How we rebuilt a Series A SaaS's CRM in eleven days

A founder reached out in early March with a problem that will sound familiar: their CRM was technically HubSpot, but the actual sales motion lived across six spreadsheets, two Notion pages, and one heroic head of sales who held the whole thing in his head.

The brief was simple, the work was not: one source of truth, in eleven days, without breaking the active pipeline. Here is exactly what we did.

Day 1 — The audit

The first day is never about building. It is about looking. We started with a two-hour pipeline walk-through: every deal currently in flight, the source of each, the last touch, and where the next action lived. We mapped this onto a single ledger and named everything that was a duplicate, ambiguous, or stale.

The output was a list of 47 deals, 11 of which existed in more than one place. Three were already closed but had not been marked as such. Two had been forgotten.

Days 2–3 — Schema

The fastest way to break a CRM rebuild is to start with the integration. We started with the schema instead. We sat with the head of sales and asked the question that should be asked first and rarely is: what decisions does this CRM need to support?

The answer was three: which deals to push this week, which to deprioritise, and which need executive cover. From those three decisions we worked backwards into fields. We ended up with 12 custom properties, down from the 38 that had accumulated over two years.

Days 4–6 — Migration

The migration itself was the easy part. We staged it: import all closed deals first (low risk, high volume), then open deals in priority order. Every deal got a single validator before going live: a real person looking at the import and confirming the next action was correct.

This is the part where most agencies skip a step. They run the import and call it done. We caught 7 import errors that would have caused real revenue impact — including one deal where the contact email belonged to the wrong company.

Days 7–9 — Automation

Once the data was clean, we built four automations:

• New inbound lead → enrich via Clearbit → score → assign → Slack alert to the right rep
• Deal stuck in stage X for > 14 days → flag for manager review
• Closed-won → trigger onboarding handoff to client success
• Closed-lost → trigger a 90-day reactivation sequence

These four automations now handle work that previously occupied the head of sales for about an hour a day.

Days 10–11 — Handover

The last two days are documentation and training. SOPs for every field, every automation, every weekly ritual. A 90-minute training session with the sales team. A 30-minute briefing with the founder.

What we measured

Three weeks after handover, the founder shared the numbers: pipeline forecast accuracy up 41 percent (from 52 percent to 73 percent), deal cycle time down 18 percent on average, and zero deals lost to "I forgot to follow up" in the trailing 30 days.

What we got wrong

One thing. We underestimated the volume of historical contact data that needed deduplication. We allocated a half-day for it; it took a day and a half. We extended the engagement at no cost — the Promise.

The lesson: every CRM rebuild has one task that takes 3× longer than estimated. Build the buffer into the engagement, not the invoice.

If you are sitting with a CRM that feels like the spreadsheets you used to have, this is the shape of the work. Two weeks. A schema-first approach. A clean handover. The compounding starts immediately.