Late project spikes to Salesforce–SAP IBP signals for time?phased demand, fewer expedites

Overview

A building materials company’s demand plan overlooked project pipeline signals sitting in the Customer Relationship Management (CRM) system, so large orders appeared late and triggered rush purchases and emergency shipments. Intelligex connected CRM opportunity milestones to the planning system, standardized how project signals translate into time-phased demand by product and location, and added an AI-assisted anomaly reviewer to flag deviations from seasonal curves. Planners made earlier, better-context decisions, procurement placed orders on steadier lead times, and expedite calls tapered.

Client Profile

• Industry: Building materials manufacturing and distribution
• Company size: Multi-region footprint with plant and distribution center network
• Stage: Mature supply chain formalizing sales and operations planning
• Department owner: Procurement, Supply Chain & Logistics
• Other stakeholders: Sales, Demand Planning, Finance, Plant Scheduling, IT/Enterprise Applications, Regional Operations

The Challenge

Demand planners built forecasts from historical seasonality and promotional calendars. Yet a significant portion of volume came from project-based orders tied to commercial builds and renovations. Those opportunities lived in the CRM with stages, dates, and bill-of-material clues, but they never reached the planning system in a structured way. When a project reached contract stage, the first signal planning saw was an urgent purchase requisition or a phone call from Sales. The result was compressed lead times, premium freight, and frequent rescheduling at plants.

Sales exported spreadsheets from the CRM to alert planners about upcoming projects, but fields varied by region and product line. Some teams applied their own weights to stages; others sent lists without any probability or timing. Planners struggled to translate opportunities into specific product and location demand. With no common logic or system integration, the organization defaulted to reactive adjustments that disrupted procurement and production.

Replacing the core planning tool or the CRM was not an option. The company needed to integrate project signals into the existing demand plan, consistently map opportunities to product families and ship-to locations, and give planners an assisted review step that highlights when the combined forecast strays from expected seasonal patterns.

Why It Was Happening

The CRM and planning system were optimized for different views of demand. Opportunities in the CRM reflected deals with milestones and probability, while the planning system expected time-phased quantities by product and location. Without a shared translation layer, Sales’ pipeline did not inform the plan until late in the cycle. Manual exports lacked uniform stage definitions and often omitted details needed to assign demand to the right warehouses and plants.

Governance was informal. Regional teams created their own spreadsheets, and planners applied ad hoc judgment to allocate quantities and timing. No automated checks reconciled project-driven adjustments against existing seasonal curves, so over- and under-corrections were common. When outcomes missed, the team debated the data rather than the plan.

The Solution

Intelligex implemented a signal integration layer that converts CRM milestones into structured, stage-weighted demand adjustments and feeds them into the planning system alongside the statistical forecast. An AI-assisted reviewer compared the combined view to established seasonal curves and highlighted anomalies for human review. Planners accepted, adjusted, or rejected suggested changes with a clear rationale. The approach kept both CRM and planning tools in place, added governance to the translation logic, and made the review step faster and more consistent.

• CRM integration to ingest opportunities, stages, close dates, products, and locations using standard APIs. Reference: Salesforce REST API.
• Planning system integration to publish external demand signals as adjustments into the existing demand plan; reference product pattern: SAP Integrated Business Planning.
• Mapping model that translates opportunities to product families, configurations, and ship-to nodes, with support for partial bills of material and regional substitutions.
• Stage-weighting rules that convert milestones into time-phased expected demand with configurable probability curves and lead-time offsets.
• AI-assisted anomaly detection that compares the adjusted plan to seasonal baselines and recent sell-through, flagging lifts and dips outside policy thresholds. Background reference: scikit-learn: Outlier and anomaly detection.
• Human-in-the-loop review screen where planners see flagged items with context (project attributes, location capacity, supplier lead times) and approve or tune adjustments.
• Data contracts and validations for identifiers, stage definitions, and timing to prevent drift between CRM and planning.
• Dashboards for Sales, Planning, and Finance showing pipeline-derived demand by product family and region, accepted adjustments, and items pending review.
• Audit trail that captures source opportunities, mapping decisions, overrides, and publish events into the plan.
• Role-based permissions so Sales can maintain opportunity data, Planning owns translation rules and approvals, and Finance sees accepted adjustments and rationale.

