- 01Why Continuous Beats Batch
- 02The 7 Stages Detailed
- 03Vercel Cron Scheduling
- 04Signal Monitoring
- 05Continuous Optimization
- 06FAQs
Why Continuous Beats Batch
Most CRE technology runs in batch mode. You pull a list of prospects on Monday, run outreach on Tuesday, check results on Friday. This cadence was designed for a world where data arrived slowly and analysis took time. That world no longer exists. Engagement data arrives in real time. Market signals change hourly. A prospect who was Warm at 9 AM may be On Fire by noon because they just visited three property pages and downloaded a market report.
The IO Pipeline Orchestrator replaces batch processing with a continuous loop. Instead of running once and waiting for human review, the pipeline cycles through seven stages and then starts again. Each cycle incorporates new data from the Monitor stage, so the pipeline gets smarter with every rotation. The question is not “did we run the pipeline this week” but “how many cycles has the pipeline completed since the last deal closed.”
For a portfolio of 78 properties, continuous orchestration means every property has an active pipeline running. No property sits idle waiting for a broker to initiate outreach. No prospect goes unscored because the weekly batch hasn't run yet. The pipeline runs whether anyone is watching or not.
"A pipeline that runs once is a report. A pipeline that runs continuously is an operating system. The difference is the difference between knowing what happened and controlling what happens next.
The 7 Stages Detailed
Each stage has a defined input, processing agent, output, and handoff protocol. The stages are sequential within a cycle but the pipeline can run multiple property cycles in parallel.
Stage 1: Seed
The Seed stage identifies new prospects by querying data sources: CoStar lease expirations, company expansion announcements, funding rounds, job postings in target markets, and inbound website traffic. It runs every 6 hours via Vercel cron, pulling fresh data and deduplicating against the existing prospect database in Supabase. New prospects enter the pipeline unscored and are immediately queued for Stage 2.
Stage 2: Score
Score runs the 14-signal qualification model against every new and updated prospect. This is the Qualification Agent from Article 07 — firmographic, behavioral, transactional, and relationship signals combined into a composite 0–100 score. Score runs immediately after Seed completes, processing the full queue in parallel batches of 50 prospects.
Stage 3: Audiences
The Audience stage rebuilds all dynamic segments based on updated scores. Prospects are assigned to segments, promoted between tiers, or flagged for disqualification. This stage also identifies segment gaps — if a high-value segment has fewer than 10 prospects, it triggers a targeted Seed query in the next cycle.
Stage 4: Creatives
Creatives generates fresh content packages for any segment that has new prospects or updated positioning. This is the Content Agent from Article 08 — brochures, articles, target profiles, and ad copy. Creatives runs in parallel with Audiences because each segment's content is independent.
Stage 5: Launch
Launch deploys outreach sequences across email, LinkedIn, and phone. On Fire and Hot prospects require broker approval before sequences launch — this is the human-in-the-loop gate. Warm and Neutral tier outreach launches automatically. Launch also schedules follow-up touchpoints based on the prospect's tier-specific cadence.
Stage 6: Monitor
Monitor runs continuously between full pipeline cycles, tracking 8 signal types: email opens, email clicks, website visits, content downloads, LinkedIn profile views, LinkedIn message responses, phone call completions, and event registrations. Real-time triggers fire when engagement exceeds defined thresholds — three email opens in one hour triggers an immediate broker alert.
Vercel Cron Scheduling
The orchestrator uses tiered scheduling based on property activity level. High-activity properties (recent listing changes, upcoming lease expirations, or high prospect engagement) run full pipeline cycles every 4 hours. Medium-activity properties run every 8 hours. Low-activity properties run every 24 hours. This ensures the most active properties get the most frequent attention while keeping compute costs manageable across 78 properties.
Signal Monitoring
The Monitor stage is unique because it runs continuously rather than on a cycle. It watches for engagement signals that should trigger immediate action, regardless of where the pipeline is in its current cycle. A prospect who opens three emails in one hour does not wait for the next Rescore cycle — they get an immediate tier promotion and broker alert.
The monitoring system tracks 8 signal types with defined thresholds for immediate action. These thresholds are calibrated per property type because engagement patterns differ — industrial prospects tend to be more decisive (fewer touches to conversion) while office prospects tend to be more deliberate (more research before engagement).
Continuous Optimization
Stage 7: Rescore
Rescore re-evaluates every prospect's composite score using data collected during Monitor. Prospects whose scores cross tier boundaries are automatically promoted or demoted. A Warm prospect whose email engagement spiked gets rescored to Hot and enters the broker-managed sequence. A Hot prospect who went silent for 30 days gets demoted to Warm.
Rescore also feeds the next Seed cycle by identifying coverage gaps. If a high-value segment is declining in size, Rescore triggers a targeted seed query to replenish it. This creates a self-healing pipeline — segments that lose prospects automatically trigger efforts to find new ones. The loop restarts. The pipeline never stops.
"The pipeline does not wait for instructions. It generates its own instructions based on what the data says happened since the last cycle. That is the difference between automation and intelligence.