All Case Studies
Abstrabit Technologies
Simplifying operations as you scale
Case Study
Demand Intelligence · ERP + CRM + Email Fusion

$1.2M in Materials
Nobody Needed.
The Answer Was in the Emails.

A packaging manufacturer was sitting on $1.2M in dead inventory while running emergency rush orders monthly. The demand signals existed — in SAP, HubSpot, and sales emails — but nobody was reading them together. We built the system that did.

$1.2M
Dead inventory
→
$340K
72% reduced · 2 quarters
Three-source demand fusion
SAP ERPOrder history
+
HubSpot + EmailQualitative signals
+
Market DataCommodity pricing
→
Forecast Engine
→
✓ Procurement recs
72%
Dead inventory reduction — $1.2M → $340K
84%
Forecast accuracy — was 58%
$0
Lost orders to stockouts — was $600K+/yr
~$200K
Rush order savings projected annually
4 hrs
Forecast time — was 3 days manual
Industry
Packaging Mfg
Revenue
~$45M/yr
Employees
200+ · 2 sites
Customers
~340 active B2B
Engagement
7 weeks
ERP / CRM
SAP B1 · HubSpot
The Problem

Too Much of What They Didn't Need.
Not Enough of What They Did.

Two real incidents — same week — that defined the problem
📦
Food Brand Volume Drop
Client ordering 200K units/quarter for 2 years suddenly dropped to 80K. A sales call 3 months earlier mentioned a packaging redesign. Nobody flagged it. Materials were already ordered.
~$180K in stranded materials
⚡
Electronics Stockout
Client doubled order volume with 3 weeks' notice — a product launch. Couldn't fulfill. Order went to competitor. A sales rep email 2 months earlier mentioned "a big Q4 launch."
$340K lost to competitor
🖥️
SAP Business One
4 years order history. Pattern data existed — but procurement read it as "last quarter = next quarter."
✓ used — simplistically
💬
HubSpot CRM + Email
Sales reps heard about launches, redesigns, volume shifts. Completely disconnected from procurement.
✕ never used for procurement
📊
Market Data
Commodity prices and industry indices correlated with volumes. Nobody had connected it to forecasting.
✕ never used for procurement
🔍
"We're spending too much on materials we don't need and not enough on materials we do." — VP of Operations. The data existed. Nobody was reading it together.
$1.2M
Dead inventory
$340K
Single lost order
~$250K
Annual rush costs
3 days
Manual forecast time
The Solution — Architecture

Three Data Sources. One Fusion Engine.
Forecasts That Read the Emails.

🖥️
SAP Business One
ERP · 4 years order history
Time-series patterns per customer
Seasonality & ordering cycles
Configuration-level demand curves
💬
HubSpot + Email
CRM notes · sales rep emails
"Launching new SKU in Q3"
"Consolidating suppliers"
"Redesigning their packaging"
📊
Market Data
Commodity prices · industry indices
Containerboard price trends
Industry production indices
Consumer spending seasonality
↓   all three sources feed the fusion engine
⚙️   Demand Intelligence Fusion Engine
Quantitative Baseline
Time-series analysis on 4 years of SAP data. Seasonality, customer cycles, trend direction.
ARIMA · seasonality decomp
Qualitative Signal Extraction
AI reads HubSpot notes and emails. Tags forward-looking signals with confidence levels.
AI email scan · confidence-weighted
Market Correlation
Commodity prices and production indices correlated with historical order patterns.
commodity price correlation
↓   engine produces three outputs procurement can act on
📈
Three-Scenario Forecasts
Base, upside, downside per customer. Procurement hedging logic — not a single number, a strategy.
base · upside · downside
🔔
Real-Time Signal Alerts
When demand signals change, procurement is flagged immediately. Before materials ship.
pre-shipment flags
💬
Conversational Interface
Plain language queries against live forecast data. No dashboards to learn.
natural language · live data
Material-level aggregation — what procurement actually orders
Q2 Raw Material Forecast — 340 Customers
80% confidence intervals
MaterialBase qtyConfidence rangeConfidence
C-flute 32ECT corrugated 847 tons
92% · firm rec
B-flute 26ECT corrugated 312 tons
76% · order base
Process color inks (CMYK) 4,200 kg
81% · firm rec
⚠ Die plates — food segment 14 plates
44% · manual review
The Proof — What Procurement Now Sees

Ask in Plain Language. Get a Strategy.
The Email That Changed the Q2 Forecast.

Conversational demand interface — VP of Operations
demand-intelligence · procurement-chat ⚡ Live data
VP
What's our corrugated board exposure for Q2? Breakdown by grade.
🤖
Here's your Q2 corrugated exposure across all 340 customers:
Q2 Corrugated Board Forecast 80% confidence
Base
1,159t
C+B+E flute combined
Upside
+18%
if 3 launches confirm
Downside
−22%
if food drops hold
📋 Order 1,159t base now. Hold upside delta until Jan 12 confirmation window.
VP
If Hartfield drops 30%, what's our waste risk?
🤖
Running scenario — Hartfield Foods −30%:
Hartfield Foods −30% · Material Exposure
C-flute over-committed41t · ~$68K
Custom die plates3 plates · ~$12K
Total exposed~$89K at risk
Cancellable before delivery~$51K if actioned by Nov 22
Signal confidenceMedium · CRM note Nov 8
Recommend escalating to sales rep before Nov 22 cancellation window.
AI signal extraction — the email that changed Q2
📡   Demand Signal Alerts
3 active
Hartfield Foods — volume review flagged
"consolidating vendors · Q2 could look different" · $68K exposure
⚠ Confirm by Nov 22
2d ago
Vantage Electronics — order frequency below pattern
Expected reorder Nov 10 — not received. 18-day delay.
⚠ Outreach recommended
4d ago
Clearbrook Consumer — launch signal detected
"new product line Q2 · needs heavy-wall triple-wall corrugated"
📈 Upside scenario added
6d ago
Results — First Two Quarters

$860K Dead Inventory Eliminated.
Zero Stockout Losses. Procurement Made Proactive.

72%
Dead inventory reduced
$1.2M → $340K
84%
Forecast accuracy
was 58%
$0
Lost orders to stockouts
was $600K+ prior year
Dead Inventory Value
Before
$1.2M
After
$340K · −72%
Forecast Accuracy
Before
58%
After
84%
Rush Orders per Month
Before
2–3/mo
After
1 per 6–8wk
AI Signal Extraction Three-Scenario Forecasting Conversational Interface SAP Business One HubSpot CRM Time-Series Analysis Alert Trigger Engine Confidence Scoring
Metric
Before
After (2 qtrs)
Dead inventory
$1.2M
$340K (−72%)
Rush orders
2–3/month
1 per 6–8 wks
Rush order cost
~$250K/yr
<$50K projected
Forecast accuracy
~58%
~84%
Lost orders
$600K+/yr
$0 in 2 quarters
Forecast time
3 days manual
4 hours review
The Real Differentiator
Quantitative forecasting on order history was table stakes. The differentiator was reading the emails — extracting qualitative demand signals from CRM notes and sales conversations, and turning them into signed forecast adjustments with confidence scoring. That's what caught the Hartfield drop before materials shipped.