We're an auto glass distributor in Delaware. We carry over 4,100 vehicle-specific SKUs from manufacturers in China, Mexico, and the US — with lead times ranging from 3 days to 130 days. We built Tru-Stock AI because every inventory planning tool on the market was designed for retail or manufacturing, and none of them understood our business.
3,889
Products Tracked
$190K
Idle Capital Identified
Nightly
AI Analysis Cycle
3–130
Day Lead Time Range
Auto glass distribution is a different animal from general wholesale. Every single product in our catalog is vehicle-specific. A windshield for a 2022 Honda CR-V does not fit a 2022 Toyota RAV4. They look similar sitting on a rack, but they're completely different parts, each identified by a unique NAGS number. There's no "close enough" — if you don't have the exact part, you lose the sale.
We were running our operation on InFlow Inventory as our ERP, supplemented by spreadsheets and gut feel. Our purchasing manager spent hours every week manually reviewing SKUs — scrolling through reports, checking stock levels one by one, trying to remember which parts were trending up and which were sitting idle. The min/max levels in InFlow were set once and basically never updated. A part that sold heavily two years ago might have completely different demand today because of vehicle model year turnover, but the reorder triggers hadn't changed.
Wildly different lead times across vendors
XYG ships from China — 90 to 130 days. CMX and Vitro come from Mexico — 15 to 45 days. NMC and Nielsen are domestic — 3 to 15 days. Pilkington is domestic OEM. You can't use a single reorder formula when your supplier lead times vary by a factor of 40.
Seasonal demand swings
Winter is peak season — road debris, ice damage, and defroster cracks drive 30 to 50 percent higher volume from November through March. Static reorder points that worked fine in July leave you stocked out by December.
Kit dependencies
Every windshield installation requires companion parts — moldings, clips, urethane adhesive. If you have the glass in stock but you're out of clips, the installer can't finish the job. You haven't just lost the clip sale — you've lost the entire ticket.
Backorder infill decisions
When your primary manufacturer is backordered on a fast-moving part, you need an alternate source mapped and ready. Waiting for China to restock when you could bridge with a domestic supplier at higher cost still beats losing the customer.
We looked at every inventory planning tool we could find. They all had the same problem — they were built for retail (predict how many t-shirts you'll sell) or manufacturing (BOM planning and production scheduling). None of them understood that our lead times span from 3 days to 4 months, that our demand is seasonal and vehicle-specific, or that a missing 50-cent clip can kill a $400 glass sale.
So we built it ourselves. Tru-Stock AI connects directly to InFlow Inventory and syncs our entire catalog — products, stock levels, sales history, and purchase orders — every night at 1 AM. By 2 AM, the AI has analyzed every single SKU and generated fresh reorder suggestions. When our purchasing team arrives in the morning, the work is done. They see a prioritized list of what to order, how much, and from which vendor.
We track lead times from actual purchase order goods-received dates — when the product physically arrived, not when the vendor said it would. Over time, the system builds a real performance profile for each vendor, including variability. XYG might quote 90 days but actually deliver anywhere from 85 to 135. That variability gets factored into safety stock calculations.
The system uses a 90-day half-life on sales data — last month's sales count more than sales from six months ago. This means seasonal ramps get picked up naturally. When windshield replacements start climbing in October, the reorder suggestions adjust before November hits, not after you've already stocked out.
We use the King formula for safety stock, which accounts for both demand variability and lead time variability. Top-moving SKUs get 95% service level protection. The rest get 85%. This isn't a flat "keep 2 weeks extra" — it's a statistical calculation that adapts to each product's actual behavior.
The system maps companion parts to every glass product — the moldings, clips, and adhesives that complete an installation. If you're about to reorder 50 windshields, it flags whether you have enough companion parts to actually use them. Think of it as the nut-and-bolt problem: 100 bolts plus 80 nuts means you can only finish 80 jobs.
How the nightly cycle works
InFlow sync pulls latest products, stock levels, sales orders, and purchase orders. Incremental sync grabs the last 72 hours of changes. Full catalog syncs in under a minute for our 4,100 SKUs.
AI batch analysis kicks off. Every active SKU with inventory or sales history gets analyzed — demand velocity, safety stock, reorder point, risk classification, and vendor-specific order suggestions.
