CONDITION INTELLIGENCE FOR FRESH RETAIL

Fresh produce changes every hour. Retail decisions should too.

Departmint is building toward combining retail signals with physical product-condition intelligence — to help fresh teams identify what needs attention, when to intervene, and whether the action creates economic value.

DECIDE HERE
Become a design partnerSee how it worksBUILT FOR FRESH RETAIL IN THE GCC
01 — THE DECISION WINDOW

The goal isn't to discount more. It's to intervene better.

Every batch of fresh produce has a window in which action still recovers value. Departmint is being built to help retailers find it.

TOO EARLY

Markdown healthy inventory

LOST MARGIN
Product is discounted that would have sold at full price. Waste falls — and so does gross margin. This failure never appears in a waste report.
THE DECISION WINDOW

Inspect or mark down at the right moment

RECOVER VALUE
Enough condition risk to justify acting, enough commercial life left for the action to work. This is the window Departmint is designed to find.
TOO LATE

Deterioration

WASTE
Condition passes the point where any discount clears the stock. The unit is written off and the margin is gone entirely.
02 — THE PROBLEM

Retail systems are excellent at what happened. Fresh needs a decision about what happens next.

Fresh produce is unusual inventory: its physical condition changes continuously, and its commercial value changes with it. Inventory and POS systems answer what arrived, what sold and what remains — with precision.

The harder question is continuous, and today it is usually answered by whoever is on shift: what should we do with what remains, and when? That decision is made periodically, manually, and with limited economic information behind it.

What arrivedSOLVED
What soldSOLVED
What remainsPARTIALLY
THE UNANSWERED QUESTION
What should we do with what remains — and when?
03 — WHAT THE SYSTEM CANNOT SEE

Products change before databases do.

HYPOTHESIS — NOT VALIDATED

A retailer's system can know that 52 boxes of strawberries remain. That doesn't necessarily mean it knows what condition those 52 boxes are actually in. Departmint is exploring the intelligence layer between physical inventory and the systems managing it.

THE DYNAMIC COMMERCIAL CLOCK
Packaged products often have an expiration date. Fresh produce doesn't.

Its commercial window changes with condition, environment, time and demand. A strawberry carries no printed date, and neither does an avocado, a tomato or most unpackaged produce. Departmint is exploring whether physical-condition signals can create a continuously updated commercial clock for perishable inventory.

NOT AN EXPIRATION DATE · NOT A SAFETY CERTIFICATION
AN ESTIMATED COMMERCIAL-LIFE SIGNAL, IF IT PROVES MEASURABLE
PHYSICAL REALITY + COMMERCIAL REALITY
PHYSICAL REALITY
What is happening to the product?
Condition
Quantity
Location
Visual anomalies
Temperature
Humidity
Time
Batch / arrival information
THE LAYER BETWEEN
DEPARTMINT
Condition is an input. The output stays a commercial decision.
CONDITION + INVENTORY + DEMAND
+ PRICE + TIME
COMMERCIAL REALITY
What is happening to demand?
Inventory
POS
Sales velocity
Price
Markdown history
Waste history
Expected demand
COMMERCIAL ACTION — ONE OF SIX
KEEP
Healthy inventory. No intervention.
PRIORITISE
Sell this inventory before newer or healthier inventory.
INSPECT
Human verification required.
SEPARATE
Potentially deteriorating inventory isolated from healthy inventory, where operationally appropriate.
MARKDOWN
Review whether a price intervention creates better economics.
REMOVE / REVIEW
Potential quality issue requiring staff action.

Same inventory. Different physical reality.

ILLUSTRATIVE SCENARIO
STORE A
Strawberries
Inventory52 boxes
PriceAED 18.95
Sales velocity11 /day
Physical conditionStable
RECOMMENDATION
HOLD
STORE B
Strawberries
Inventory52 boxes
PriceAED 18.95
Sales velocity11 /day
Physical conditionDeclining
RECOMMENDATION
INSPECT / REVIEW MARKDOWN

Traditional commercial data can make these two situations look similar. Physical-condition intelligence may reveal that they require different actions. Whether that difference is large enough to be worth measuring is exactly what the first pilot has to establish.

