。还有 @julianacidralvanvossen 2026-09-24 发的「Você já ouviu falar no site Sr. Watson? Se você tem ou já teve um CNPJ, corra para dar u[ma olhada]」,只有 127 播放,但这是**隐私恐慌型** 内容,和人物页被人搜到的情况对得上(TikHub 搜索)。

- **Ads** : Meta and Google ad libraries show no records with srwatson or sherlocker in product_domain. The card says srwatson.co is also running ads, but I couldn't verify that (srwatson.co is an old domain, used in a YouTube video description from 2024).

- X: x-radar has no recent mentions of it in the last 90 days. No press coverage found.

### 5. Competitors (Brazilian CNPJ lookup / company investigation)

| Product | Form | Scale (Tranco 11-05 / CrUX br 2026-09) | How it differs from Sr. Watson |
|---|---|---|---|
| cnpj.biz | Web pSEO | 19,215 / top 1k | Category leader; pure CNPJ records, no "investigation" or relationship-graph positioning |
| Escavador | Web | 14,832 / top 1k | Leader in person-name and court-case search; far deeper on the person dimension than Sr. Watson |
| Jusbrasil (/consulta-cnpj, /nome/) | Web | 1,276 | A legal portal that also builds CNPJ and name pages as a side product; hundreds of times its size |
| Econodata | Web pSEO + B2B leads | 38,361 / top 5k (down from 1k in September) | Mainly sells sales leads, with /contato/ person pages |
| Casa dos Dados | Web pSEO + lead export | 60,971 (still rising) / top 5k | Mainly exports company lists in bulk by filter |
| CNPJá(cnpja.com) | Web + API | 66,717 (once 42k in August) / top 5k | Developer API and lookup |
| cnpj.ws | Web + API | 36,564 (310k in July, surged over the same period) / top 100k | Another pSEO/API site that took off in the same window |
| Serasa Experian free company lookup | Big-company features | — | Free SEO pages from a credit-bureau giant |
| advdinamico, maiscnpj, monitorcnpj, consultas.plus, cnpjcheck, informecadastral, consultasocio, solutudo, empresaqui | Web pSEO | Between 80k and 300k names | Homogeneous long-tail page farms, many of which also build shareholder / person pages |
| Receita Federal / gov.br official lookup | Official | — | Free but behind CAPTCHA, with no relationship network |
| Sherlocker (same company) | Paid SaaS | Tranco 820k to 980k, marginal | Paid in-depth investigations, downstream of the Sr. Watson funnel |

Takeaway: the **category is badly crowded** , and all of it is built on the same public data (the full Receita Federal CNPJ dump). Sr. Watson started late and now has about 100k names, mid-pack among CNPJ sites.

### 6. Why it took off

1. **Copy a playbook that's already been validated and use a new domain to pick up incremental traffic** : 67M company pages plus person pages and full JSON-LD. Redesigned by May 3, indexed by Google through August 6, CrUX br moved from the 500k bucket up to the 50k bucket. See the sitemap and indexing check above.

2. **Person pages capture extra name long-tail** : shareholders are individuals, so building a page per name means searches for a person's name land on it (the "exposto" queries return /pessoa/ pages directly).

3. **Brand foundation** : Bruno Fraga is known in Brazil's OSINT circles, and lawyers and investigators recommend the tool organically (IG tool lists, OSINT tool libraries, investigative press citations). Google may give it some credit on E-E-A-T and backlinks, though this can't be quantified.

4. **Why now** : the only thing I can confirm is the spring 2026 redesign itself. No external platform event is visible.

### 7. Differentiation

- The stated differentiation is "investigation / due diligence" positioning: relationship graphs, UBO, economic group maps, heat maps, with funnel traffic routed to the AI deep-investigation product (Sherlocker).

- **But that differentiation is not what drives the traffic** . The SEO pages drive it, and their content is basically the same as cnpj.biz, econodata and others (same Receita data). The differentiation only matters at the conversion end: for the same CNPJ record page, competitors push "buy a lead list," while Sr. Watson pushes "paid investigation / Sherlocker," which means higher ticket size and a more professional audience (lawyers, compliance teams, investigators).

- Risks: Tranco has been declining since 10-27. Page farms like this are vulnerable to Google core updates (thin / duplicate content), and the person pages carry LGPD complaint risk (TikTok already has "go check your data" content, and Reclame Aqui's check gave it a 1/5 shield).

### 8. Visibility

Almost nobody knows about it. No press coverage, no public revenue, no one talks about it on X. It only circulates in small circles of Brazilian lawyers and OSINT people. The approach itself (CNPJ pSEO) has been a red ocean in Brazil for a long time and doesn't need copying. It's just one of the late followers.

### 9. Takeaways

- "Build pages from the full public government dataset + entity linking (company ↔ person) + JSON-LD + a paid in-depth report entry point on every page": this works in any country with public company and officer registries.

- Split the work: use the verticals that creators and educators already reach (lawyers, investigators) for conversion, and use SEO for acquisition.

- Before building person pages, prepare the removal flow and privacy compliance first.

### 10. Uncertain

- Real traffic, paid conversion and PRO billing cycle (is R$90 monthly or annual?) are all unknown.

- Whether the 3 May redesign also launched "67M pages" at the same time can't be confirmed: the Wayback homepage snapshots don't show company links, since the homepage doesn't link to them, so there's no way to tell which day the sitemap went live.

- Srwatson.co ad activity and the conversion relationship between Sherlocker and Sr. Watson are unverified.

- Whether the late-October dip is caused by a Google update or is normal fluctuation needs 2–4 more weeks of observation.

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.