Mastering Demand Forecasting and Navigating Amazon’s Vendor Central Order Volatility

At Balkan eCommerce Summit 2026, Bruno Ferreira, founder of BlueDot eCommerce, explored one of the most complex operational challenges in modern eCommerce:

demand forecasting and managing Amazon Vendor Central order volatility.

The presentation focused on:

• how Amazon Vendor Central forecasting actually works
• how brands should interpret Amazon’s demand signals
• the relationship between commercial and supply chain teams
• and how companies can reduce forecasting errors and operational friction.

Understanding Amazon Vendor Central

Bruno began by explaining the difference between:

• Amazon Seller Central
• and Amazon Vendor Central.

While Seller Central allows brands to sell directly to customers,
Vendor Central works differently:

Amazon itself becomes the buyer.

In this model:

• Amazon purchases stock directly from brands
• predicts future customer demand
• and automatically generates purchase orders.

According to Bruno:

one of the biggest misconceptions is believing that Amazon’s forecast equals future orders.

In reality:
Amazon provides forecasts based on:

• expected customer demand
• clicks
• traffic
• and sales probability

but:

brands still need to calculate what Amazon will actually order.

Forecasting Is Probability, Not Certainty

A core theme throughout the presentation was:

forecasting is never a guarantee.

Bruno described forecasting as:

a statistical model based on probability and confidence levels.

Amazon Vendor Central provides forecasting visibility:

• 26 to 48 weeks ahead
• depending on the category

but these numbers represent:

expected sell-through, not confirmed purchase quantities.

He emphasized that:
brands must learn to:

• interpret the data
• adjust expectations
• and build internal forecasting models around probability ranges rather than exact numbers.

Understanding P70, P80 and P90 Forecasting Levels

One of the most practical sections of the lecture focused on:

confidence levels in forecasting.

Bruno explained the meaning of:

• P70
• P80
• P90

These represent:

the probability that sales volume will reach or stay below a certain number.

Example:

• P70 = 70% confidence
• P80 = 80% confidence
• P90 = 90% confidence

His recommendation was:

• use P70 for standard catalog products
• use P80 for “hero ASINs” with stable and predictable demand
• avoid relying too heavily on P90 because it removes operational flexibility.

The overall advice was:

forecasting should aim for balance, not maximum optimism.

The Conflict Between Commercial Teams and Supply Chain Teams

A major part of the presentation focused on:

internal organizational alignment.

Bruno explained that:
commercial teams and supply chain teams often interpret forecasts very differently.

Commercial teams focus on:

• promotions
• pricing
• growth
• retail media
• product launches
• maximizing sales

Supply chain teams focus on:

• warehouse capacity
• stock availability
• lead times
• packaging
• fulfillment risks
• operational stability

This creates tension when forecasts suddenly change.

Example:

• a forecast initially predicts 1,000 units
• later updated to 3,000 units

For the supply chain team, this creates immediate operational pressure:

• Is the inventory available?
• Can production scale fast enough?
• Can the warehouse handle the order?
• Is packaging prepared correctly?

Bruno stressed that:

forecasting is not only a sales exercise – it is a cross-functional operational process.

Forecasting Errors Start in the Warehouse

Another important insight was:

forecasting problems often begin long before the customer places an order.

Bruno highlighted that:
many brands:

• overproduce
• overstock
• or incorrectly estimate procurement needs

because they fail to align:

• demand forecasts
• warehouse realities
• supplier lead times
• and Amazon’s operational behavior.

He explained that:
some products require:

six to seven months of production and replenishment lead time.

This means:
small forecasting errors today can create:

• inventory problems
• excess stock
• or shortages months later.

Amazon’s Fulfillment Logic Creates Volatility

Bruno also explained that:
Amazon’s ordering behavior is heavily influenced by:

warehouse and fulfillment center capacity.

Amazon does not order products simply because demand exists.
It also considers:

• available shelf space
• fulfillment center logistics
• shipping optimization
• carton sizes
• and operational efficiency.

One important trend mentioned was:

Amazon increasingly prefers shipping-ready packaging and master carton optimization.

Brands that align better with Amazon’s operational model:

• reduce friction
• improve reliability
• and improve forecasting consistency.

Weekly Forecasting Cadence Improves Accuracy

A key operational recommendation from Bruno was:

forecast weekly, not occasionally.

Using a weekly cadence allows teams to:

• identify forecasting errors faster
• adjust models continuously
• learn from historical deviations
• improve prediction accuracy over time

He explained that:
if teams consistently see:

a 20% gap between forecasted demand and actual orders,

they can gradually refine future models accordingly.

The message was clear:

forecasting is a continuous learning process, not a one-time calculation.

Forecasting Must Be Practical, Not Perfect

Bruno emphasized that:

operationally balanced forecasting is more valuable than theoretically perfect forecasting.

He demonstrated this through examples comparing:

• aggressive forecasting
• versus balanced forecasting.

The goal is not:

• maximum sales projections
• or unrealistic optimism

but rather:

creating reliable communication between commercial and operational teams.

The best forecasting models are the ones that:

• reduce surprises
• stabilize operations
• and improve long-term supply reliability.

Key Takeaways

The lecture demonstrated that:
forecasting in eCommerce is fundamentally based on probability, not certainty
Amazon Vendor Central forecasts represent expected customer demand, not guaranteed orders
P70 and P80 confidence levels provide more balanced operational planning
commercial and supply chain teams often operate with conflicting priorities
weekly forecasting processes significantly improve long-term accuracy
inventory and forecasting problems usually begin far earlier than most companies realize
Amazon ordering behavior is heavily influenced by fulfillment center logistics and operational efficiency
forecasting should balance growth ambitions with operational reality
continuous adjustment and historical learning are essential for improving forecasting accuracy

The overall conclusion was:

successful eCommerce forecasting is not about predicting the future perfectly – it is about building operational systems that can adapt intelligently to uncertainty and volatility.

The presentation is not available for sharing.