There are plenty of tools that generate lead lists. Most of them offer large volumes of contact data — email addresses, job titles, company names — often scraped from public sources or pulled from third-party databases.
But if you’re looking for specificity — like companies with 50–200 employees that operate in both Ireland and the Netherlands, manage physical locations with wellness facilities, have recently submitted public tenders, rent from a larger real estate provider — these platforms quickly reach their limits.
In cases like this, building a lead list becomes a data collection problem. It requires identifying relevant sources, defining filtering criteria, and writing logic to connect the dots across different formats and structures. Once done, it becomes possible to semi-automate the process and start using tools to regenerate or update the list as needed. That requires an approach where initial trying & testing strategies are all mapped (or just discussed) and turned into a general format. Compare it to scientific research: first do experiments, then link a general theorem to explain (and be able to replicate) results.
Having well-structured, high-context lead data also makes testing messaging and segmentation more straightforward. Instead of A/B testing on job title alone, one can work with industry-specific variables, really adding any type of niche filter imaginable. Generic platforms are great for getting lists of contacts and searching for standard attributes like industry, job title.
Despite advances in scraping tools and language models, this kind of work still requires significant human input. Some common friction points:
Most scraping use cases start as manual workflows. Someone has to go through the pages, determine where useful information sits, and establish a consistent way to extract it. This step is often skipped, but without it, attempts at automation tend to break or produce unreliable data. There is no template — someone has to define it.
Data pulled from multiple sources rarely aligns. Formats differ, naming conventions vary, and data may be incomplete or inconsistent. Matching entities, standardising fields, and dealing with missing or ambiguous entries requires both rules and judgment. This can’t be fully outsourced to generic AI models (yet).
Some websites explicitly block automated access. Others don’t, but they present content in ways that make scraping fragile — with dynamic loading, inconsistent structure, or content hidden behind interactions. In these cases, using standard scraping tools can still work for simple cases, but more robust setups require conditional logic, retries, and alternative data paths.
Web scraping can still be useful for building specific lead lists — especially when off-the-shelf databases fall short. But the useful part of the process is rarely the scraping itself. It’s in designing the system: deciding what data matters, figuring out how to get it, and combining sources in a way that produces something usable.
If you’re currently experimenting with this kind of process or have found ways that work better, we’re curious. Feel free to share your approach or reach out at stephanie@pearstop.com.
Discover how Pearstop revolutionizes research for sales, procurement and due diligence. Access internal data swiftly, interact with large data rooms reliably, automatically label, categorise and price ledger data.
We believe in instant answers from your data. No hours spent on tedious, repetitive work—what you need, when you need it.