04 — HOW DEPARTMINT WORKS

From retail signals and product condition to a decision.

SEE — SIGNALS
Inventory
POS / sales velocity
Markdown history
Waste history
Product observations
Batch & arrival information
Temperature, where available
UNDERSTAND — DEPARTMINT
DEPARTMINT
Identifies which inventory may need attention, why, and how urgently — then states the action.
WHAT NEEDS ATTENTION?
WHY?
WHEN?
WHAT SHOULD I DO?
ACTION — ONE OF SIX
Keep
Prioritise
Inspect
Separate
Markdown
Remove / review
Then: measure the outcome, and feed it back.
EXECUTION PHILOSOPHY
Prove the decision.
Then automate it.

We are not building complicated AI first and looking for a problem afterwards. The order is deliberate: start from the retail decision and its economic outcome, learn the workflow and the data around it, and only then automate — because a model is only worth building once you know precisely what it must predict.

SEE → UNDERSTAND → DECIDE → ACT → LEARN
05 — PRODUCT VISION

One recommendation, with the reasoning attached.

PRODUCT VISION — SIMULATED DATA
DEPARTMINTFRESHFRESH OVERVIEW
DEMO STORE · SIMULATED DATA
Strawberries
Candidate category — selected on retailer evidence
INVENTORY
52 boxes
SALES VELOCITY
11 /day
CURRENT PRICE
AED 18.95
CONDITION TREND
Declining
INVENTORY EXPOSURE
High
SELL-THROUGH RISK
High
INTERVENTION WINDOW
ACT BY HERE
NOW+6H+12H+24H
RECOMMENDED ACTION
Review markdown now
AED 18.95AED 14.95
REASON
Expected sell-through is below inventory exposure within the current intervention window.
CONFIDENCEMEDIUM / HIGH
Approve
Inspect first
Dismiss
EVERY DECISION FEEDS THE NEXT ONE
The outcome is the product.

Whether staff accepted the recommendation, whether the product sold, how much margin was recovered, whether the intervention was unnecessary — each answer is a labelled example that makes the next recommendation better.

01Data
02Risk
03Recommendation
04Staff action
05Outcome
06Learning
06 — BUSINESS VALUE

A margin product with a sustainability consequence — in that order.

01

Protect margin

Avoid unnecessary or poorly timed markdowns. A discount given to product that would have sold at full price is a permanent margin loss that no waste report will ever show you.

02

Recover value

Intervene while the action still works — before inventory reaches the point where no price clears it and the entire margin is written off.

03

Reduce waste

Less avoidable disposal, as a consequence of better commercial decisions rather than as the goal itself. Waste falls because the decisions improved — not because everything was discounted.

07 — THE PILOT

Start with one decision.

No retailer needs to transform its fresh department on day one. The first deployment is deliberately narrow enough that the result can be attributed with confidence.

01
1 retailer
02
1 store
03
1 fresh category
04
1 measurable decision
01 — BASELINE
Understand the current process
Who decides, when, on what information, what action follows — and what happens economically.
02 — OBSERVE
Collect the relevant data
POS, inventory, markdown, waste, product observations, and temperature where it is available.
03 — RECOMMEND
A small number of clear actions
Inspect now. Prioritise sale. Review markdown. Continue monitoring.
04 — ACT
Staff decide
Your team keeps every decision. Whether they follow the recommendation is itself a measurement.
05 — MEASURE
Against your existing process
Your current workflow is the control. A result we cannot attribute is not a result.
WHAT THE PILOT COMPARES
A — CONTROL
The retailer’s current process
MANUAL INSPECTION EXISTING REPORTS EXISTING MARKDOWN RULES
B — COMMERCIAL DATA
Decision from retail signals alone
INVENTORY POS SALES VELOCITY HISTORY
C — CONDITION-AWARE
Commercial data + physical condition
EVERYTHING IN B + PHYSICAL-CONDITION OBSERVATION
Does C materially outperform A and B?

