Field manual
This page is the whole system with no gloss: what runs, in what order, what the machine decides, and what waits for you. Ten steps, four phases, one memory. If you only read one page before a demo, read this one.
PHASE 01 — MAP
Everything starts with a brief in plain language, not a filter panel. The engine does the translation from intent to research plan.
You describe who you sell to the way you'd tell a colleague: "every mid-market 3PL in the Benelux still running a legacy WMS." The engine converts that into a research plan — which registries, directories, review platforms, communities, and career pages to read, and what counts as proof of fit. You see the plan before it runs.
Worker agents read those sources in parallel and turn what the market publishes about itself into structured accounts — company, domain, size, geography, evidence links. Duplicates collapse by domain; ambiguous entities are held back rather than guessed at.
Every account is scored against your ICP criteria — and it is not a black-box number. "Matches 7 of 8" means seven criteria, each citing the evidence it was scored on. Kill a bad account and the engine learns which criterion misled it.
Output: your total addressable universe as a living map you own — filterable, exportable, and never rented back to you.
PHASE 02 — ENRICH
A map tells you who exists. The dossier tells you who decides, what they run, and when the window opens.
Each high-fit account gets a dossier: the people who actually decide, their roles and seniority, the tech stack in use, funding history, hiring activity, and timing evidence — every claim with its source attached.
Contact data runs through a waterfall of licensed providers: each source is checked against the next until confidence is high enough to act on, and email addresses are verified before they ever enter a sequence. Roughly a third of B2B contact data decays every year — the waterfall is why that decay dies in the pipeline instead of in your sender reputation.
Buying signals attach to accounts the moment they surface — and age out when they go cold. The map is a delta stream, not a CSV export: what changed this week, ranked by how much it matters to you.
Output: dossiers with verified people, cited evidence, and live timing — refreshed by agents, not by an annual data purchase.
PHASE 03 — ENGAGE
Personalization isn't a first-name token. It's writing from what you verifiably know, at the moment it's true.
Openers and follow-ups are written from the dossier — the signal that justified the touch, the stack they run, the role of the person reading. Every draft can point at the evidence behind every sentence.
Nothing sends itself. Drafts land in a review queue; you approve, edit, or kill. Your edits are the steering wheel — the system learns your voice from what you change, not from a tone dropdown.
Approved touches run across email and LinkedIn with deliverability guarded — verified addresses only, volume ramped, domains warmed. Replies are classified; the moment there's real interest, a human takes the conversation.
Output: conversations opened on evidence, not spray — with your reply rate protected by the same verification that built the map.
PHASE 04 — CLOSE
The handoff is where most stacks drop context. Here there is no handoff — the CRM was in the room the whole time.
A positive reply becomes a deal with the full dossier attached — every signal, every touch, every reason. Next steps are proposed as tasks you confirm; the account memory keeps compounding through the sales cycle and after the win. That memory is why the second deal in a segment closes faster than the first.
Output: pipeline where every deal carries its own intelligence file — and a system that gets sharper with every closed-won.
Reference
Ten classes of buying signal the engine watches for, why each one predicts a purchase, and how long the window typically stays open. Signals attach to accounts automatically and age out when they go cold — timing is data, not intuition.
| Signal | Why it predicts buying | Typical window |
|---|---|---|
| Funding round closed | New budget plus a mandate to grow — tooling and vendor decisions follow within the quarter. | 4–12 weeks |
| Leadership hire (VP / C-level) | New owners re-evaluate the stack they inherited. The best moment to be found is before their plan is set. | First 90 days |
| Hiring spike in one function | The pain is present-tense — headcount is the expensive fix your product replaces or multiplies. | While reqs are open |
| Tech switch detected | The stack is already in motion; the switching cost objection is being paid anyway. | Weeks |
| Negative reviews of incumbent | Public frustration with a competitor is an account mid-decision, telling you its evaluation criteria. | Days–weeks |
| New country entity registered | Expansion creates operational gaps overnight — compliance, logistics, hiring, tooling. | 1–2 quarters |
| Public tender / RFP | Explicit, budgeted, deadline-bound demand — the rarest and clearest signal there is. | Fixed deadline |
| Champion changed jobs | People who bought from you once buy again in new seats — if you notice the move. | First 90 days |
| Pricing or positioning rewrite | A strategy shift in public: new segments, new budgets, new gaps a vendor can fill. | Weeks |
| Physical expansion (sites, warehouses) | New locations force process and tooling decisions that were postponable before. | 1–2 quarters |
Why verification is the moat
Any single contact-data source is meaningfully wrong: people change jobs, companies rename, inboxes close. Industry studies put B2B data decay around 30% a year — which means a list bought in January is a coin flip by summer.
Hyves treats no source as truth. Each fact is checked across a waterfall of licensed providers and public evidence until confidence clears the bar; email addresses are additionally verified live before they can enter a sequence. What fails verification never reaches your team — and never burns your domain.
Verification also runs backwards: dossiers carry a freshness clock, and facts past their reliable lifetime get re-checked or retired. The result is a number most teams have never seen on their own stack — a bounce rate low enough that deliverability stops being a project.
Rules of engagement
APPROVAL
Every outward-facing action — a sent email, a new sequence, a task — passes a human approval gate. This is architecture, not a settings toggle.
LAWFUL BASIS
Public business data and licensed B2B datasets under legitimate interest, suppression and deletion on request, no scraped private profiles, no gray-market databases.
ISOLATION
Tenant separation is enforced at the database layer. Your map, your dossiers, and what the engine learns from your edits belong to you alone.
JURISDICTION
EU infrastructure, EU legal home. Your data doesn't take trips it doesn't need to take.
Bring one sentence about who you sell to. You'll have a scored map before the first call ends.