Sell More at Full Price.
Hold Less of the Stock.
Agile Replenishment Simplified
Merchandise planning, store replenishment and AI style replacement for fashion retail - one continuous loop, built on the Databricks Lakehouse. We enable organisations to refill winners, swap dead options for live lookalikes, pull back slow movers. Implemented with a leading Indian menswear retailer - ERP-integrated replenishment into Microsoft Dynamics AX
of inventory is sold in end-of-season sale — after only ~45 days of full-price shelf life
design-to-shelf lead time, committed against a selling season of ~12 weeks
size-set breakage on the floor — core sizes gone, fringe sizes left for markdown
of brands still buy to long-range forecasts — for styles with no sales history
of in-season fixes — transfers, restocks — are manual, planner by planner
of stock carried by Indian apparel retailers — industry days-on-hand hit an all-time high in 2024, +14% vs pre-2020
The season is decided before it starts. Then the customer decides otherwise.
Your OTB is largely spent and the option plan locked 90–120 days before the range hits the floor. Once selling starts, the plan is rarely re-cut. When live sell-through diverges from it, and it always does, the losses are structural, not occasional.
Buy committed
Range, options, depth and ratio packs signed off on a forecast for styles that haven't sold a unit.
Range pushed out
Initial allocation goes by store grade and gut feel, so the DC empties into doors that may never sell it.
Winners break
Bestsellers lose core sizes in A-grade doors while replen waits for next week's WSSI review.
Slow lines clog
Dead options hold fixture slots and cover climbs, while fresh newness has nowhere to go.
Markdown clears it
EOSS discounts what never sold. The winners sold out at full price weeks ago.
Your POS already generates the sell-through signal every day. What's missing is a way to act on it at store × option speed, before the range sells down and the season window closes.
Why your ERP, WSSI and forecast-led push planning fall short in-season
Fashion demand at store × option × size is too noisy to forecast well, and the further from the customer you place stock, the harder it is to move. Theory of Constraints flips the question: stop pushing on a forecast, and pull on what actually sold.
Why push planning falls short
- Forecasts each style before it has sold a unit, then pushes most of the buy to stores on day one
- Allocates by store grade, so A-doors stock out while C-doors sit on weeks of cover
- Reads a broken size or empty fixture as "no demand", so a winner's rate of sale looks weaker than it is
- Re-plans weekly at chain level, always a week behind what each door actually sold
- Blankets every store with safety stock, tying up OTB and floor space in the wrong places
- ERP records what happened; it was never built to decide tonight's store-option moves
The Velo advantage: TOC replenishment
- Replenishes to actual consumption: what sold today is what gets refilled tonight
- Holds reserve depth at the DC, where variability pools, and releases it only where demand proves out
- Reads true rate of sale by excluding the days a size was genuinely out of stock, so a bestseller never looks like a slow mover
- Subordinates every call to your option plan, fashion-bucket min/max and fixture capacity
- Replenish, Replace or Remove: refills winners, swaps dead options for live lookalikes, pulls back slow movers
- Gives planners a ranked, explained exception list in place of a spreadsheet
Have more questions?
Explore FAQ sectionOne planning layer, from pre-season assortment to in-season replenishment
Simbus Velo sits on top of your ERP and WMS. It doesn't replace them. Every night it turns the day's sell-through into store × option calls your buying, merchandising and allocation teams can trust.
