Product Engineering

How to Build a Custom CRM, and What Off-the-Shelf CRMs Miss

Home Blog Product Engineering

Product Engineering

Salesforce, HubSpot, Zoho and Pipedrive serve millions of teams well. But we regularly meet businesses whose sales process doesn't fit those tools, and who spend more time working around their CRM than with it. Here's when a custom CRM makes sense and how to build one sensibly.

Signs you've outgrown a generic CRM

  • Your pipeline isn't a simple funnel. Think multi-party deals, tenders, long project-based sales, or approvals that span several departments.
  • Critical data lives in spreadsheets next to the CRM because the CRM can't model it.
  • Per-seat costs are high for a large team that only needs a few features.
  • Integrations are fragile. Data has to move between the CRM and your ERP, billing or operations systems through brittle connectors.
  • Industry-specific needs, such as property, education, healthcare or logistics, require fields and workflows that generic tools treat as afterthoughts.

The core features (start here)

  1. Contacts and organisations, with deduplication.
  2. Leads and deals with your actual pipeline stages.
  3. Activities: calls, meetings, notes and tasks, linked to contacts and deals.
  4. Email integration, so conversations are logged automatically.
  5. Roles and permissions, so reps see their own records and managers see their team's.
  6. Dashboards: pipeline value, conversion by stage, activity per rep, forecast.

That's enough for a first release. Resist building marketing automation, a quoting engine and a customer portal all at once.

A sensible data model

Organisation ─┬─ Contact ── Activity
              └─ Deal ──┬── Activity
                        ├── StageHistory   (who moved it, when)
                        └── Product / LineItem

Two details that pay off later: keep a stage history table (it powers conversion and cycle-time reports), and use soft deletes so nothing important disappears by accident.

Technology choices

For most custom CRMs we use Django on the backend. Its ORM, permissions and admin interface cover a large part of a CRM's back office. A React frontend handles the fast, interactive screens sales teams live in, such as kanban pipelines, search and inline editing. PostgreSQL with full-text search handles finding contacts quickly, and Celery runs background jobs like email syncing and reminders.

Integrations that matter

  • Email and calendar (Google Workspace, Microsoft 365),
  • web forms and landing pages that create leads,
  • telephony and WhatsApp Business for call and message logging,
  • accounting or ERP for invoices and payment status,
  • a clean REST API, so future tools can connect easily.

A phased plan

  1. Weeks 1–2: workshops to map the real sales process and the data it needs.
  2. Weeks 3–8: core CRM (contacts, deals, activities, permissions) used by a pilot team.
  3. Weeks 9–12: email integration, dashboards, data migration from the old system.
  4. After launch: add automation and integrations based on what users actually ask for.

Don't forget data migration

Moving years of contacts and deals out of spreadsheets or an old CRM is often the most underestimated task. Plan for cleaning, deduplication and a test migration before go-live.

Wondering whether a custom CRM is worth it for your team? Tell us about your sales process and we'll give you an honest recommendation.