Measured on recovered margin, sell-through, waste, markdown cost, staff effort, false positives, operational burden and net economic value. If condition does not change the decision enough to pay for collecting it, that is a finding worth having before any sensing hardware is built.

CHOOSING THE CATEGORY
The category is selected on your evidence — not our assumption.

Berries are a plausible starting candidate, not a decision. We look for categories with meaningful shrink value, frequent markdowns, sufficient volume, real margin exposure, a workable intervention window, and outcomes that can actually be measured.

MEANINGFUL SHRINK VALUE
FREQUENT MARKDOWNS
SUFFICIENT VOLUME
HIGH MARGIN EXPOSURE
WORKABLE INTERVENTION WINDOW
MEASURABLE OUTCOMES
08 — HOW SUCCESS IS MEASURED

Lower waste is not enough.

Any retailer can reduce waste to nearly zero by discounting everything aggressively — and destroy the category doing it. The intervention has to beat the existing process on economics, not on tonnage.

01
Waste rate
02
Sell-through
03
Gross margin recovered
04
Markdown cost
05
Staff time
06
Recommendation adoption
07
False positives
08
Incremental operating cost
WHY FALSE POSITIVES MATTER
A recommendation can be technically correct and still destroy value.

Flagging healthy product for inspection or markdown costs staff time and margin every time it happens. So the question the engine has to answer is not "is this product deteriorating?" — it is the harder one:

Will acting now produce a better economic outcome than waiting?
09 — WHO IT'S FOR

Built for the people who own fresh performance.

PRIMARY

Head of Fresh

Owns overall fresh-category performance.
PRIMARY

Produce Director

Owns produce economics, availability, markdowns and waste.
PRIMARY

Fresh Category Manager

Owns category-level margin and sell-through.
PRIMARY

Retail Operations

Owns store execution and operational consistency.
SUPPORTING — FINANCE
Validates whether interventions actually recover incremental margin.
SUPPORTING — DIGITAL / INNOVATION
Supports systems integration and technology deployment.
SUPPORTING — SUSTAINABILITY
Supports food-waste objectives and reporting — as a beneficiary of the commercial case, not the basis of it.
10 — WHY GCC
GCC RETAIL FOOD WASTE, 2022
1.3Mt
Retail sector only
ANNUAL ECONOMIC LOSS
$4–7B
Corresponding to the above
UAE FOOD RETAIL, 2023
$40B
6.5% five-year CAGR
SAUDI FOOD RETAIL, 2023
$62B
4.2% five-year CAGR

These figures describe waste generated at retail level. They exclude household waste and upstream supply-chain loss, which are larger and are not the problem we address. Retail waste is the smaller share — but it is the share a retailer directly controls, books as shrink, and can act on within a single shift.

SOURCE — OLIVER WYMAN, “HOW TO SUCCESSFULLY REDUCE
RETAIL FOOD WASTE IN THE GCC”, MARCH 2025.
FIGURES AS PUBLISHED.

Built for the realities of GCC fresh retail.

Fresh produce can travel through multiple commercial stages before it reaches a GCC shelf. Departmint is being designed to help retailers understand what action makes sense once that inventory is in their hands.

Large, consolidated modern grocery networks
Substantial imported fresh-food flows
Complex supply chains across many fresh categories
Food-security priorities at national level
Government food-waste objectives
Rapid retail digitisation and sophisticated groups
A concentrated market is hostile to cold outbound sales and favourable to a company that earns one reference customer and expands inside it.
GEOGRAPHIC SEQUENCE — INTENDED, NOT ENTERED
01
UAE
Design partners and first pilots
02
Saudi Arabia
Enterprise expansion
03
Wider GCC
Qatar · Kuwait · Bahrain · Oman
04
Wider MENA / international
11 — BUSINESS MODEL

Built as enterprise B2B software.

01 — PILOT

Controlled retailer pilot

A narrow, instrumented deployment. Commercial terms agreed with the design partner.

02 — STORE DEPLOYMENT

Annual software subscription

Priced by store, module and category once the value is understood.

03 — ENTERPRISE

Multi-store chain agreement

With integrations, analytics, implementation and enterprise support alongside.

