Stale leads are a design problem, not a hygiene problem

A decision-maker at one of your accounts changed jobs four months ago. Your CRM still has their old title, old company email, old company.

Next time the AE runs outreach, the email bounces. The exact pattern that should be triggering renewed attention — a champion landed at a target account, a buyer moved to a competitor — is invisible to your system.

Most teams call this a database hygiene problem. Once a quarter, someone runs a contact-refresh job. Done.

The trouble is the moments that matter don’t move on a quarterly schedule.

A funding round announced last Tuesday changes how an account scores today, not in three months. A buyer who took a new role at a different company is an opportunity the day they update LinkedIn, not 90 days later. The signal is live; the response is dead.

Hygiene is the wrong frame. What you need is a freshness pattern — a continuous workflow that re-processes records the moment the world around them changes.

A freshness workflow has four moving parts, and they all have to run continuously — not on a calendar.

1. Signal detection. The system watches a set of monitored signals and fires when one changes. Person-level signals: job changes, role changes, email bounces. Company-level signals: funding rounds, headcount shifts, M&A, leadership changes. Engagement-level signals: long stretches of silence, suddenly active again, opened the last newsletter after six months of nothing.

Not every signal applies to every record. The system needs an opt-in or a rule-based filter — “monitor job change on contacts flagged as champions, on decision-makers at accounts that have gone quiet, on anyone whose company is in your top ICP segment”. Watching everything indiscriminately produces noise; watching nothing produces stale data. The trick is opting in deliberately.

2. Re-enrichment. Once a signal fires, the system fetches whatever new data the signal implies. New title? Re-pull the LinkedIn profile. New company? Re-enrich the company record. Funding round? Update the company’s headcount, segment, and capacity signals. The point is to make the structured record match the current truth, not last quarter’s truth.

3. Re-scoring. With new data, the ICP score recomputes. A contact who scored B at a 30-person startup might score A after a Series B doubled their headcount. A company that scored A two quarters ago might fall to C after their main competitor acquired half their customer base. The score isn’t a one-time calculation — it’s a function of data that changes over time.

4. Re-routing. If the score change crosses an owner-assignment boundary — segment, region, deal size, account tier — the system reassigns ownership. The old owner gets a handoff note; the new owner gets the account with the signal that triggered the change preserved as context. No record drifts into a “still owned by someone who doesn’t fit anymore” state.

Detect → re-enrich → re-score → re-route. Continuously, not quarterly.

Three things go wrong with most freshness implementations:

Partial enrichment failure. The system detects a job change, finds the new company, but can’t find the new email yet. The right behavior is not to update what it has and leave the email field empty — that creates a half-broken record an AE downstream might trust. The right behavior is to update what it found, hold the rest, and create a task for human review. Half-done is worse than not-done when the data is going to drive an action.

Conflicting signals. A contact and the company they work for fire signals in opposite directions on the same day — the contact got promoted (ICP score up), but the company just announced layoffs (ICP score down). Which signal wins? The pattern needs an explicit precedence rule. At Novrith we weight company-level signals over contact-level when they conflict on the same day, on the theory that the company is the buying unit.

Re-routing without context loss. When an account changes owners because the score crossed a threshold, the new owner inherits a record stripped of context. The fix is structural: the routing automation has to write a handoff note that captures the signal that triggered the change, what the prior owner had done, and where the prior owner left things. The new owner shouldn’t have to ask.

None of these are exotic — they’re the unglamorous failure modes that decide whether the pattern actually works in practice.

A freshness pattern that’s running well surfaces three healthy metrics:

• Freshness rate by segment. What percentage of your A-tier accounts have firmographic data refreshed in the last 90 days? If that number is below 70%, the trigger logic isn’t catching enough signals. If it’s above 95%, you’re probably re-processing too aggressively and burning enrichment budget on no-change deltas.
• Re-enrichment hit rate. Of the signal-triggered re-enrichments that fired this week, how many actually found a delta worth writing back? A low rate (say, <20%) means the trigger conditions are too loose. A very high rate (>80%) probably means you’re not catching enough records — the trigger conditions are too tight.
• Days from signal to action. When a signal fires, how long does it take to land in a human’s queue with all the context attached? Under 24 hours is the target. Over 72 hours and the pattern is too slow to capture the moment.

If any of these three drifts, you have a design issue — not a hygiene issue. The fix is in the rules, not in another cleanup.

Tomorrow on LinkedIn, I’m shipping a 2-minute walkthrough of one specific implementation of this pattern: a job-change-detection automation that catches when monitored contacts move companies and turns it into a refreshed Person record plus a queued task. One signal type, one re-process workflow, one human decision point — a clean example of what this pattern looks like wired up.

That’s the slice. The newsletter is the pattern.

Talk soon,

Marco

Founder / CEO, Novrith