Lead Routing Best Practices: Stop Losing Deals to Slow Follow-Up

The stakes are mostly about time. Research published in Harvard Business Review found that firms contacting web leads within an hour were roughly seven times likelier to have a meaningful conversation with a decision maker than those waiting even a couple of hours — and many companies took days. That study is old enough to drink, and yet in audits we still routinely find hand-raised demo requests sitting unowned for 12+ hours because a queue rule silently failed in 2024 and nobody noticed. This is the field guide: the prerequisites routing depends on, how to choose a routing model, SLA design, and the five failure modes to check this week.

The prerequisites most teams skip

Routing logic is the visible 20% of a routing system. The invisible 80%: Deduplication before assignment. The most common routing bug isn’t a broken rule — it’s a duplicate record. The new lead matches an existing contact owned by Rep A, but the duplicate routes to Rep B, and now two reps are calling the same prospect, or worse, neither is (each assumes the other owns it). Dedupe-on-create is a routing feature. If your database hygiene is shaky, fix that first — routing distributes whatever you feed it, including the mess. (Our data hygiene system guide covers the intake controls.) Defined lifecycle stages. Routing fires on a lifecycle transition — usually MQL, always hand-raise. If your MQL definition is contested, routing inherits the argument: reps deprioritize leads they don’t believe in, and your speed-to-lead metric quietly becomes a rep-trust metric. Stage criteria come first; see our lifecycle stages framework. Enrichment before rules. Territory and segment routing need firmographic fields — employee count, region, industry — populated at routing time. If enrichment runs asynchronously after assignment, your rules route on empty fields and everything lands in the fallback queue. Sequence the automation: create → dedupe → enrich → route. A complete owner roster. Routing tables rot through personnel changes. A rep leaves; their routing slot becomes a black hole. Ownership rosters need a review trigger tied to offboarding, not to quarterly cleanup.

Choosing a routing model

There is no best model — there’s a best model for your motion and team size. The honest comparison:

Model	How it works	Best for	Watch out for
Round robin	Rotates leads evenly across a pool	Inbound-heavy, interchangeable reps, small teams	“Fairness” ≠ fit; add availability checks or vacations create black holes
Territory / geo	Assigns by region or country	Field sales motions, compliance-bound markets	Depends entirely on clean, present geo data at routing time
Segment / size	Assigns by employee band, industry or tier	Teams split SMB / mid-market / enterprise	Enrichment latency; disputed boundary accounts need a tiebreak rule
Account-based / named	Matches lead → account → account owner	ABM motions, existing-customer expansion	Lead-to-account matching is its own engineering problem — fuzzy domain matching, subsidiaries
Weighted / capacity	Skews distribution by quota, seniority or current load	Scaling teams, ramping reps	Complexity creep; document weights or the model becomes unexplainable
Direct-to-calendar	Hand-raisers book a meeting directly; routing assigns the booked rep	Demo requests, pricing inquiries	Still needs ownership + dedupe behind it, or the CRM forks from the calendar

Two model-selection principles from the field:

1. Hand-raises get the fastest possible path, always. A demo request should hit a rep’s calendar or phone within minutes, bypassing scoring and nurture logic entirely. Sophisticated routing that slows down your hottest leads is optimizing the org chart, not the revenue.
2. Start simpler than you think you need. A round robin with availability checks and a working escalation beats an elegant weighted-territory hybrid with a silent failure mode. Every branch you add is a branch that can rot.

SLA design: routing without a clock is just labeling

Assignment is half the system. The other half is what happens if the owner doesn’t act:

• Set the first-touch SLA by lead type, not globally. Hand-raise: minutes. MQL: same business day. Recycled re-engagement: 48 hours. One global SLA is either too slow for demos or unrealistic for everything else.
• Escalate on breach — don’t just report it. After the SLA window, reassign to the next rep or alert the manager automatically. A weekly SLA-breach dashboard is an obituary; an escalation rule is a defibrillator.
• Measure from creation, not from assignment. Speed-to-lead clocks that start at assignment hide the system’s own latency — sync delays, enrichment lag, queue time. The prospect’s clock started at form submit; yours should too.
• Route to availability, not just to names. Working hours, PTO calendars, and capacity caps belong in the rules. The #1 cause of blown SLAs isn’t lazy reps; it’s leads assigned at 6:01 pm Friday to someone who’s out Monday.

The five failure modes to audit this week

Run these checks; they take an afternoon and regularly find revenue.

1. The unrouted graveyard. Query for open leads with no owner or a queue owner, older than 24 hours. Anything above ~zero means your fallback logic is doing more work than it should.
2. The sync gap. In two-system stacks (marketing platform + CRM), measure the timestamp delta between record creation and routed assignment. Sync batching and integration errors add silent hours; if HubSpot and Salesforce are both in the path, the sync’s filters and field mappings are part of the routing system whether you designed them or not.
3. The stale roster. Compare active routing-rule members against the current team list. Departed reps and changed territories are the classic silent leak.
4. The duplicate fork. Sample recent leads that matched an existing account: did they route to the account owner or to the pool? Inconsistency here means your lead-to-account matching needs work.
5. The after-hours cliff. Chart speed-to-lead by hour of submission. If evenings and weekends are 10× slower, decide deliberately what should happen then — a booking link, an on-call rotation, or an honest SLA — rather than letting the cliff be your policy.

Iterating: routing is a product, not a project

The teams whose routing works treat it like a product with a backlog: a named owner, a monthly review of SLA and distribution metrics, and a change log for the rules. The metrics worth watching are few — median speed-to-lead by type, SLA breach rate, distribution variance across reps, and fallback-queue volume. When those four are healthy and boring, routing has done its job: nobody talks about it, and the deals that were being lost to slow follow-up simply stop being lost.