Systems / Blueprint 03
Your leads get answered in under ten seconds.
One event-driven pipeline takes every form fill, ad conversion and chat message through validation, dedupe, enrichment, scoring and routing. Median time from capture to assigned rep: eight to nine seconds. The average B2B company answers in 42 hours. You will not be average.
Every hour of delay burns the deal
In B2B sales the first response wins. Most companies take 45 minutes to 4+ hours to route a lead to a rep. While they deliberate, buyer intent cools and conversion drops by 40 to 60%. Revenue operations end up scaling with headcount instead of capturing demand in real time.
- Response delay: the average B2B response time is 42 hours against a 5-minute best practice.
- Inconsistent qualification: every rep applies their own standard to the same lead criteria.
- Territory conflicts: manual assignment creates disputes and coverage gaps.
- Data duplication: the same lead enters three times and nobody trusts the record.
- Missing context: reps start cold with no firmographic or intent data attached.
The pipeline
Event-driven microservices built for low latency. Asynchronous queues prevent blocking, circuit breakers protect downstream APIs and distributed caching cuts external lookups. Every stage scales horizontally for high-volume inbound periods. The stage-by-stage build is in my automated lead routing guide.
01 · Enrichment and routing
- CaptureWebhooks / Forms / Ads
- ValidationSchema and sanity
- DedupeRedis + fuzzy
- EnrichClearbit / Apollo
- ScoringML fit + behavior
- RouterTerritory / skills
- AssignRep / queue
02 · Write-back and alerting
- ConflictResolver
- CRM WriteCreate / update
- NotifierSlack / SMS / Email
- DashboardRep / mobile
main path conflict-resolution feedback into the CRM writeRouting hands the lead to a rep and the notifier; scoring conflicts enter the resolver before write-back.
Capture Gateway
One unified webhook endpoint accepts form fills, chat transcripts and ad conversions. Payloads get normalized and schema-checked within 50ms.
Duplicate Resolver
Deterministic matching on email and phone, fuzzy matching on company name and domain, Redis lookups in under 100ms. Merge strategies prevent data loss.
Enrichment Engine
Parallel lookups against Clearbit, Apollo and LinkedIn. Results are cached to keep API costs down, and the engine falls back to cached data when a provider goes dark.
Scoring Service
An ML model blends firmographic fit with behavioral interest into a single 0-100 score. It retrains on conversion outcomes, so every closed deal sharpens the next routing decision.
Router Service
Territory rules, load balancing and rep availability applied in one pass. Account-based routing, round-robin and skill-based assignment all run through the same engine.
Notification Hub
Slack, SMS, push and email alerts with priority queuing. If the primary rep does not acknowledge inside the SLA window, the lead escalates to the next rep automatically.
Where the leads come from
Form Submissions
Website forms, landing pages, webinar registrations and content downloads, each arriving with UTM parameters and session context attached.
Ad Conversions
Facebook Lead Ads, Google Ads conversions and LinkedIn Lead Gen Forms, carrying full campaign attribution from the first click.
Chat Conversations
Drift, Intercom or custom chatbot transcripts, with qualification answers and engagement scoring already captured.
Built to fail safely
Latency budget
Roughly 8 to 9 seconds median from capture to assignment, measured per stage rather than guessed at the end.
Enrichment fallback
When a provider times out, the pipeline uses cached data and keeps moving. A lead never waits on a lookup.
Merge audit trail
Every duplicate merge is logged with its inputs. Nothing disappears from a record without a paper trail.
Human approval
High-value or conflicted leads route to a manual queue with an SLA timer, and the router keeps a dead-letter queue for anything it cannot place.
Deployment roadmap
Typical build runs six to eight weeks from kickoff to cutover. Each phase ends with something live against your real traffic.
Instrumentation and dedupe
Baseline response times measured, dedupe layer running on your highest-volume lead source.
Enrichment and scoring
Providers wired in, caches warmed, scoring model trained on your historical conversions.
Routing and notification
Territory rules, escalation chains and Slack/SMS alerts running end to end.
Load test and cutover
Full-volume load testing and failure drills, then every remaining source switches over.
What a routing decision looks like
One real decision object from the router. The reason field is what makes it auditable.
{
"lead_id": "ld_4b91c2",
"received_at": "2026-10-14T09:12:03Z",
"dedupe": { "status": "new", "matched_on": null },
"enrichment": { "company": "Acme Logistics", "employees": 120, "industry": "logistics" },
"score": 78,
"routing": { "owner": "rep_maria", "territory": "EU-north", "skills": ["logistics", "mid-market"], "sla_minutes": 10 },
"decision": "route",
"reason": "territory match + logistics skill + score above 70 threshold"
}Related systems
The pieces this one plugs into.
Frequently asked
How fast is fast enough for a lead response?
Under 10 seconds from form submit to rep assignment. After 5 minutes, contact rates drop sharply. The architecture targets 8 to 9 seconds end to end.
What happens when enrichment fails?
The pipeline continues with whatever data it has. Enrichment is best-effort; routing still happens on the fields that exist, and the failure is logged for retry.
Can it handle duplicate submissions?
Yes. Redis-backed dedupe catches the same email or phone within a configurable window before it reaches scoring.
What about leads that arrive at 2 AM?
They get scored and routed immediately. The rep's SLA clock starts at assignment, and escalation rules handle off-hours according to your policy.
Stop paying the slow-reply tax.
Paste your current lead flow into the free diagnostic and I will tell you what sub-10-second routing takes on your stack, or where a full CRM API workflow overhaul fits better.
Stay in the loop.
Technical breakdowns, new system builds, and lessons from deploying autonomous AI. Direct to your inbox.
Weekly architectural updates. Unsubscribe anytime.