Frequently asked questions

Surfacer is an intent engine for B2B sales teams. Every morning it tells you which companies in your market are showing signs of buying behaviour today, new hires, leadership changes, tech migrations, news, homepage shifts, and ranks them against your ICP and your closed-won pattern.

The output is a daily call list with reasoning, not a static database to filter through. See the platform tour for screenshots of each surface.

B2B sales teams in the DACH SMB and mid-market space — typically companies selling SaaS, IT services, consulting, or industrial products to organizations with 10–500 employees. The product is sharpest when you have a defined ICP and an existing CRM (HubSpot today, more soon).

No. The default feed works out of the box once you sign up — it learns from your swipes (yes / pass) and from your HubSpot closed-won pattern if you connect it. Custom signal rules are an optional power-user feature for when you want to alert on specific patterns.

You see ranked companies on day one. After about 5 swipes the personalization layer kicks in; by ~50 swipes the per-user preference model is fully warmed up and shifting scores noticeably. Connecting HubSpot accelerates this because we can build a lookalike centroid from your closed-won customers immediately.

Six families:

• Hiring, open positions, role types, seniority, hiring intent (new team vs. scale vs. backfill), failed-repost detection.
• Product, product/service catalogue changes, new launches, pricing-page redesigns, feature additions.
• Funding, rounds raised, investors named, exits, acquisitions, ownership changes.
• Press, mentions in DACH trade press, regional papers, podcasts, category coverage.
• Leadership, managing-director changes, exec hires and departures, board moves.
• Tech & growth, CMS migrations, stack changes, company-size drift, homepage semantic drift between crawls.

No individual signal is decisive, combinations of three or four signals tightly clustered in time are. See the longer write-up on /about.

Each company gets a blended score: ICP fit × 0.6 + your preference fit × 0.4, with a HubSpot lookalike centroid layered in when available. ICP fit uses embeddings + taxonomy matching; preference fit uses Q-values updated from every swipe; lookalike uses cosine similarity against the average of your closed-won customers’ homepage embeddings.

Every score travels with a confidence range — sparse companies get wider ranges, signal-rich companies get tighter ones.

Every yes / pass updates a per-user preference profile across four dimensions: size, industry, location, and feature signals. We also capture dwell time and pass reasons to debias the learning — a fast pass weighs differently from a deliberate one. After ~50 swipes the model surfaces a learned-preferences card on your dashboard showing how it’s reweighting signals on your behalf.

Publicly available sources only: company websites, careers pages, the major job-board APIs and ATS providers, news and RSS feeds, German Impressum pages, the Handelsregister, the Bundesanzeiger, and the Marktstammdatenregister for the energy sector. No personal-profile scraping. No grey-zone data brokers.

Event-driven, not cron-driven for monitored companies — if a company opens a job posting, you’ll typically see it in the next dashboard refresh. Companies you actively pin into your monitoring list get a nightly diff and a daily digest email summarising state transitions only (no noise about unchanged signals).

Significantly better than US-built tools. Our extraction pipeline is engineered for German Mittelstand sites, including those behind Cloudflare, Radware, and other WAFs that US tools tend to give up on. We use a six-tier HTML fetch cascade (aiohttp → cloudscraper → curl_cffi → Playwright → Google Cache → Wayback) so we routinely pull data where US-built tools return nothing.

Impressum pages (legally required on every .de site), Handelsregister filings, Bundesanzeiger announcements, and Marktstammdatenregister for the energy sector. We extract managing director, legal form, VAT number, and address even when the company isn’t on LinkedIn.

Yes — a dedicated second pass detects multiple legal entities on the same Impressum page (common for German holdings with several brands) and infers the connections between them. Names with German titles like Prof. Dr., Dipl.-Ing. are normalized automatically.

Yes — the Impressum requirement and Handelsregister-equivalent exist in both AT and CH and we extract similarly. Coverage on AT and CH is currently a bit thinner than DE because the bulk of our crawl history is German, but the pipeline is the same.

Yes — fully. OAuth connect, push single or bulk companies into HubSpot as companies + contacts with a five-minute undo window, bidirectional blocklist sync (import your HubSpot exclusion list, export domains you block in Surfacer), and a HubSpot closed-won lookalike model that ranks new prospects against your existing customers automatically.

