Web-scraping data audits

Your vendor delivers data. We show you what's missing.

We run one of your vendor's scrapes ourselves, compare it head-to-head, and hand you the gaps: what's missing, why, and the evidence. Yours to keep and verify. Free.

Healthcare data Real estate Pricing intelligence Alt data Advertising intelligence Travel & hospitality

Start with only a URL, instructions, and 5–20 sample rows. No system access. No vendor contact. NDA welcome.

400+ Sites under management
150M+ API calls served every month
500M+ Raw records processed monthly
Diagnostic

Missing data is silent.

Spot-checks catch errors in what your vendor delivers. They can't catch what isn't there. Samples are blind to absences by design. Only an independent comparison surfaces them.

Standard sample coverage 0.005%
Q1

How do you know your vendor's deliveries are 100% accurate?

Q2

How do you know what they didn't get?

Q3

If they can't scrape a site, could someone else unlock that revenue?

Q4

Could someone do this cheaper, with modern approaches, and catch more than your current vendor?

The Gap Audit answers all four, and hands you the proof. →

Method

How the Gap Audit works. Nothing hidden.

No black box. Here's the exact procedure, so you can judge the test up front and re-check our numbers against your own delivery after.

Step 01Independent re-run

We scrape the same source ourselves, from scratch

Your URL list and instructions, run against the live source independently — no contact with your vendor's pipeline. Two independent pulls of the same ground truth.

Step 02Head-to-head comparison

We diff our pull against your vendor's last delivery

Record by record, field by field — every absence, every missing field, every value that disagrees. You start with a sample; the full comparison needs the vendor's last delivery, which you send once we begin.

Step 03Dual-layer QA

Dropped-record check + LLM-assisted targeted sampling

A dropped-record check catches every silent absence; LLM-assisted sampling concentrates human review on the records most likely to be wrong. We tell you exactly which records were checked, and how.

Step 04Like-for-like discipline

A fair test, not a stacked one

In-scope against in-scope only, de-duplicated, same time window. If the gap shrinks to zero under a fair comparison, the report says so.

Definitions

We only count a record if it's real, in-scope, and yours to use.

Every number in the report is built on this bar. A record counts only when all of it holds — otherwise it doesn't go in our favor.

UniqueA distinct entity, not a second copy of one already counted.
In scopeInside the scrape's agreed boundary — right state, right category, right target.
Live at sourcePresent on the live page right now, not a removed or expired entry.
Within instructionsCovered by the URL list and scraping instructions you gave us.
Parsed to schemaStructured into the fields the delivery requires, not raw or malformed.
Not staleCurrent as of the comparison window, not carried over from an older pull.
Not duplicatedDe-duplicated across every source before anything is counted.
Not out of scopeAn out-of-state or off-target page isn't "more data." It's the wrong data.
Evidence

We audited the data they trusted. Here's what was missing.

What a finding looks like, row by row.

Illustrative structure, not real client data. Every figure in your report traces to rows like these — each with the live source URL, so you can re-check any one of them yourself.

Vendor recordDataweav recordSource URLField issueWhy it counts
MissingPresentexample.com/provider/123Provider omittedLive page, in-scope state, matches schema
PresentPresentexample.com/provider/456Identifier missingHidden ID found in the page source
PresentExcludedexample.com/provider/789Out of scopeDifferent state / duplicate / stale — not counted
Where it goes from here

Starts with the audit. Goes as far as the proof earns.

01

The Gap Audit

We run one of your vendor's scrapes independently and hand you the findings: what's missing, why, and the evidence. Yours to keep and verify.

Free
02

The walkthrough

A 30-minute live review of the findings: what's missing, why it matters, and what it's worth fixing. No obligation to go further.

Free
03

Fixing the gaps

Where you want it closed, we re-engineer the broken scrapes and re-run the audit so the gap can't reopen silently.

04

We take over the scraping

When it makes sense, we run the scrape end-to-end: our own infrastructure, the dual-layer QA bundled in, priced as a swap, not another line item.

The artifact

Exactly what you'd get.

Not a teaser. The Gap Audit report is the deliverable, and it's yours to keep. Below is a redacted sample from a real engagement: the gap, the cause, the evidence trail, and the per-source breakdown you can verify against your own delivery.

View the sample report →
Gap Audit Report Redacted · Sample
Client████████████ data co.
Source auditedNationwide provider directory
Records in vendor's last delivery1,212,000
Records in our independent run1,291,000
Coverage gap (like-for-like)
In-scope records missing+79,000 · +6.5%
Likely cause██████ traversal depth
Verification trail
Dropped-record checkPass · full reconcile
Records re-checked at source██,███

Every figure above is traceable to the row-level appendix, so you can re-run the comparison yourself.

Gap Audit

Request your Gap Audit. Yours to verify.

Free · 10 Days

Run by the operators behind 400+ live scraping sites. You'll review the findings with the people who built the audit — not an SDR.

Not sure what to share? Book a 20-minute call to scope the audit together, before any data changes hands. Scope before sending data →
What you give us URL + instructions · Plus 5–20 sample rows from your vendor's last delivery.
What we'll need to run it Your vendor's full last delivery. You start with a sample; the full comparison needs it. NDA available before you send, redacted fields where possible, and we don't contact your vendor.
What you keep The PDF Gap Audit Report + evidence trail · yours to keep and re-verify, with a 30-min live walkthrough of the findings. No obligation to engage us.
What's the engagement Taking over the scraping, fixing what's missing, and the ongoing bundled QA: the managed service the audit opens.
Timeline 10 business days from handoff.
Sample-first You start with 5–20 rows. We never ask for the full dataset before you've decided to proceed.
Data handling Used only to run your audit — never resold, never used to train models. No access to your systems. Deleted on request, NDA available before full delivery. Privacy policy →

Best fit: teams already paying for recurring scraped data — not one-off jobs.

Use your business email. Personal addresses (Gmail, Hotmail, etc.) won't be accepted.

No files selected

Upload 5–20 sample rows from your vendor's last delivery (CSV, XLSX, JSON). We'll ask for the full dataset once we begin.

We don't share your data.

Questions, or want to send data another way? Reach us directly: hello@dataweav.com
FAQ

"What's free, and what do I pay for?"

The audit and its findings are free and yours to keep: what's missing, why, and the evidence. Verify it and walk away if you like. The paid part is the managed service: we take over the scraping, fix what's missing, and ship the bundled QA on every delivery going forward.

"What if your numbers are wrong?"

You'll catch it, because every figure is traceable to row-level evidence you can re-check at the source. The dropped-record check catches what spot-checks miss, and the report tells you why in plain language.

"We don't have budget for another tool."

You don't need new budget. The audit is free, and the managed service replaces your current vendor's line item. We price comparable and bundle the QA at no premium. A swap, not an addition.

"How does the audit wrap up?"

We send the report and walk you through it: what's missing, why, and what it's worth fixing. The 30-minute walkthrough is where the value lands. The report stands on its own whether you proceed or not.

Next step

We'll prove it on your real data. Then you decide.

Free. Ten business days. We run one of your vendor's scrapes, hand you the findings to verify yourself, and walk you through them. Worst case: you learn your current setup is fine, and you keep the report anyway.