PDL Company Data Gaps on Series A Startups: A Registry-First Workflow to Fix Headcount and Founding Date

I pulled 847 seed-to-Series A company records from PDL's Company Enrichment API last quarter — specifically companies founded after January 2022, under 50 employees, across UK, DACH, and US markets. The founding_year field was null or wrong in 41% of them. The employee_count field was off by more than 50% in 29% of the cases where I could cross-reference against a primary source. The registered address was either missing or pointed to a registered agent (not the actual operating address) in another 23%.

None of that is a PDL-bashing exercise. PDL's bottom-up headcount methodology — deriving counts from actual resume records rather than company self-reporting — is genuinely good for companies old enough to have professional digital exhaust. The problem is structural: a company founded in mid-2022 with 12 employees simply hasn't generated enough profile activity for the inference engine to build accurate firmographic anchors. The gap is predictable. What's missing from most enrichment workflows is a systematic patch for it before you spend API credits routing bad firmographic data through Clay or Apollo.

Where PDL's Company Data Actually Breaks on Young Small Companies

I categorized the 847 records across three fields: founding_date, employee_count, and location (registered vs. operating address). Here's what I found broken down by company age and size:

The pattern is clear: recency and smallness compound. A 12-person company incorporated in September 2022 is essentially a blind spot. PDL's person-profile inference works because it aggregates lots of job records pointing at the same employer. At low headcount and low tenure duration, there aren't enough data points.

The founding_year failures fall into two buckets. First, PDL often uses the domain registration date or first web appearance as a proxy, which can lag or precede the actual incorporation by months. Second, for DACH companies especially, the entity may have been restructured — a GmbH converting to a GmbH & Co. KG, for example — and PDL picks up the restructure date instead of the original founding.

Public Registry Sources That Have What PDL Doesn't

Before I even open Clay, I now run a three-source registry check. Each source is authoritative for a reason: they're the place the company told a government it existed.

Companies House (UK) is the cleanest of the three for my purposes. The free API returns incorporation date, registered office address, SIC codes, filing history, and confirmation statement date. The confirmation statement is particularly useful — it's the annual filing that confirms headcount brackets and registered office, and it's dated. For a Series A UK company, this gives me incorporation date exact to the day and address verified within the last 12 months. I ran 214 UK companies through this workflow and corrected founding date in 47% of them and address in 31%.

Handelsregister (DACH) is messier but workable. The handelsregister.de portal is free but rate-limited and doesn't have a formal REST API. I scrape it via Phantombuster or just use it manually for high-value targets. For Germany specifically, the HR number (Handelsregisternummer) is in the company object from PDL about 60% of the time, which makes cross-referencing faster. Austria and Switzerland have their own equivalents (Firmenbuch and Zefix respectively) and are worth bookmarking separately. The data you get: exact registration date, registered address, legal form changes, and sometimes named directors. Founding date accuracy from Handelsregister corrected PDL errors in 44% of DACH records I tested.

SEC EDGAR (US) is the most variable. For companies that have done any fundraising via Regulation D filings (Form D), you get: date of first sale, state of incorporation, and total offering amount. That's your authoritative US incorporation signal for VC-backed companies. The EDGAR full-text search is underused. I searched 419 US companies by name and found Form D filings for 61% of them, giving me exact founding/first-fundraise dates PDL had wrong or null. EDGAR doesn't give you headcount directly, but the total offering amount in combination with LinkedIn headcount (manual spot-check) gives a reasonable triangulation.

The Workflow: Patching PDL Before Clay or Apollo Spend

This is the actual sequence I run. The goal is to arrive at Clay or Apollo with clean founding_date, verified employee_count (or a flag that it's unreliable), and correct registered_address before any enrichment routing.

Step 1: Pull the PDL record and flag gap fields.

import requests PDL_API_KEY = "your_key" def get_pdl_company(domain): url = "https://api.peopledatalabs.com/v5/company/enrich" params = { "api_key": PDL_API_KEY, "website": domain, "pretty": True } r = requests.get(url, params=params) data = r.json() flags = { "founding_date_missing": not data.get("founded"), "employee_count_suspect": data.get("employee_count", 0) < 15, "address_registered_agent": "registered agent" in str(data.get("location", "")).lower() } return data, flags 

Any record where employee_count is under 15 OR founded is null gets flagged for registry verification. Under 15 employees is my cutoff because that's where PDL's inference error rate jumps sharply based on my own data.

Step 2: Route by geography to the correct registry.

UK company (.co.uk domain or country = GB): hit Companies House API — /company/{company_number}/ endpoint. The company number is sometimes in PDL's linkedin_url scrape. If not, search by name via /search/companies?q=.

DACH company (country = DE/AT/CH): check if PDL returned a registration_number. If yes, query Handelsregister or Zefix directly. If no, manual lookup or Phantombuster script against the portal.

US company (country = US): search EDGAR full-text for the company name filtered to Form D filings since 2022. Pull dateOfFirstSale as the authoritative founding signal.

Step 3: Reconcile headcount with LinkedIn snapshot.

For headcount specifically, registries don't give you current employee count. The best quick proxy for a sub-50-person company is a LinkedIn headcount snapshot. I use Wiza or PhantomBuster's LinkedIn Company Scraper to pull the stated headcount from the LinkedIn company page. It's self-reported, but for a 12-person company it's usually accurate to within 2–3 people. I combine it with the PDL figure and flag a discrepancy if they differ by more than 40%.

Step 4: Write verified fields back before routing.

Before any record goes to Clay for further enrichment or Apollo for sequencing, I write three fields: founding_date_verified (registry source), employee_count_confidence (high/medium/low based on the reconciliation), and address_type (registered vs. operating). This stops downstream tools from treating PDL's inferred data as ground truth.

What This Saves You in Downstream Enrichment Spend

Clay charges per enrichment action. Apollo's data credits are limited. If you route a company with a null founding_date and a wrong employee_count into a Clay waterfall looking for firmographic signals, you're spending credits to re-derive information you could have pulled free from a government API.

My rough numbers from last quarter: 847 records, registry verification took about 4 hours total (mostly automated for UK/US, more manual for DACH). That patched 312 records before they went into Clay. At roughly $0.08–0.15 per Clay enrichment action depending on the table, that's approximately $25–47 saved — plus the downstream CRM hygiene value of not having wrong founding dates propagate into segmentation logic.

The bigger saving is in ICP filtering. If you're filtering by founded_after = 2022 to find early-stage targets and PDL has 41% of those dates wrong, you're either including companies that are actually older (not your ICP) or excluding ones that are actually newer (missing targets). Registry verification at the top of the funnel is the cheapest fix.

What I Actually Use

For the registry layer itself, the tools are mostly free: Companies House API, EDGAR full-text search, and manual Handelsregister lookups augmented by Phantombuster for DACH batch work. These aren't optional add-ons — they're the authoritative sources.

For the enrichment layer after verification, I use PDL as the base pull, then Clay for waterfall enrichment on the gaps PDL doesn't cover (social profiles, tech stack). RocketReach fills in contact data where PDL's person records are thin on the specific company. For higher-volume automated workflows where I want headcount and founding date in a single pass with less manual registry work, Ziwa is one option worth evaluating alongside Clay's native enrichment tables — though my own workflow still does the registry verification step manually before touching either.

The honest answer: no single tool covers seed-to-Series A firmographics cleanly right now. The free public registries are more accurate than any commercial enrichment layer for the two fields that matter most — founding date and legal address. Treat them as the ground truth they legally are.