Most teams ask us to automate their CRM. What they usually need first is an honest look at what the CRM is already doing to their pipeline. Automation on top of a broken data model just moves bad records around faster. So every engagement starts the same way: a two-day audit, twelve checks, one prioritised fix list.
Start with the pipeline, not the tooling
A pipeline should describe what the buyer has done, not what your rep intends to do. Stages named 'Following up' or 'Nurture' are activity labels — they tell you nothing about whether the deal moved, so forecasting collapses and automation has nothing reliable to trigger on.
We rewrite every stage as an observable buyer event: booked a call, attended the call, received a proposal, agreed commercials. Once stages are events, exit criteria become obvious and a deal can no longer sit in one column for six weeks without anyone noticing.
- Every stage has a written entry and exit condition
- No more than seven stages in a primary sales pipeline
- Lost reasons are a fixed list, not a free-text box
- One pipeline per motion — new business and onboarding stay separate
Trace where a lead actually enters
We map every genuine entry point: website forms, calendar bookings, paid landing pages, inbound calls, live chat, referral intros, and the spreadsheet someone still keeps for events. Each one gets tested with a real submission that we follow all the way through to a CRM record.
This is where most leaks show up. Forms writing to a mailbox nobody owns. Calendar bookings that create a contact but no deal. Ad campaigns landing without source data, so paid spend can never be tied to revenue. You cannot automate a route you have not proven exists.
Audit the data before you trust it
Duplicate contacts, inconsistent phone formats, missing owner fields and half-populated custom properties will each break an automation in a different way. We measure fill rates on the fields that workflows depend on and treat anything under 90% as a blocker rather than a nice-to-have.
The fix is usually structural: fewer required fields, stricter input validation at the form layer, deduplication rules on email plus phone, and a single owner field that every downstream rule reads from.
- Duplicate rate across contacts and companies
- Fill rate on source, owner, lifecycle stage and deal value
- Consistency of phone, country and currency formats
- Records with no activity in the last 90 days
Time the hand-offs
Speed-to-lead is the single metric most correlated with close rate in the accounts we work on, and it is almost always worse than the team believes. We pull the timestamp gap between enquiry creation and first genuine human contact, then break it down by source and by hour of day.
The pattern repeats: fast on weekday mornings, hours slower in the evening, effectively dead at weekends. That is not a discipline problem, it is a routing problem — and it is the first thing worth automating.
What comes out of the audit
You get a twelve-point scorecard, a ranked list of fixes with the effort and expected impact of each, and a build order. Roughly a third of the items are usually configuration changes your own team can make in an afternoon. We are explicit about which those are.
Only after that do we start building workflows — on a pipeline that reflects reality, with data clean enough to trigger on, and hand-off rules that hold up at 9pm on a Sunday.