LAND AND EXPAND
One category
Multiple fresh categories
Multiple stores
Entire retail group
Other perishable departments
PRICING
Custom enterprise pricing.
Pilot programmes available for selected design partners.
STATED PLAINLY
Pricing has not yet been validated with a customer. We would rather arrive at it from what a pilot demonstrates than reverse-engineer it from a target.
12 — WHERE THIS GOES

The wedge is small. The platform is not.

ROADMAP — NOT BUILT TODAY
PHASE 1 — PROVE THE DECISION
Manual observations, retail data, simple recommendations, measured economics
PHASE 2 — PROVE THE CONDITION SIGNAL
Does physical condition materially improve the recommendation?
PHASE 3 — AUTOMATE OBSERVATION
Computer vision, environmental sensors, batch information, retail integrations
PHASE 4 — CONDITION INTELLIGENCE
Continuously estimate commercial-condition risk
PHASE 5 — AUTOMATED DECISION INTELLIGENCE
What needs attention, when, why, and which intervention to consider
PHASE 6 — MULTI-STORE INTELLIGENCE
Learn across stores, categories, origins, conditions, interventions and outcomes
FROM ONE DEPARTMENT TO ALL PERISHABLES
DEPARTMINTFRESH
Produce first — because it has the shortest windows, the most frequent decisions and the clearest measurement.
Bakery
Meat
Seafood
Dairy
Prepared
Nothing in the approach is produce-specific. Every perishable category is a quantity with a condition and a deadline.
FRESHNESS PASSPORT
LONG-TERM CONCEPT
Every batch could eventually carry its own commercial history.
ORIGINNETHERLANDS
ARRIVAL — UAE10 AUG
DCDUBAI
STORE ARRIVAL11 AUG
OBSERVED CONDITIONSTABLE
COMMERCIAL RISKMEDIUM
RECOMMENDED ACTIONMONITOR
Illustrative. This dataset does not exist today — it is what enough deployments could eventually build, connecting origin, transport, environment, store, condition, price, demand and outcome.
POTENTIAL DATA FLYWHEEL — NOT AN EXISTING MOAT
More observations. Better predictions. Better outcomes.

Each deployment could eventually relate product, origin, condition, time, temperature, price, demand, intervention, sell-through and waste. That relationship is the eventual defensibility — and it does not exist yet. We would rather call it a potential flywheel than a moat.

PRODUCTVARIETYORIGINBATCHSTORELOCATIONTIMETEMPERATUREHUMIDITYVISUAL CONDITIONDETERIORATIONINVENTORYPRICEDEMANDINTERVENTIONSELL-THROUGHWASTEECONOMIC OUTCOME
13 — FUTURE PRODUCT ARCHITECTURE

What if the shelf could understand what was sitting on it?

FUTURE PRODUCT VISION — NOT BUILT
CONCEPT — VISION MODULE
A camera positioned above a produce display
Potential observations, if the sensing proves reliable enough to matter:
Visual inventory
Quantity
Product identification
Visible condition
Visible anomalies
Shelf occupancy
OPTIONAL ENVIRONMENTAL SENSORS
TEMPERATURE · HUMIDITY
DEPARTMINT CONDITION ENGINE
Observations combined with commercial data
POS
Inventory
Price
Sales velocity
Markdown history
Waste history
OPERATIONAL RECOMMENDATION
One of six actions, with its reasoning
KEEP
PRIORITISE
INSPECT
SEPARATE
MARKDOWN
REMOVE / REVIEW
CONDITION WHERE NEEDED. TRACEABILITY WHERE AVAILABLE.
Not every perishable category needs the same technology.

Packaged products — dairy, packaged meat, juices, prepared foods — often already carry a barcode, a batch and a best-before date. A future Departmint system could combine that with shelf location, inventory, demand and pricing to improve stock rotation.

Expiry management and FEFO already exist in retail. Departmint's differentiation is not barcode scanning alone. This is a future extension, not the initial wedge.

