EXPERIENCE, NOT THEORY.
Prospect research:
the limitations of platforms,
dirty data, and what changes
when AI comes into play
B2B prospect research in 2026 relies on multiple tools, each designed to serve a different purpose: commercial databases such as Kompass, Cerved, Bureau van Dijk and Orbis for verified company data; LinkedIn Sales Navigator for accurately mapping roles and decision-makers; sales engagement platforms such as Apollo, Lusha and ZoomInfo for direct contact information; and, in our case, an AI-powered sales intelligence solution for semantic profiling and contextual analysis. Each platform has its own strengths and limitations, and the real differentiator lies in orchestrating multiple sources, diversified by geography and industry, rather than relying on a single platform. AI accelerates profiling and analysis, but it stops before the point of contact: the relationship-building phase remains human-centric.
Prospect research today relies on four types of tools that serve different purposes and, when used correctly, work together to create something that none of them could achieve alone. Certified commercial databases provide verified company information. LinkedIn Sales Navigator offers granular insights into company roles. Sales engagement platforms such as Apollo, Lusha and ZoomInfo provide access to direct contacts at scale. AI platforms enable analysis and contextualisation at a speed that manual work cannot match.
The challenge is that each of these tools has its own specific biases which, if ignored, can lead to ineffective prospect lists. The most common mistake we see every day is treating them as interchangeable sources, choosing one and building an entire strategy around it.
Prospect research platforms all share the same characteristic: they look impressive when you buy them, but become an obstacle course once you start using them. It is worth understanding why before investing time and budget in tools that promise perfect lists but ultimately deliver noise.
Certified commercial databases: verified data as the starting point.
Certified commercial databases such as Kompass, Cerved, Bureau van Dijk and Orbis are the most reliable source for company data. They cover different geographies and levels of data depth, and none of them is sufficient on its own.
They aggregate verified data: revenue, number of employees, industry, corporate structure, organisational charts and ownership stakes. These are declared and verified information, supported by an underlying methodology. When building a list from these platforms, the filters applied correspond to real-world data.
Each platform has a different area of specialisation. Cerved is particularly strong in the Italian business landscape, with granular data on domestic SMEs. Kompass offers strong European coverage and a detailed sector classification. Bureau van Dijk, through Orbis, provides international coverage with in-depth insights into multinational groups and complex corporate structures.
The question is not “which one is the best?”, but rather “which one is needed right now?”. Prospect research on Lombardy-based manufacturing SMEs requires different tools compared to research on industrial groups with a presence in Asian markets. The limitation of these platforms, even the best ones, is that they stop at the company profile level. They tell you what a company is, but not what stage it is in, what is happening within it, or who is actually making the decisions. For that, additional sources are needed.
Sales Navigator excels at job title searches but is affected by four structural distortions: profile optimisation, delays in reporting job changes, unverified company data, and uneven geographic coverage.
Sales Navigator is probably the most widely used tool for B2B prospect research.
It is also the tool most often misunderstood, not because it is a poor tool, but because it is used with expectations that do not match what it actually does.
Its strength is real and specific: job title search. No other platform offers comparable coverage when it comes to mapping company roles. If you search for “Export Manager” in manufacturing companies in Northern Italy with 50–200 employees, Sales Navigator will return a list with a level of granularity that no traditional database could match.
The problem begins when we forget that this data is self-reported.
Every LinkedIn profile, whether a personal or company profile, is completed directly by the user. There is no verification process and no external validation. This creates four distortions that anyone conducting prospect research needs to understand.
Let’s analyse your sales research process?
Apollo, Lusha, ZoomInfo: the promise of direct contact data and its limitations.
Sales engagement platforms such as Apollo, Lusha and ZoomInfo provide access to millions of direct contacts, including business emails, phone numbers and job roles… but they raise specific questions around data quality, regulatory compliance and responsible use.
In recent years, a new category of tools has changed the way many sales teams approach prospect research. Apollo, Lusha, ZoomInfo, as well as Cognism, Lead411 and Seamless.ai, promise something that traditional platforms do not offer: direct contacts at scale. Business emails, phone numbers, job roles, all within a single interface, with instant filters and the ability to export lists in just a few clicks.
The potential is real. For outbound sales, having access to a direct mobile number rather than a company switchboard can make the difference between starting a conversation and receiving no response. Sales engagement platforms have undoubtedly accelerated the speed at which sales teams can build a pipeline.