Implementation

• Discovery: Cataloged project attributes in the CRM, current stage definitions, and regional differences. Mapped the planning system’s product-location hierarchy and how external demand adjustments are ingested. Identified high-impact categories and typical lead-time constraints for suppliers and plants.
• Design: Defined the translation model from opportunity to product and location demand, including stage probabilities, milestone-based timing, and partial BOM handling. Specified data contracts for opportunity fields, product codes, and locations. Designed the anomaly review workflow and decision capture.
• Build: Implemented CRM ingestion with field-level validations, built the mapping and weighting logic, and configured outbound integration to publish adjustments into the planning system. Added the AI-assisted reviewer and dashboards. Established governance for rule changes with versioning and approvals.
• Testing and QA: Ran in shadow mode, generating proposed adjustments without publishing them, and compared outcomes to historical orders and manual alerts from Sales. Injected edge cases like slipped close dates, multi-location projects, substitutions, and canceled deals to confirm fallbacks and review prompts. Verified that anomalies were sensible and that accepted adjustments flowed correctly.
• Rollout: Started with selected regions and product families where project-based demand was most volatile. Kept manual spreadsheets visible during the first cycles for confidence checks. Enabled publish after planners validated several rounds of proposals. Phased in additional regions as rules stabilized.
• Training and hand-off: Delivered role-based training for Sales on opportunity hygiene, for planners on reviewing and tuning adjustments, and for Finance on interpreting dashboards and accepted changes. Published a playbook on stage definitions, translation rules, and how to handle exceptions.
• Human-in-the-loop review: All flagged anomalies required planner approval with a recorded reason; low-impact, policy-compliant adjustments could auto-apply with planner visibility. Any change to stage weights or mapping rules required a documented approval and effective date.

Results

Planners began incorporating project-driven demand early and consistently. The translation logic turned CRM milestones into time-phased quantities at the right product and location granularity, while the anomaly reviewer surfaced only the items that warranted attention. Procurement placed orders against a steadier horizon, and plant schedules stabilized because late project spikes were rare.

Sales, Planning, and Finance worked from the same view of pipeline-derived demand. Sales focused on maintaining milestone accuracy rather than emailing lists. Finance understood forecast adjustments and their impact on inventory and working capital. Expedite calls and late rescheduling eased as decisions shifted earlier in the cycle and were grounded in shared data.

What Changed for the Team

• Before: Project alerts arrived as ad hoc spreadsheets; After: CRM milestones translated automatically into governed, time-phased demand signals.
• Before: Planners guessed at timing and product mix; After: Stage-weighted rules and mappings placed demand by product family and location.
• Before: Forecast changes debated after the fact; After: AI-assisted anomaly flags with a clear approval screen and rationale capture.
• Before: Late purchase orders and premium freight; After: Earlier procurement decisions aligned to a combined plan.
• Before: Regional definitions varied; After: Standardized stage definitions and data contracts enforced at integration.
• Before: Limited visibility for Finance; After: Dashboards showing accepted adjustments and their impact on the plan.

Key Takeaways

• Translate CRM milestones into structured, time-phased demand; don’t rely on manual lists to bridge Sales and Planning.
• Keep core systems—CRM and planning—intact; add a thin layer that standardizes mappings, weights stages, and governs changes.
• Use AI-assisted anomaly detection to focus planner attention, but keep approval authority with humans.
• Define data contracts for products, locations, and stages so integrations resist drift across regions and teams.
• Start with volatile categories and run in shadow mode; only publish to the plan once stakeholders trust the proposals.
• Capture rationale and effective dates for rule changes so Finance and Operations can trace adjustments to decisions.

FAQ

What tools did this integrate with?
Salesforce Sales Cloud supplied opportunity stages, dates, and project attributes via the REST API. The planning system ingested external demand adjustments alongside the statistical forecast; the pattern aligned to platforms like SAP Integrated Business Planning. The anomaly reviewer used established time-series and outlier detection techniques; background reference: scikit-learn.

How did you handle quality control and governance?
Data contracts defined required CRM fields, stage definitions, and product/location identifiers. Validations blocked incomplete or mismatched records. Stage-weighting and mapping rules were versioned, with approvals and effective dates. The reviewer flagged deviations from seasonal curves, and planners approved or tuned adjustments with a recorded rationale. All publishes into the plan were logged with source opportunities and decisions.

How did you roll this out without disruption?
We ran in shadow mode first to compare proposals to historical outcomes and manual alerts. Early cycles focused on selected regions and product families. Planners kept manual spreadsheets visible for side-by-side checks. Publishing to the plan began only after stakeholders confirmed that rules and mappings behaved as intended. No core systems were replaced.

How were project milestones translated into demand?
Opportunities mapped to product families and ship-to nodes using product hierarchies and regional rules. Stage probabilities and milestone dates drove time-phased allocations with lead-time offsets. Where partial bills of material existed, the translation apportioned demand across components. The logic produced location-ready quantities that the planning system treated as external demand signals.

How did you avoid double counting and late changes?
Opportunity-to-demand links used stable identifiers, so updates replaced prior adjustments rather than stacking them. The integration invalidated adjustments when stages slipped or deals were lost, and re-timed expected quantities when milestones moved. The reviewer highlighted large changes for approval, and dashboards showed pending adjustments so Sales and Planning could reconcile differences quickly.