Purchasing team opens Tru-Stock and sees prioritized reorder suggestions organized by vendor, with quantities, urgency flags, and companion part alerts ready to go.
These are real numbers from our live dashboard — not projections, not estimates. After connecting Tru-Stock AI to our InFlow ERP, here's what the system surfaced across our 3,889-product catalog:
Live Inventory Health Snapshot
353
Healthy SKUs
$82,725 — properly stocked
1,209
At-Risk SKUs
$237,381 — trending toward stockout
431
Overstocked SKUs
$132,822 idle capital
551
Dead Stock SKUs
$58,021 trapped in non-moving parts
281
Out of Stock
Completely depleted — sales at risk
$190,843
Total Idle Capital
Overstock + dead stock combined
336 items flagged for immediate attention
Before Tru-Stock, none of these numbers were visible. We had a vague sense that some parts weren't moving and others were running low, but no way to quantify it across 3,889 SKUs. Now every product has a health classification updated nightly, with dollar values attached.
We found $132,822 sitting in overstock and another $58,021 trapped in dead stock — parts that haven't sold in months. That's nearly $191K that could be redeployed into fast-moving inventory. Before Tru-Stock, this capital was invisible, buried across thousands of SKUs.
Over 1,200 SKUs worth $237,381 were trending toward stockout. The system flagged them with enough lead time to reorder before they went to zero. Without this visibility, we'd discover these gaps only when a customer called and we didn't have the part.
551 products with $58,021 in cost value have zero sales velocity. These are parts we can liquidate, return to vendors, or mark down — freeing up warehouse space and cash. Liquidating dead stock alone could fund hundreds of reorder cycles for fast-moving parts.
The purchasing review that used to consume most of a day — pulling reports, cross-referencing spreadsheets, checking stock levels manually — now starts with a prioritized list that's already been calculated overnight. The team reviews and approves instead of researching and guessing.
We now have actual performance data on every vendor. When XYG quotes 90 days but consistently delivers in 110, the system knows that and calculates safety stock accordingly. No more stockouts because we trusted a vendor's optimistic estimate.
Kit tracking shows when we have glass on the shelf but are short on the moldings or clips needed to install it. A missing 50-cent clip can kill a $400 glass sale — now we see those gaps before the installer calls.
The honest summary
We didn't build Tru-Stock AI because we wanted to start a software company. We built it because we were drowning in spreadsheets and losing sales to stockouts while simultaneously tying up $190K in dead and overstocked inventory we didn't even know about. The same tool we built for ourselves is now available to other distributors who face the same problem — whether you sell auto glass, HVAC parts, plumbing supplies, or any other distribution vertical where lead times are long, demand is variable, and getting the order wrong costs real money.
For distributors who want to understand the math, here's what powers the recommendations:
SS = Z x sqrt(LT x sigma_d^2 + d^2 x sigma_LT^2)
The King formula. Z is the service level factor (1.65 for top SKUs, 1.04 for the rest). Accounts for both how much demand varies and how much your vendor's delivery time varies.
90-day half-life exponential decay
Recent sales weigh more heavily than old sales. A sale from last week influences the forecast more than a sale from six months ago. Seasonal shifts get captured naturally without manual adjustment.
IQR outlier filter + median
Outlier POs are filtered using interquartile range analysis. For vendors with fewer than 10 POs, we use the median. For vendors with more history, we use a weighted average. No single bad shipment skews the whole calculation.
Full methodology details available on our methodology page.
This morning our business looked like it was failing. By lunch, it wasn't.
For over a year, a silent mismatch between our inventory system and QuickBooks was stacking up fake losses. Every night. Automatically. Without anyone noticing.
$441,000 in false adjustments. Gross profit of negative $337,000. Equity deeply negative.
On paper we looked broken. In reality we were growing. Once we found it, we fixed it in a single morning.
Gross profit swung to +$275,000
Net income went from -$719,000 to +$65,000
Equity flipped from -$523,000 to +$261,000
Same business. Same team. Same revenue. Just clean data.
The lesson? Your inventory isn't just a warehouse problem. It's a financial problem. A decisions problem. A growth problem. If your books don't feel right — trust that feeling. The numbers are trying to tell you something.
Brian H.
Distribution Company Owner
If you're managing inventory with spreadsheets and gut feel, you're solving the same problem we had. Upload your data and see what AI-powered inventory planning looks like for your catalog.