insights revenue os architecture

Where leads fall through the cracks: eight states a lead stops moving in

Direct answer. "Leads falling through the cracks" is a symptom, not a diagnosis. It resolves to a small number of specific states where a lead stops having a record, an owner, or a next action — and each of those states is something you can test for directly rather than argue about. The eight below cover the path from arrival to an owned next action. Each names what the record looks like when a lead is stuck there and the single export that finds it, so the question moves from "are we losing leads?" to "which of these, and how many."

related system · Lead management · Revenue OS audience · Operators and revenue leaders who can see that inquiries are not turning into worked conversations, and who need to locate where that stops happening before choosing a repair.
On this page 8 sections

Why the phrase is hard to act on

"Leads are falling through the cracks" describes an outcome, and outcomes are hard to fix because everyone in the room can hold a different theory about the cause without either of them being testable. The phrase is accurate — it is just not yet a finding.

What makes it actionable is that a lead can only stop moving in a limited number of places. Each one is a state the record is sitting in right now, which means each one can be queried rather than debated. The useful question is not whether you are losing leads — at some level the answer is always yes — but which of these states your records are actually in, and how many.

Settle the denominator first

None of the counts below means anything until "a lead" is defined. If spam, existing customers, recruiters, suppliers and genuine inquiries all sit in the same table, every number you produce is a mix of populations and the exercise argues in circles.

Write down what makes an inquiry eligible before you count anything, and apply the same rule to every number. That contract is the first step of a lead management implementation for exactly this reason, and it is the same discipline as denominator integrity elsewhere in the estate: a rate is only comparable when both sides of it were counted the same way.

The eight states

This is first-party diagnostic methodology, written to be run against your own records. It names where a lead stops and what to look at. It is not a benchmark, it carries no industry figures, and it does not claim how common any state is [1] .

State What you hear What the record shows The export that finds it
1 · Never captured "They told me they filled in the form." No record exists at all for an entry action that completed. Completed entry actions per channel for one period, against records created in the same period — same window, same timezone.
2 · Captured without origin "I cannot tell where these came from." The record exists; the campaign, source and referring context are empty. Share of records created in the period with a null source, campaign or referrer field.
3 · Captured twice "Two of us called the same person." One human being, two records, with different owners and divergent history. Normalised email and phone collisions within the period, grouped by owner.
4 · Captured but unowned "I thought someone else had it." The record exists with no assignee, or sits in a queue nobody works. Records with a null owner, plus records that have an owner and zero activity rows.
5 · Owned but never actioned "Nobody called them back." An owner is set and no first outbound action was ever recorded. First activity timestamp minus creation timestamp, filtered to eligible records only.
6 · Actioned once, then silent "We tried them." Exactly one attempt, no sequence enrolment, and no open task. Records with one activity row, no open task and no active sequence.
7 · Sequence ended, no next state "The automation is running." The cadence completed and returned the record to the same undifferentiated pool it started in. Records whose last sequence step is complete and whose status still equals its creation status.
8 · Closed without a reason "Those were all junk." Closed or disqualified, with nothing recorded about why, so the cohort can never be re-examined. Closed or disqualified records with a null reason field.

The three states people argue about most

State 4 is the one that hides best. A record with no owner is not an error in any system — it is a valid row. Nothing fails, nothing alerts, and the queue gets worked when somebody remembers rather than when something fires. That is the shape described as an unowned queue, and it is why "is anything broken?" returns a clean answer while leads sit still.

State 7 is the one most often mistaken for success. A finished sequence looks like completed work. Whether it was depends entirely on whether finishing put the record somewhere different from where it started. A cadence that ends by returning the record to the general pool has automated the contact and left the decision undone — the distinction held by terminal state configuration and, for the people who never replied at all, by non-responder escalation.

State 8 is the one that costs you the diagnosis. Closing without a reason is cheap on the day and expensive later: it is the field that would have told you whether the leads were genuinely unqualified or whether the qualification rule is wrong. Without it, "those were all junk" can never be checked, so the same argument recurs every quarter with no new evidence.

The check, in an afternoon

You do not need new tooling to locate the states you are in. You need thirty recent inquiries and the discipline to include every channel you actually receive on — form, phone, email, live chat, direct message, referral, walk-in. Sampling only the channel that is easiest to export is how state 1 survives audits.

  1. Take thirty recent eligible inquiries, spanning every channel, using the eligibility rule you wrote down above.
  2. For each one: is there exactly one record for it?
  3. Who owns it, by name? "The team" is a null owner with better manners.
  4. What was the first outbound action, and when did it happen relative to arrival?
  5. If it is open: what is the next action, and what date is it due?
  6. If it is closed: what reason was recorded?
  7. Any inquiry you cannot answer all five questions for is a finding. Sort the findings by state and by channel — the concentration tells you where to look first, and it is usually one channel and one state rather than a general malaise.

The output of the check is a list of states with counts against them, not a score. Thirty records is a sample that locates a structural gap; it is not a rate, and it should not be reported as one.

What this does not establish

  • It does not establish what any gap costs. Nothing here supports a lost-revenue figure, and none is offered.
  • It does not establish a rate. Thirty records locate a problem; they do not measure its frequency across the whole population.
  • It does not establish a cause. A missing next action can be a process gap, a tooling gap or a capacity gap, and the record looks identical in all three cases.
  • It does not establish that the leads were worth having. A lead that was never real is not a leak, which is why the eligibility rule comes first.
  • It does not claim that repairing any state produces a particular improvement. No recovery rate is stated or implied.
  • None of the eight is observable from a public surface. These are internal-record findings and require access to the records to make at all.

Where this goes next

  • For the design question rather than the diagnostic one — what the handoff between acquisition and CRM has to enforce, and the failure modes a design must expose — read how to connect lead acquisition to CRM and follow-up. This page finds where a lead stopped; that one specifies the contract that stops it happening again.
  • For the wider method — how to grade the evidence behind any suspected revenue failure, not just this one — read the revenue leak taxonomy.
  • For the implementation that owns capture, routing, response and follow-up as a working system, see lead management automation. For the record model and lifecycle underneath it, see CRM automation.
  • If you would rather walk your own path with someone than run the check alone, a Diagnostic Review is a live working session, not a form submission.

Source and evidence notes

  • [1] OmniLabs Systems first-party diagnostic methodology. The eight states and the thirty-record check are our own working method for locating where a lead stops moving. Limitation: Methodology only. It carries no industry benchmark, no frequency data, no financial quantification and no claim about outcomes from repairing any state.

Related entities

  • Unowned Queue A list of records a system already produces that has no assigned owner, no ageing threshold and no rule turning an entry into work.
  • Non-Responder Escalation An explicit post-sequence state an operator may define when a follow-up sequence completes without a recorded response.
  • Terminal State Configuration The setting that determines what status a record is left in when an automated sequence exhausts its attempts, and which systems are told about that status.
  • Denominator Integrity The property of a conversion metric whose denominator includes every inquiry that occurred, not only the inquiries a system of record happened to receive.
  • Follow-Up System The owned operating layer that turns a captured inquiry into explicit record states, next actions, exceptions, and verification evidence.
RELATED IMPLEMENTATION

From operating model to delivery scope.

Move from locating the state a lead stopped in to the system that keeps capture, ownership, response and follow-up from producing it again.

Lead management automation
[CLAIM BOUNDARY] All eight states are internal-record findings. A public surface can frame the question but cannot establish any of them. Finding a state establishes that a structural gap exists — never how much money it costs, how often it occurs across the whole population, or whether the cause is process, tooling or people.
Book Diagnostic Review Back to insights