AI Retail Playbook: Lessons from Itochu’s Profit Surge

Japan’s Itochu didn’t hit a record nine-month profit by “getting lucky” with commodities. It did it the boring way: by making consumer businesses (food, textile, convenience retail) run predictably and profitably.

On 6 Feb 2026, Reuters reported that Itochu’s nine-month net profit (to 31 Dec) rose 4% to 705.3 billion yen, a record for the period. The standout detail wasn’t just the headline number—it was where it came from. Food division profit jumped 38% to 82.5 billion yen, helped by higher banana production and sales, and growth in packaged foods at its Dole agriculture unit. Even as resource sectors slowed, Itochu’s CFO said non-resource “core profits” hit a record, and the company doubled down on shareholder returns with an additional share buyback of up to 20 billion yen, on top of a previously announced 150 billion yen programme.

For Singapore retailers and e-commerce teams reading this as part of our “AI dalam Peruncitan dan E-Dagang” series, the signal is clear: consumer businesses win when demand planning, inventory, pricing, and merchandising are treated as data problems—then solved with analytics and AI. You don’t need to be a Japanese sogo shosha to apply the same playbook.

What Itochu’s results really say about consumer profitability

Itochu’s numbers point to a simple truth: the most defensible profits in retail and consumer goods come from execution, not hype. When “food and textile” offset a resource slowdown, it’s because everyday categories can generate steady margin—if you control supply, predict demand, and react faster than competitors.

The non-resource advantage: predictable demand beats cyclical swings

Resource profits swing with prices. Consumer profits swing with operations.

Itochu highlighted that consumer strength offset weaker metals and delayed turnaround plans at U.S. and Australian coking coal mines. That contrast matters. In retail and FMCG, your “commodity” isn’t coal—it’s:

• shelf availability and fulfilment speed
• markdown discipline
• basket size and repeat purchase
• shrink, waste, and returns

Those are measurable. And that’s exactly why AI adoption in retail is accelerating: it targets repeatable levers.

Why the 38% food division jump is a data story

Bananas and packaged foods may sound old-school, but they’re highly optimisable. Agriculture and packaged foods both depend on forecast accuracy and yield planning. Small improvements compound quickly:

• If you forecast demand better, you buy/produce closer to reality.
• If you improve allocation, you sell more at full price and waste less.
• If you detect regional shifts early, you avoid stockouts and emergency shipments.

Even a 1–3% improvement in forecast accuracy can materially change waste and availability in perishables—especially in Singapore, where cold-chain costs and limited backroom space punish mistakes.

How AI translates “consumer strength” into profit (without buzzwords)

AI in retail isn’t one system. It’s a stack of decisions—some automated, some assisted—made faster and more consistently than humans can manage at scale.

1) Demand forecasting you can actually act on

Answer first: AI demand forecasting is valuable when it drives ordering, allocation, and staffing—weekly, not quarterly.

Most companies get this wrong by treating forecasting as a report, not an operating system.

What works in practice:

• Use hierarchical forecasting (SKU → category → store/region) so teams can drill down without losing the big picture.
• Add causal signals beyond sales history: promotions, paydays, holidays, weather, lead times, competitor pricing, marketplace events.
• Build forecasts around decisions: order quantities, replenishment cadence, and safety stock targets.

Singapore angle: planning for seasonal demand spikes around Chinese New Year, Hari Raya, Deepavali, year-end gifting, and 9.9–12.12 mega-sales isn’t optional. AI helps you model the spikes rather than “guess and pray”.

2) Inventory optimisation: fewer stockouts, fewer dead piles

Answer first: inventory optimisation is the quickest way to protect margin because it reduces both lost sales (stockouts) and forced discounting (overstock).

Itochu’s consumer businesses likely benefit from disciplined supply allocation and inventory turns—especially in convenience retail where assortment is wide but space is tight.

A practical AI approach retailers in Singapore can adopt:

1. Segment SKUs by volatility and margin (A/B/C plus “fast/slow movers”).
2. Set service-level targets per segment (e.g., 98% for essentials, 92% for long-tail).
3. Optimise reorder points using lead time variability and demand uncertainty.
4. Add exception alerts: “this SKU will stock out in 5 days at Store 18.”

The reality? Your team doesn’t need 100 dashboards. It needs 10 alerts that are correct.

3) Personalised recommendations that don’t feel creepy

Answer first: recommendations increase revenue when they mirror shopper intent (mission-based baskets), not when they push random cross-sells.