Pre-season planning
- Store–SKU affinity model, stock-out corrected
- Constrained fill: bucket min/max, planogram
- AI-decided store vs reserve split
- Size-depth norms by bucket × grade
- Non-pivotal ratio caps & override guardrails
- Mid-season plan versioning & re-validation
In-season execution
- 3-level triggers: options, pivotal sizes, slots
- Replenish → Replace → Remove decision tree
- Warehouse-first, then inter-store transfers
- Non-pivotal sizes in replenish & replace
- Pull-back orders on slow movers (60-day rule)
- ERP order-ID handshake & auto-cancel
AI under the hood
- Visual Replace: image embeddings + vector search
- Match blends look, category, colour, fit, price
- Every Replace explained in plain English (agents)
- Databricks Genie — ask-in-English insights for management
- End-to-end on the Databricks Lakehouse
Built for fashion and lifestyle retailers who run their own stores
Whether you run 30, 100 or several hundred exclusive brand outlets, wherever a range is bought ahead, phased by season and refreshed with drops, the same two problems bite: markdowns on what didn't sell, and lost sales on what did.
Apparel & fashion
Seasonal ranges and frequent drops; core styles stock out mid-week while fashion lines pile up.
Footwear
Deep size runs, and blanket safety stock tying up cash across every door.
Sportswear & athleisure
Never-out-of-stock core running alongside seasonal colourway launches.
Accessories & bags
Wide option counts, shallow depth and constant newness churn.
Replenish to consumption, not to forecast
The Velo approach — Theory of Constraints
Pull, not push
Depth stays at the warehouse — the point of highest aggregation. Stores are seeded light and replenished daily from actual sales.
Dynamic buffer management
Penetration decides what gets ordered. When warehouse stock is short, store grade and predicted sell-through decide who gets it first. Thresholds are configurable by store grade × fashion bucket — never hardcoded.
A fashion bucket = a look
Velo plans and protects option count and size integrity per look — not just total units.
Configuration, not code
Buffer zones, grade triggers, depth norms, planogram and style holds are all configurable — every change captured with a reason and an audit trail.
Pull-back is a separate discipline. Stock that has sat 60 days with almost no sales is proposed for return or hold, so capital is not stranded on the floor.
Buffer per store × fashion bucket
options on the floor vs the option norm for that look
- Green · 67%+healthy — monitor
- Yellow · 33–67%replenish per grade rule
- Red · 5–33%top of the order queue
- Black · 0–5%the look is off the floor
A stores: any sale · B: at Yellow · C: at Red. Zones, boundaries and grade rules are configuration, not product constants — shown here is our reference client's setup.
Executive scorecard —Throughput: full-price sell-through upInventory: working capital downOpEx: planner effort & markdowns down
One continuous loop: Assort › Replenish › Replace
Pre-season
Buy plan & OTB intake
season budget, options, bucket mix
Constrained assortment
bucket min/max · planogram · size depth
Initial fill
AI split: store vs warehouse reserve
In-season · Daily loop
Buffer check
daily, per store–SKU–size
Replenish
warehouse first, then inter-store
Replace
AI visual match for sold-outs
Remove / hold
pull-back orders · 60-day rule
ERP write-back
order-ID handshake · exceptions
Repeats nightly. Orders are in your ERP before stores open — decisions land in role-based dashboards, with Databricks Genie for ask-in-English insights.
More of the season sold at full price, less of it cleared in the sale.
Velo moves the merchandising KPIs your leadership already reviews every week, from full-price sell-through and markdown spend to in-stock and stock carrying cost.
Velo is built by our Databricks practice
The same team that builds production data platforms, ML pipelines and AI agents on the Databricks Lakehouse for enterprises. Here's some of what they're writing about.
On-Demand State Repartitioning in Databricks Streaming
Databricks now lets you resize stateful streaming partitions without losing checkpoint data. Learn what it means and how Simbus can help you adopt it.
Building AI Agents on Databricks: Tips for 2026
Practical tips for building AI agents on Databricks in 2026 - key tools, considerations, and where to start.
5 Powerful AI Supply Chain Wins Kinaxis and Databricks
How Kinaxis Maestro and the Databricks Data Intelligence Platform are transforming demand forecasting, inventory optimization, and supplier risk management together.
Stop clearing your bestsellers' margin in the sale.
Start with a free sell-through and availability health check on a sample of your stores. We'll show you where full-price sales are walking out and where OTB is stuck, using your own data.