Inbound Leads Die in the Handoff Not in the Inbox

Your inbound leads aren’t “low intent.” They’re getting stranded in the handoff between your form, your inbox, and someone’s memory of what a good follow-up looks like, and by the time you notice, the trail is cold and the CRM is full of polite fiction. That gap is the system. Not the salesperson. This playbook builds a working lead triage and response loop using four tools with clear jobs: Webflow (capture), n8n (routing logic), Supabase (state + audit trail), and HubSpot CRM (ownership + pipeline). Start with Webflow: one form, no variants, no “contact us” alternatives. Every lead must enter through the same throat so you can measure latency, not anecdotes. On submit, Webflow posts to an n8n webhook. That workflow does the dirty work: normalize fields, enrich lightly (company size from domain if you have it), and compute a lead score using explicit rules you can argue about. Then n8n writes a record to Supabase, not the CRM. This is the unpopular part. Supabase becomes your source of truth for processing state: received_at, routed_to, first_touch_due, first_touch_at, and a running event log. It’s your black box recorder when someone claims “I never got the lead.” Finally, n8n pushes the lead into HubSpot with an owner, SLA clock, and a required next step. If no first touch event hits Supabase before the due time, n8n escalates: reassign owner, ping a channel, create a task, and tag the lead as “missed_sla” so you can see patterns. The outcome isn’t faster replies. It’s fewer invisible failures. You stop debating effort and start interrogating latency, routing policy, and where work actually stalls.

Maya runs growth at a 14-person B2B SaaS. She’s the one who gets yelled at when “inbound is trash” because deals don’t materialize. Except the leads are there. Somewhere. In a Slack ping. In someone’s inbox. In HubSpot as duplicates with slightly different emails. Quietly rotting. Monday, 9:12am. A prospect fills the Webflow form. Same form for everyone. No “book a demo” side door. Webflow hits the n8n webhook. n8n normalizes the payload, strips weird whitespace, grabs a company size guess from the domain, computes score. High intent. It writes the lead to Supabase first, with received_at and first_touch_due set to 10 minutes. Here’s the friction: the first version didn’t write to Supabase first. They pushed straight into HubSpot and then tried to “sync back” state. It worked until it didn’t. HubSpot rate-limited for 6 minutes one afternoon, n8n retried, and HubSpot created two contacts because the email field had a trailing space the first time. Ownership split. Two reps “followed up.” Neither logged it correctly. Supabase had no record of the original receive event because it only wrote after HubSpot success. The black box was missing the crash. Now, with Supabase first, Maya can see it: received_at exists even if downstream systems choke. n8n pushes into HubSpot, assigns owner, creates a task with a due timestamp. Rep replies from Gmail but forgets to log the email, so no first_touch event hits Supabase. The SLA clock keeps running anyway. At 9:22am, n8n escalates. Reassigns owner. Pings #inbound. Creates a second task. Tags missed_sla. Maya isn’t thrilled. The rep isn’t either. “I replied, I just didn’t click the stupid button.” True. But did the prospect get it? Did it bounce? Did they open it? Who knows. That’s the uncomfortable part. The system stops accepting vibes as evidence. It forces a question nobody likes: are we measuring follow-up, or just measuring what we bothered to record?

Here’s the part people hand-wave: this loop is only “clean” while the team is small and the paths are few. Once you add a second form, a second product line, SDR + AE splits, partner leads, and a calendar tool, n8n turns into the place where logic goes to die. Not because n8n is bad. Because every exception feels reasonable in isolation. “If they’re enterprise, route to Sam unless it’s EMEA unless it’s a competitor unless the domain is Gmail.” Six months later, nobody can explain why a lead went to the wrong owner without opening a workflow that looks like a circuit board. Supabase as the black box helps, but it also introduces a new failure mode: you’ve now created an internal system that people will start treating like an app. They’ll ask for dashboards, permissions, data retention, PII policies, “can it backfill tasks,” “can it dedupe across subsidiaries,” “can we merge two leads without breaking the audit log.” Congrats, you’re running a mini product. If we’re being honest, the hidden complexity isn’t routing. It’s the definition of “first touch.” Is it an email sent? A reply received? A call attempt? A LinkedIn message? A meeting booked? Each answer changes incentives and creates new ways to game the metric. The rep who hates logging will find the lowest-effort event that stops the clock. The manager will optimize for compliance. The prospect will still ghost. So the scalable version isn’t “add more rules.” It’s deciding what you’re willing to treat as evidence. Pick one or two signals you can actually capture automatically. For example: meeting booked (Calendly/HubSpot meetings) or outbound email sent via a connected inbox with tracking. If it can’t be verified, it doesn’t stop the SLA. That sounds harsh until you realize the alternative is letting everyone redefine reality in private. And yes, you’ll lose some nuance. But you’ll gain something more useful: a system that doesn’t require a tribal expert to babysit it.