Convenience retail and packaged foods are perfect for this because baskets are frequent and patterns are strong.

Recommendation systems that work for omnichannel retail:

• “Complete the basket” prompts (noodles → eggs → chilli sauce)
• replenishment reminders based on typical cadence
• personalised bundles (“weekday lunch set”) instead of item spam

In e-commerce, recommendation quality shows up in:

• higher average order value (AOV)
• higher conversion rate
• better repeat purchase rate

But you only keep those gains if inventory is aligned—otherwise you recommend items you can’t deliver.

4) Pricing and promotions: stop buying revenue with discounts

Answer first: AI-driven pricing works when it protects margin by predicting promo lift and cannibalisation.

Many retailers run promotions on gut feel. The result is predictable: sales go up, profit doesn’t.

What AI can do better:

• estimate incremental lift vs “customers would’ve bought anyway”
• detect cannibalisation (promo SKU steals from full-price sibling SKU)
• optimise promo depth and duration

If you sell packaged foods, toiletries, apparel basics, or household goods in Singapore, this is the difference between a campaign that looks good in GMV and one that actually improves contribution margin.

A Singapore-ready checklist: applying Itochu’s playbook in 90 days

Itochu can announce buybacks because it trusts the durability of its earnings. For most SMEs and mid-market retailers, the equivalent is simpler: get to predictable weekly profit drivers.

Here’s a 90-day plan I’ve seen work because it’s operational, not theoretical.

Days 1–30: Fix the data pipeline (no, you can’t skip this)

Answer first: if your product, inventory, and sales data don’t match, every AI output will be wrong.

Minimum viable foundation:

• a clean product master (SKU, variant, cost, supplier, lead time)
• unified sales by channel (POS, Shopify, marketplaces)
• inventory snapshots (store + warehouse) daily
• promotion calendar and price history

Deliverable: one “source of truth” dataset your team trusts.

Days 31–60: Build forecasting + replenishment decisions

Answer first: start with your top 20% SKUs that drive 80% of revenue and operational pain.

Pilot scope (keep it tight):

• 2 categories (e.g., packaged foods + health/beauty)
• 1–2 fulfilment nodes (one warehouse, one store cluster)
• weekly forecast and reorder recommendations

Track these KPIs weekly:

• stockout rate (% of time out of stock)
• inventory turns
• waste/expiry (if perishable)
• gross margin after markdowns

Days 61–90: Add customer-facing AI (recommendations + targeting)

Answer first: only scale personalisation after inventory is stable—otherwise you create demand you can’t fulfil.

Practical wins:

• personalised email/WhatsApp segments (replenish, new arrivals, bundles)
• onsite recommendations for top missions
• post-purchase cross-sell flows tied to actual stock

If you’re running e-commerce in Singapore, this is where “AI marketing” becomes real: fewer blanket campaigns, more relevant messages, and lower acquisition waste.

One-liner worth pinning: AI in retail isn’t magic—it’s discipline you can measure.

People also ask: does AI help even if you’re not Itochu-sized?

Yes—often more.

“We’re small. Do we have enough data?”

If you have 12–18 months of transaction history and stable SKU definitions, you have enough to start with forecasting and segmentation. For newer brands, you can use external signals (ad spend, traffic, marketplace trends) to supplement.

“Will AI replace my merchandisers and planners?”

No. It replaces spreadsheets and “manual copy-paste planning.” The best teams use AI to:

• automate routine decisions
• surface exceptions
• test promo ideas with predicted outcomes

“What’s the fastest ROI use case?”

Inventory optimisation usually pays back first because it reduces:

• lost sales from stockouts
• capital trapped in slow movers
• margin erosion from heavy markdowns

A final take for Singapore retailers watching Japan

Itochu’s record nine-month profit is a reminder that consumer business strength is engineered, not hoped for. The company’s food and textile performance—plus its confidence to return capital—points to systems that keep demand, supply, and customer behaviour aligned.

In this AI dalam Peruncitan dan E-Dagang series, we keep coming back to the same theme: AI is most valuable when it’s tied to a specific operating decision—reorder, allocate, price, recommend, or target.

If you’re running retail or e-commerce in Singapore, the next step isn’t buying “an AI tool” and calling it transformation. It’s choosing one profit lever (forecasting, inventory, promos, or personalisation), instrumenting it properly, and running weekly improvements until the numbers move.

Landing page/source URL: https://www.channelnewsasia.com/business/japans-itochu-posts-record-9-month-profit-consumer-business-strength-5912401