ILLUSTRATIVE — FEFO ROTATION
Expires tomorrowFRONT / PRIORITISE
Expires in 4 daysNORMAL
Expires in 9 daysBACK
POTENTIAL PRODUCT ARCHITECTURE
NOT LAUNCHED PRODUCTS
DEPARTMINT VISION
Physical condition intelligence for condition-dependent perishables.
PRODUCE · BAKERY
OTHER APPROPRIATE FRESH CATEGORIES
DEPARTMINT TRACE
Expiry, batch and location intelligence for packaged perishables.
DAIRY · PACKAGED MEAT
PREPARED FOOD · JUICES
DEPARTMINT INTELLIGENCE
The decision layer both would feed. This is the part being validated first.
KEEP · PRIORITISE · INSPECT · SEPARATE · MARKDOWN · REMOVE / REVIEW

Computer vision is a potential enabling technology, not the product. A retailer does not buy fruit recognition — they buy better sell-through, more recovered margin, less avoidable waste and better store execution. Condition intelligence only matters here if it changes those numbers.

14 — LANDSCAPE

A real category, with a specific wedge inside it.

Companies including Smartway, Wasteless, Freshflow, OneThird, Clarifresh and Focal Systems are working on parts of this problem — dynamic markdown, demand forecasting, freshness sensing, shelf monitoring. Their existence is evidence that the category is real.

We are not claiming to be first, and we are not competing on a feature grid. Departmint is testing a specific intersection, starting in GCC fresh retail.

THE INTERSECTION WE'RE TESTING
Physical condition
Inventory
Sales velocity
Intervention timing
Economic outcome
15 — DESIGN PARTNERS

We're looking for GCC retail design partners.

We are currently speaking with fresh-category and retail-operations teams to validate the first Departmint workflow. The first pilot is intentionally narrow — one retailer, one store, one category, one measurable decision.

THE IDEAL DESIGN PARTNER
Supermarket or grocery retailer
UAE or Saudi Arabia
Meaningful fresh-produce operation
Willing to share relevant historical data
Interested in testing fresh-category economics
WHAT YOU GET
A controlled test on one category, measured against your own current process, reported in category margin rather than tonnes. You keep the data and the finding — whichever way it goes.
16 — WHERE WE ACTUALLY ARE

We are pre-seed and pre-product, and we would rather you heard it here.

NOT CLAIMED ANYWHERE ON THIS PAGE
No customers, pilots or letters of intent
No revenue or ARR
No accuracy or computer-vision performance
No waste-reduction or margin results
No proprietary dataset
No store deployments
No validated pricing
WHAT IS TRUE TODAY

A solo founder — an industrialization and design engineer with a background in complex physical systems, product development and entrepreneurship — building at the intersection of AI and retail operations.

CV/ML engineering and fresh-retail operating expertise are planned as key hires and advisors immediately after initial customer validation. Hiring them before knowing which decision creates value would be building a capability in search of a problem.

Every dashboard on this page is simulated. Every forward-looking capability is labelled as roadmap, concept or hypothesis.

WHERE EACH CAPABILITY ACTUALLY STANDS
VALIDATING NOW
INSPECTION TIMING MARKDOWN TIMING ECONOMIC INTERVENTION CONDITION RELEVANCE
NEXT
STRUCTURED CONDITION CAPTURE RETAIL INTEGRATIONS DECISION MODELS
PLANNED
COMPUTER VISION ENVIRONMENTAL SENSING AUTOMATED CONDITION ASSESSMENT
LONG TERM
COMMERCIAL-LIFE MODELS MULTI-STORE LEARNING FRESHNESS PASSPORT PERISHABLE INTELLIGENCE PLATFORM
17 — INVESTORS & STRATEGIC PARTNERS

Building toward the condition intelligence layer for perishable retail.

Departmint is at pre-seed stage and building its first GCC retail design partnerships. If you invest in retail technology, food technology, enterprise AI, waste reduction or GCC technology, we are happy to share the full thesis and pilot design.

RETAILTECHFOODTECHENTERPRISE AIWASTE REDUCTIONGCC TECHNOLOGY
Every perishable product has a commercial clock. Departmint is being built to help retailers read it before value disappears.