However, beneath the surface there are three limitations that need to be understood before building a strategy around these tools.
Apollo, Lusha, ZoomInfo and similar platforms have real value when used as one source among many, with data verified before use and outreach processes designed to stand out from the background noise. They become a problem when they are treated as a complete solution to the prospect research challenge.
AI platforms for prospect research.
ChatGPT, Gemini, Copilot and Claude are useful tools for accelerating prospect analysis and qualification, each with different capabilities in a rapidly evolving field.
At this stage, Claude handles complex prompts and long contexts with greater consistency compared to other tools, but the sector is evolving rapidly and the gap is narrowing every quarter. The key point here, as well, is not which tool you choose, but understanding what you can make it do, and where it may lead you in the wrong direction.
Building a prospect list: why getting the fundamentals wrong comes at a high cost.
A prospect list built by applying standard filters on a platform reflects the quality of the platform’s data, not the quality of the actual target. The result is commercial noise that wastes contacts, time and reputation.
Building a prospect list from scratch seems simple. You choose a platform, apply filters, industry, company size, geography, and get your contacts. What you obtain reflects the quality of the platform’s data, not the quality of your actual target. Companies that match the formal criteria but not the real profile. Misclassified industries. Wrong decision-makers. Sending outreach based on a list built this way means creating noise. And in sales, noise has a measurable cost: it burns contacts, wastes time, and damages the reputation of the person generating it.
For years, the solution was manual. Someone opening one website at a time, verifying information, cross-referencing data and making decisions. Slow, expensive, not scalable, and in some markets, practically impossible. In many Asian countries, company websites may be inaccessible due to firewalls, public information can be limited, and traditional profiling reaches a dead end before it even begins.
Manual work also has a limitation that is often underestimated: it is subjective. Two people profiling the same database using the same criteria can reach different results because human judgement introduces variables that cannot be fully controlled. The quality of the profiling depends on experience, focus and even the time of day. That isn’t a system.
How our sales intelligence works: from semantics to the hook.
Mallei’s sales intelligence is based on semantic search: our data analysts describe the ideal customer profile in natural language, the system interprets the description and provides a reasoned assessment of each potential company. It identifies decision-makers, relevant news and contextual signals that can support the initial outreach.
What we have built starts from a simple idea: profiling should be an exercise in understanding, not filtering. The difference is substantial.
A filter tells you whether a company meets certain formal criteria: industry, size, geography. Understanding tells you whether that company has the right characteristics, timing and context to become a real prospect.
Our method starts with the orchestration of sources. The platforms we have described so far, certified commercial databases, Sales Navigator, sales engagement platforms and general-purpose AI tools, are not alternatives. They are complementary layers that need to be integrated and diversified based on the specific context of each research project. Researching food and beverage SMEs in Southern Italy requires a different combination of sources compared to researching industrial groups with a presence in North America or Asia. Geography and industry determine which sources are primary, which are secondary, and which should be excluded because they are less reliable in that specific context.
The core of the system is semantic search. Instead of applying rigid filters based on predefined categories, our data analysts describe the ideal customer profile in natural language: who they are, what they do, what stage they are in, what challenges they face and what objectives they are pursuing. We build a prompt that describes a real-world scenario: “a company that produces industrial components, has an internal sales structure, is looking to expand into international markets, likely with an export manager or sales director directly managing negotiations and seeking companies with a strong focus on technological innovation.”
Semantic search interprets this description and compares it with publicly available company information, such as websites, press releases, news, job postings and communication style. The system returns a reasoned assessment: why this company is a good fit, what makes it relevant, which specific signals support this conclusion, and why another company is not a match and what does not align.
This reasoning is valuable for two reasons. The first is operational: it allows the outreach team to immediately understand the context without having to conduct additional research. The second is systemic: over time, by analysing these assessments, our data analysts understand which qualification criteria genuinely work and which ones generate false positives. The system therefore improves and becomes more accurate through continuous use.
News, decision-makers, LinkedIn: the hook that changes everything.
Once target companies have been identified, our system searches for recent news that highlights specific moments in a company’s journey: management changes, acquisitions, new markets, and significant hires. It also analyses the public communication of decision-makers on LinkedIn to understand how they think and which topics they focus on.
Anagraphic profiling is only the first level. What truly makes the difference during the outreach phase is understanding what is happening within that company at that specific moment and how the person the team is about to speak with thinks and operates.
