Extraction
- Public websites, respecting their terms and without overloading them
- PDFs, invoices and scanned documents with OCR
- Excel, CSV and exports from other software
- Official APIs and open data portals, when they exist
Nexura Digital service
Data extraction and cleaning so you can work with reliable information
We collect data from public websites, documents and spreadsheets, clean it and deliver it in a stable format, with a repeatable process.
We extract data from public websites, PDFs, Excel files and other sources, normalize it, remove duplicates and deliver it ready for your CRM, your store or your analysis. It's for companies with messy lists, catalogs or records that want the work done legally and documented. You can start with a 7-day trial on one specific dataset.
What we do with your data
Cleaning and normalization
- Dates, phone numbers and countries in standard formats such as ISO 8601 or E.164
- Duplicates caught even when they're written differently
- Split fields and unified units
- Doubtful records flagged for review, not deleted
Loading and repeating
We deliver in a spreadsheet, a CSV file or straight into your database or CRM. If the source keeps updating, we leave an n8n workflow or a scheduled script.
Legal and orderly: the limits we respect
This is technical assistance, not legal advice: if a project raises legal questions, your lawyer should assess it before we start.
Terms of use and robots.txt
We read each site's terms of use and its robots.txt, the standard (RFC 9309) a site uses to say what it doesn't want crawled. We don't bypass logins, paywalls or CAPTCHAs, and we limit request rates.
Personal data and GDPR
Public doesn't mean usable for any purpose: the GDPR also protects personal data that has been published. We collect only the fields needed and never build lists for unsolicited commercial messages.
How the work runs
Five steps
- We define what you'll use the data for and which fields you need
- We review sources, terms and, if there's personal data, how it fits the GDPR
- We test extraction on a sample and measure its quality
- We agree with you on cleaning and deduplication rules
- We process the full dataset and deliver it with a quality report
Projects that fit and projects that don't
Good fits
Supplier catalogs in PDF that need to go into your store, customer databases full of duplicates before a CRM migration, public prices in a market you want to track, or open data for a study.
Not a fit
Collecting emails or phone numbers for cold campaigns, copying protected content to republish it, or extracting data from sites that expressly forbid it. Not if you need a legal opinion either: that's a lawyer's job.
What each delivery includes
Data and documentation
- A clean dataset in the agreed format
- A data dictionary with each field, its origin and its format
- A quality report: records processed, duplicates merged and doubtful cases
- A documented script or n8n workflow to repeat the process
- A log of sources and the date of each extraction
If the process repeats
The workflow runs on a server you pay for, n8n Cloud or your own VPS, with its cost included in the quote, and it stays in your name.
7-day trial with a real dataset
From sample to result
A realistic option is cleaning and deduplicating your customer base or a catalog, or automating extraction from one specific public source. We start from a sample, agree on the rules and deliver the result with its quality report. Before we touch a single record, we agree in writing on scope, dates and terms.
FAQ
Frequently asked questions
Is it legal to extract data from a public website?
It depends on which data, which site and what for. The site's terms of use, the rights over its content and, if personal data is involved, the GDPR all play a part. We apply cautious technical criteria and point out the risks we see, but we don't give legal advice: in doubtful cases, decide with your lawyer.
What happens if the website changes its design?
Website extractions break when the page structure changes. That's why we prefer official APIs or downloads when they exist, add checks that alert you if data arrives empty or in odd formats, and document how to adjust the process. If you need it, maintenance is agreed separately.
Can you extract data from scanned PDFs?
Yes, with OCR, and for complex documents we combine it with AI models that pull out specific fields such as amounts, dates or references. Quality depends on the scan, so we validate on a sample, flag low-confidence fields and check totals when the document includes them.
How do you detect duplicates?
We combine exact rules, such as the same tax ID or email, with fuzzy matching on names and addresses that accounts for abbreviations, accents or typos. Clear cases merge automatically and doubtful ones go to a list for your team to decide. Our request automation demo shows the idea: it's an in-browser simulation, not a real workflow.
Contact
Which data costs you the most time today?
Send us a sample or describe the source and its use. We'll tell you whether it's feasible, with what limits and what the 7-day trial would look like.
Tools
Find out where to start in one minute.
Tools
What do you want to improve?
Pick a goal and we'll suggest the services that fit best.