Salesforce and Pipedrive are on the near-term roadmap. In the meantime: CSV export of any feed / company list works today, and we expose a webhook + REST API for custom integrations.

Yes, pin up to 100 companies (Pro plan) into your monitoring list. We diff them nightly across growth, drift, news, jobs, and pricing signals and only surface state transitions (no noise about unchanged data). A daily digest email summarises everything. See the platform tour for what the monitoring view looks like.

Three options today: (1) the /dashboard Today’s Call List with bulk-push to HubSpot, (2) the swipe feed at /feed, or (3) CSV export. Slack / Teams / daily email push are on the roadmap — see below.

Yes — the Signal Composer at /settings/signals/new lets you assemble rules from a closed vocabulary of atoms: industries (192 categories), regions, size buckets, revenue bands, money bands, event types (hiring spike, leadership change, M&A, funding, …) and time windows. Plain-English summary auto-renders. Six starter presets are one-click.

It tells you how expensive the rule is to evaluate against new articles — in real time as you build it:

• Green — fully cached. Atoms-only rule, matches against an existing customer’s rule, zero LLM cost.
• Yellow — new atom combination. One pass to seed the cache, then green forever.
• Orange — you added a free-text rider. LLM verifier runs per article.
• Red — broad, vague free-text rule. LLM calls fan out widely; consider tightening with atoms.

Yes — on save, the rule backfills against the historical article corpus and against any matches already cached from other customers with the same rule shape. You see results within seconds for cached rules, and a few minutes for fresh ones. New articles match in real time from that point on.

Free tier: 50 credits / month, 5 monitoring slots, no credit card. Pro: €47 / month with 1,200 credits and 100 monitoring slots, 14-day free trial. Team plan is in private preview — see /pricing for the full breakdown.

One credit surfaces one new company into your feed. Contact lookups (Hunter.io enrichment) cost five credits; bulk discovery tasks cost ten. Unused monthly credits don’t roll over, but rollover credits from top-up packs never expire.

Yes — subscriptions cancel cleanly from /billing/portal. No commitment, no clawback. You keep access until the end of the paid period.

Yes — 14 days of Pro, full features, no credit card friction on the free tier itself. /gratis-test walks you through the German-language onboarding.

Yes. We only process publicly available data about legal entities (not private individuals), every signal traces back to a transparent source, and we offer opt-out for companies that don’t want to be in the active monitoring pool.

Google Cloud, region europe-west3 (Frankfurt). All Cloud SQL data and all storage stays in the EU.

Yes — we have a German AVV (Auftragsverarbeitungsvertrag) and an English DPA template ready. Write us at hello@surfacerai.com and we’ll send it the same day.

No. Contact discovery is a per-user, on-demand integration with Hunter.io. Found contacts are stored per-user (never merged into the shared company database) and every Hunter API call is audited. Your contact data is yours.

Data portability is part of the contract. Everything you put in comes out as a structured export (JSON + CSV). For enterprise customers we offer code-escrow on request.

Premium social-network filters tell you a person exists in a role. They don’t tell you their company just opened three DevOps roles, swapped CRM, or had a leadership change last week. We watch behavioural signals across the open web (jobs, news, homepage drift, hiring intent), score them in time, and learn from your closed-won pattern. Different category, not a better filter.

Workbench tools are flexible kits where you build columns, chain APIs, and write prompts per row. We’re the assembled product: signal taxonomy, scoring, ranking, and a per-user learning loop ship out of the box. You don’t maintain workflows; the system gets sharper on its own as you swipe and as your CRM grows.

Static contact databases tell you who exists. We tell you when to call. Most large databases are US-built and lose a lot of coverage on German SMBs (in our internal tests, the leading US options sit around 20–30% DACH SMB coverage). We’re complementary, if you already have one, we add the timing layer on top.

The strongest DACH-native tools today focus on inside signals, who visited your website. Surfacer focuses on outside signals, who’s moving in the market before they ever hit your site. The two together give you a 360° view; many customers run both.