Once our system has identified target companies, it searches for recent news: new office openings, acquisitions, management changes, expansion into new markets, significant hires and trade fair participation. These are signals that indicate a specific moment in the company’s journey, and timing matters enormously in sales. A company that has just appointed a new Sales Director is likely going through a period of transformation. A company hiring export roles is building something new. A company that has just closed a funding round is entering an acceleration phase. These contexts completely change how a conversation should be approached.
Then there are the decision-makers. The system identifies relevant figures within the organisation based on their role, professional background, time in their current position and how they communicate publicly. This is where LinkedIn, when used in this way, becomes a completely different tool compared to simple contact research.
Analysing a decision-maker’s communication on LinkedIn means understanding how they think, what matters to them and which language resonates with them. What they publish, what they comment on, what they engage with and which conversations they participate in. A manager who consistently posts about international expansion and new markets has a specific mindset. A manager who engages with content about sales team management is likely focused on particular challenges. These are valuable entry points, and a contact team that understands them before the first interaction starts from a fundamentally different position compared to someone sending a generic message.
Where AI stops (and why it is right that it stops there)
AI accelerates profiling, analysis and contextualisation. It stops before the contact phase, which remains human-centric: in B2B sales, people immediately recognise when they are interacting with an automated process, and that recognition closes the door.
All of this work, semantic profiling, news research, decision-maker analysis… stops before the contact phase. And that is a choice, not a limitation.
People recognise when they are interacting with an automated process, often in subtle ways: a feeling that something does not quite align in the tone, timing or relevance of the response. In B2B sales, where trust is the foundation of any commercial relationship, that recognition closes the door before it even opens.
AI analyses, cleans, qualifies and contextualises data at a speed and scale that no human team could achieve. Building a real commercial relationship requires presence, active listening, real-time adaptation and the ability to interpret what is left unsaid. These are skills that remain human.
A contact team that approaches a prospect with accurate profiling, knowledge of the company’s latest developments, a genuine understanding of the decision-maker’s profile and a clear contextual reason why the conversation makes sense is respecting the other person’s time. And in the long term, that is what builds a company’s commercial reputation.
Frequently asked questions about B2B prospect research.
There is no single best platform overall. Certified commercial databases such as Kompass, Cerved, Bureau van Dijk and Orbis are the most reliable sources for company data, each with different geographic specialisations. LinkedIn Sales Navigator is essential for job title searches and analysing decision-makers’ communication. Sales engagement platforms such as Apollo, Lusha and ZoomInfo provide access to direct contacts at scale, with limitations around data quality and regulatory compliance. AI platforms accelerate analysis and semantic profiling. An effective prospect research strategy combines multiple sources, understanding the specific limitations of each and diversifying based on geography and industry.
Only partially. Company role data is generally accurate at a category level, but it is affected by four structural distortions: profile optimisation (inflated or generic job titles), delays in reporting job changes (weeks or months), unverified company data (self-reported company size and industry), and uneven geographic coverage (lower penetration in some international markets). It should always be cross-referenced with other sources.
They have real value, but they need to be used with awareness. Contact quality varies because the data comes from heterogeneous sources, scraping, user contributions and algorithmic inferences, and actual bounce rates are often higher than those reported by vendors. In Italy, there is also the limitation of the Public Register of Objections: people registered with the RPO have exercised their right not to receive unsolicited commercial communications, and these international platforms do not have native mechanisms to filter out those contacts. They should be treated as one source among many, not as a complete solution.
AI is effective during the profiling and analysis phases: semantic search to identify target companies, gathering relevant news, and analysing decision-makers’ public communication. It stops before the contact phase, which remains human-centric. The value of AI in sales lies in accelerating preparation, not replacing relationships.
A qualified prospect list combines three levels: alignment with formal criteria (industry, size, geography), alignment with the ideal customer profile (business model, company stage, organisational complexity), and the presence of contextual signals (recent news, intent signals, and decision-makers’ public communication). A list that only meets the first level creates noise. In some cases, completing the profiling process is only possible through the first direct interaction with the prospect, because there is information that cannot be identified through desk research alone.
Manual profiling is slow, expensive, not scalable and subjective. Two people profiling the same database using the same criteria can reach different results. In markets with information barriers, such as many Asian countries, but not only, traditional profiling is effectively impossible. AI-based sales intelligence enables thousands of companies to be processed in parallel while maintaining consistency in evaluation criteria.
