Our AI pipeline, in detail
How we find your one perfect deal in 1,000 insolvency notices.
Others just aggregate. We aggregate, deeply research, embed semantically, and match against your profile. Six stages. Transparent. Publicly documented.
No credit card for the 14-day trial. Public data sources. GDPR-compliant.
Six stages. From raw signal to ranked deal.
Click a stage for details. All numbers are live values from the production system (as of July 2026).
All stages in detail below ↓
Stage 1
Aggregate from 5 public sources.
Every night (and hourly for new publications) we crawl §9 InsO, DUB, nexxt-change, Mabya and dinsob — the only five sources that publish meaningful volumes of DACH company-acquisition signals.
- §9 InsO — the official insolvency stream, ~450 notices/day across Germany
- DUB, nexxt-change, Mabya, dinsob — direct for-sale listings from owners
- Cross-source deduplication: same case on 3 boards = 1 lead with us
What competitors do
Competitors cover only §9. Direct for-sale listings (often the more interesting deals) are missing entirely.
Stage 2
Filter out 70 % — only companies survive.
Two thirds of all §9 notices are consumer insolvencies (private individuals). Worthless for buyers. We filter them out with a trained legal-form classifier.
- Consumer insolvencies automatically discarded
- Only Kapitalgesellschaften, partnerships, and sole proprietorships kept
- ~170 usable companies per day instead of ~450 raw notices
What competitors do
Competitors deliver the full 500-count in a table. The buyer scrolls and filters themselves.
Stage 3
12 additional data sources per company — automated.
For each remaining company an enrichment job runs: website, Handelsregister, Bundesanzeiger, Impressum, Wayback Machine, Trustpilot, Google News, SearXNG, HR-AD extracts, DPMA trademark registry, WHOIS and an LLM summary. 30-40 seconds per case. Fully automated.
- Handelsregister extract + business purpose
- Bundesanzeiger financials (revenue, employees, balance total)
- Wayback Machine — works even if the company is offline today
- News sentiment via Google News + SearXNG
What competitors do
Competitors show you 3-5 data points from the §9 feed. Everything beyond that, you research yourself — 3-4 hours a week.
Stage 4
Semantic embedding — 1536-dimensional per company.
From the enriched company description we generate a 1536-dimensional vector with OpenAI text-embedding-3-small. This vector IS the semantic signature of the company — it understands meaning, not just keywords.
- 1536 dimensions — industry standard for semantic search
- Stored in Postgres with the pgvector extension
- Nightly batch — new companies immediately searchable
What competitors do
Competitors have full-text search only. "D2C Shopify" finds nothing when the text says "online retail".
Stage 5
Match against your buyer profile.
You describe in one sentence what you want to buy: "D2C Shopify shops in NRW, €20-100k revenue, English-friendly, burnout shutdown." We vectorise your profile — and compute cosine similarity against every company.
- Natural-language profiles instead of rigid filter forms
- Cosine similarity → 0-100 % match score per company
- Multiple profiles in parallel (e.g. "solo deal" + "holding deal")
What competitors do
Competitors offer dropdowns: sector, region, revenue band. Everything else falls through the cracks.
Stage 6
Rank with the Buyer-Interest Score (0-100).
On top of semantic Match %, we compute a proprietary Buyer-Interest Score from five signals (20 points each). Together they form your daily ranking.
- URL resolved = 20 points (website reachable → business still operating)
- Rich deep summary = 20 points (lots of public info available)
- News activity = 20 points (recent press footprint)
- Debtor-led procedure = 20 points (Eigenverwaltung, restructuring signal)
- Bid-phase timing = 20 points (now IS the critical window)
What competitors do
Competitors sort chronologically (new → old). No assessment of attractiveness.
The Buyer-Interest Score — transparent, not magic.
Five signals, 20 points each, openly documented. Drag the sliders to see how the score changes.
Drag the sliders ↔
The company's website is still reachable (via Wayback + live check). Signal: still operating, addressable assets.
Handelsregister, Bundesanzeiger, news and web footprint yield lots of public info. Fewer blind spots on the first call.
Fresh press mentions (last 90 days). Signal: active proceedings, media attention, possibly early investor conversations.
Eigenverwaltung or Schutzschirm rather than regular insolvency. Signal: management cooperating — better chances for a structured transition.
Filing deadline running, report/examination date upcoming. Signal: NOW is the outreach window.
Match % — not an IDEAL/STRONG mystery, just a number.
Your buyer profile is semantically compared to each lead. Green lines = fit, grey = doesn't. No "excellent" label without reasoning.
- Sector
- D2C E-commerce
- Revenue
- €20-100k
- Region
- NRW
- Buyer type
- Operator-buyer (solo)
- Capital
- €25-75k cash
- Sector
- Online-shop cosmetics
- Revenue
- €48k (2025)
- Region
- Cologne (NRW)
- Procedure
- Eigenverwaltung
- Asking
- €120k asset deal
- Sector
- D2C E-commerce
- Revenue
- €20-100k
- Region
- NRW
- Buyer type
- Operator-buyer (solo)
- Capital
- €25-75k cash
- Sector
- Online-shop cosmetics
- Revenue
- €48k (2025)
- Region
- Cologne (NRW)
- Procedure
- Eigenverwaltung
- Asking
- €120k asset deal
Non-matching criteria are also shown — so you decide whether it's a deal-breaker.
Data completeness per lead — head-to-head.
We re-checked 50 identical cases in our system and in the three largest competitors. Result:
Based on public analysis of 50 random cases from the §9 stream (June 2026). Values in % of cases where the field is populated.
From court notice to email — in under 30 minutes.
Court publishes. Our poller picks it up. Enrichment runs. Your alert fires. All in one coffee break.
- 14:23Amtsgericht Hamburgpublishes §9 InsO opening decision for Musterfirma GmbH
- 14:41Übernahme-Radaringest cron picks up the notice, enrichment starts automatically
- 14:52Match engine3 active buyer profiles match — alerts get triggered
- 14:53Your inboxFirst email with administrator contact, Match %, deep summary
vs. competitors
Competitors run 24-hour batches: "Next import: sometime tomorrow morning."
A real example. This is what you get per lead.
Anonymised deep summary from the production system. This is what a lead looks like in your inbox:
M. Kosmetik & Beauty GmbH
D2C cosmetics, online retail · Cologne, NRW
Smaller online shop for natural cosmetics, founded 2019, Shopify-based. Latest published revenue (2024): €48k, declining. Imprint + Instagram active, latest news mentions a Q3 2025 product-line extension. Insolvency in Eigenverwaltung — points to a structured transition. Administrator: Dr. M. Schmitt (firm has been active in D2C for 12 years, historical response rate ~72 %). Filing deadline runs until end of July. Attractive for: operator-buyers with €50-100k cash, D2C experience, interest in the beauty category.
Already available for 470 companies in the system.
See public example →14 days free.
No credit card. One buyer profile, unlimited deep-research runs. Cancel monthly on every plan.