Deals don't die in a stage. They die between two of them, where nobody's name is written.
When a sales or operations team complains about pipeline leakage, the first reaction is usually to add a stage or add automation. Both normally make it worse, because the problem is not the number of stages.
What an SLA is inside a CRM
An SLA inside a CRM is a named owner plus a clock: who is responsible, and by when they must act.
Without either half there is no SLA. A stage with a deadline and no owner is a deadline nobody misses. An owner with no deadline is a responsibility that never comes due.
The definition is deliberately brutal, because most implementations I have seen are missing one of the two.
Map stages to decisions, not to activities
The most common design error is stages that describe activity: "called", "proposal sent", "following up".
The problem with those is that leaving them has no criterion. When does "following up" end? Whenever someone decides. So deals accumulate there.
A stage defined by a decision has an exit criterion, because it asks a yes/no question:
- "Has the customer confirmed they have this problem?"
- "Are budget and decision authority identified?"
- "Has the decision maker seen the proposal?"
Each of these is answered or it is not. A stage whose answer is "roughly" has not finished.
The handoff table: who hands what to whom
Leaks happen between stages, so handoffs have to be written explicitly.
| From | To | What transfers | Deadline |
|---|---|---|---|
| Marketing | Sales | Lead source, stated problem, urgency level | 4 working hours |
| Sales | Onboarding | Agreed scope, promises made, technical contact | 1 working day |
| Onboarding | Support | Configuration, open items, customer expectations | 2 working days |
| Support | Sales | Renewal or churn signal, issue history | Continuous |
Column three matters most and is almost always missing. "Promises made" is the largest single source of first-month dissatisfaction — something a salesperson said that was never recorded in any field.
Where leaks appear
Unowned stages. If ownership is assigned to a team rather than a role, nobody owns it. "Sales" means nobody.
Silent expiries. A deadline passes and nothing happens. If an SLA breach raises no notification, the SLA is decorative.
Reopened deals. A record that moves from stage four back to stage two and restarts the clock. If reopenings are not counted, the pipeline looks healthy while the same deals circle.
Three reports that expose a leaking pipeline
- Time in stage, by owner. If one person averages three times the rest, the issue is either workload or the stage definition.
- SLA breach rate by handoff. Shows which boundary is broken, not which person.
- Count of moves backwards. The highest diagnostic value and the least commonly used.
Assign ownership to a role, not a person
A subtle distinction that decides everything in practice: the owner of a stage should be a role, and that role should map to one person at any given moment.
Assign ownership directly to a person and every holiday, resignation and team change breaks the flow. Assign it to a team and nobody is responsible. A role is the only level that solves both.
Practically this means the CRM needs one small table: role, current person, deputy. Changing one row in that table moves every SLA correctly. Without it, every personnel change becomes a configuration project.
When automation helps and when it hides
Automating an unowned process only makes the leak faster. The working rule: automate nothing until the owner and deadline of every handoff are written down.
After that, three things are worth automating and most others are not: assignment to the owner, a reminder before the deadline rather than after it, and automatic escalation once it passes.
What usually gets automated and should not: follow-up emails to the customer. That raises the activity number, lowers the quality of contact, and looks like improvement in the report.
A worked example: the four hours nobody measured
A seven-person services team watched inbound lead conversion fall from 22 to 13 percent over six months. The team had grown, not shrunk.
The CRM held an SLA: respond to a new lead within four working hours. Nobody measured it.
When they did, the average was 9 hours and the 90th percentile reached 31. The reason was mundane. With three people, whoever saw a new lead picked it up. With seven, everyone assumed someone else already had.
The fix was not sophisticated automation — automatic assignment to a named person, plus a notification at hour three. Conversion returned to 20 percent within two months.
The point of the story is that the problem existed from the start. Team growth simply broke an assumption nobody had written down.
Rolling it out without a mutiny
SLAs turn into surveillance quickly, and then the team learns to game the system — ticking stages early, parking deals in a safe stage.
Three rules that work in practice. First, measure only for the first month and send no manager alerts. Second, read breaches as a capacity signal rather than a performance signal. Third, when an SLA is missed constantly, assume the number is wrong before assuming the person is.
The third is the hardest and buys the most trust.
SLAs on handoffs, not on every stage
A common implementation error is putting a deadline on every stage. It almost always produces resistance, and for good reason.
How long a stage takes depends on external factors — the customer does not reply, the decision maker is travelling. Setting deadlines on things the team does not control only makes the numbers meaningless.
Handoffs are different. The time to pass a record from one role to another is entirely under internal control. Deadlines on handoffs are fair, enforceable, and land exactly where the leakage happens.
The trade-off
Tight SLAs raise speed and can lower quality. Mandate a response within four hours and you get more shallow responses.
The workable approach is defining the SLA on "first meaningful response" rather than "first contact". The number gets harder to hit and starts measuring something that converts.
How many stages is ideal?
Four to six for most businesses. More than seven almost always means some stages are activities rather than decisions.
Does a three-person team need SLAs?
Ownership yes, deadlines no. At three people implicit coordination works. Around twelve is where it stops.
What if the team won't fill in the fields?
Usually that means the field exists for reporting rather than for work. A field that does not inform the next decision should be deleted.
Closing
Your pipeline probably does not lack stages. It lacks an owner and a clock. Every handoff missing those two is a place where work stops without anyone noticing.
If you want this layer built into your own operations, that is what intelligent business systems are for.
Comments (0)
You need to log in to post a comment.
Login / Sign up