Rebuilding Nextmol’s CRM into a system that qualifies without one person’s memory

The organisation

Nextmol sells computational chemistry to R&D teams in the chemical industry, both as a service and as a software platform. The engagements are long and technical. Around twelve months pass from first contact to signature, and every proposal is written from scratch by the company’s scientific director. The commercial team is small, and the CEO carries most of the relationships.

The situation

Tools had been added as they were needed, never designed as a system. HubSpot held the history of the work but was consulted rather than run on, and the record of any one deal sat across five places: the CRM, the meeting platform, the document store, a personal spreadsheet and a paper agenda. Only the CEO could reconstruct where something stood.

The real problem

The obvious brief was a CRM cleanup. That was never the difficult part.

The pipeline had been built to record what had already happened rather than to decide what happened next. Deals were opened on prospects who had never had a commercial conversation, so deal count, win rate and deal age each measured something other than what they claimed to.

What was broken

Two pipelines for one sales process. The funnel was split by how the work would be delivered, though the selling was identical either way. Reporting was split with it.

Leads and opportunities in the same place. Deals were opened on prospects who had never had a commercial conversation.

Nothing to forecast or attribute with. Most deals carried no value, and most contacts no record of where they came from. Nothing flagged a deal gone quiet.

The architectural decisions

One pipeline, with delivery model as a property. The two pipelines recorded how the work would be done, not how it was sold. Making that a field preserved the distinction without fragmenting the forecast.

A separate object for everything not yet an opportunity. A deal is created at one moment only: qualification passed, first meeting held, capability match confirmed. Everything earlier, and everything gone cold, sits in a lead pipeline. That rule is what makes win rate and deal age readable.

Scoring on the lead, so qualification can be handed on. It had lived entirely in one person’s judgement, so it could not be taught or delegated. Company fit and engagement score separately rather than blended: an account that is wrong and quiet and one that is right and quiet need opposite actions. A third score orders open deals, so proposal writing queues by evidence.

Automation that watches for silence. Nothing in the old system could say which deals had gone quiet. A deal or lead with no activity for forty-five days now raises a task and an alert for its owner, and the threshold sits in one place so it can be tuned as history accumulates.

What we built

A six-stage pipeline with entry criteria in each stage and required fields gating movement. A lead pipeline, eleven deal properties, a nine-channel taxonomy wired into every web form. Twelve workflows covering lifecycle, attribution, task creation and stall detection. Three scores and two dashboards.

How the work ran

Two audits came first, one of the HubSpot configuration and one of how the team actually sells, both delivered in June with every figure measured in the portal rather than estimated. The architecture was proposed and approved before anything live was touched. The foundation went in at the end of July, built alongside the existing pipelines so nothing in use was disturbed.

The migration ran in August against an approved map, deal by deal, with the early-stage records routed into the new lead pipeline rather than deleted. Documentation was written as the system was built, then reviewed line by line against the live portal before handover. That review is what caught a set of claims that no longer matched what had been built, and corrected them before the client read them rather than after.

What changed

The open pipeline holds only live opportunities, and nothing sits quiet without someone being told. Qualification runs on a model rather than on one person’s memory: every account carries a fit and an engagement score, and open deals carry a priority score that orders the proposal queue. What the model cannot supply is the part only the team knows, and the build established exactly which fields those are.

Why this matters

There are no campaign metrics in this case. The system went live in the closing weeks of the engagement, and the honest position is that the numbers which would prove it out do not exist yet.

What exists is a commercial system whose rules are written down. A deal means one thing. Qualification that lived in one person’s head now has a model behind it. Nothing goes quiet without someone being told. Those are the conditions under which metrics become worth measuring, and none of them were true before.

Who this is for

Companies where the CRM holds the history but cannot say what is live. Where the pipeline has become a list of everything anyone ever spoke to, so nobody trusts the forecast. Where qualification works because one person holds it in their head, and everyone knows that is a risk but no one has had a quarter to fix it.

The problem is rarely the tool. It is that nobody ever decided what the pipeline is for.