A B2B sales engine for an SME-scale operation

What it is

A complete B2B sales engine, built for an SME-scale operation. Enterprise sales teams run on process: a pipeline with stages, clocks on every deal, a system that remembers every conversation. A lean commercial operation almost never gets that. It runs on memory and good intentions, and the good prospects slip while nobody is watching. This system takes the enterprise discipline and rebuilds it so a small team, or a single operator, can actually run it. Enterprise stages, clocks, and drop-ball detection, run by a lean team, with every message still written and sent by a human.

The problem

Two failures quietly kill a lean pipeline. The first is drift: a warm prospect goes cold because nobody put a clock on the follow-up, and past a few dozen active relationships, memory stops being enough. The second is noise: too many possible prospects, too many sources to watch, so the ones worth real effort get buried under the merely available. Same root cause. Without a forcing function, the highest-value move is always the one that gets postponed.

What it is built from

Five layers, joined into one system.

The discovery layer pulls prospects from several sources at once and turns raw signal into a clean, ranked list. It resolves entities, so the same account arriving from two sources is recognized as one, and it holds a dedup ledger so nobody already worked is surfaced again. Every candidate is scored for fit against a defined ideal-customer profile. What lands is a short, ranked list worth a real touch, not a raw dump to triage.

The pipeline layer treats every relationship as a position in a stage machine, from first identified, through connected, in conversation, meeting booked, to outcome. Each stage carries its own SLA clock. The layer watches those clocks, flags a stall the moment a stage runs long, and re-ranks by priority so the account about to go cold rises to the top on its own.

The memory layer logs every interaction the same session it happens, into a structured relationship record that outlives any one person's recall. The history is append-only: nothing is quietly overwritten, so the full arc of a relationship is always there to read. The system carries the working memory, which frees attention for the selling.

The quality layer is a set of mechanical pre-send gates. Before any message goes out, the draft is checked against a personalization floor, linted for banned phrases and format, and run through consistency checks. The gate blocks bad output and nothing more. The human writes every message and sends every message by hand.

The reporting layer reads the whole pipeline on a weekly cadence and hands back funnel health, aging analysis on every stalled thread, and conversion metrics stage to stage. The dropped balls surface as a list: the warm windows about to close, who is due now, what has gone red.

How it holds together

One rule sits under all five layers: every prospect, status change, and park is written to a single source of record the same session it happens, never deferred. A decision that lives only in someone's memory does not exist. An integrity check runs at open and close and refuses to pass if the record contradicts itself. The layers are only as good as the record they share, so the record is guarded hardest of all.

The point of the case

Discovery scores fit and refuses to resurface an account already worked. The pipeline puts a clock on every stage and lifts the one about to go cold. Memory is append-only. A gate can block a draft; it cannot send one. The weekly report is a list of dropped balls, not a dashboard nobody opens.

One of a set of engagement and systems case studies. If a problem like this is on your desk, start with a conversation.