Decision Intelligence for Amazon Marketplace
The consolidating publication: why marketplace automation erodes margin while working as designed, and what a decision layer must contain to replace it.
Download the paper →Elias Sultan
Marketplace Decision Systems ResearcherFounder of SellerFlow · creator of the Decision Intelligence Framework for Amazon Marketplace
Researching intelligent decision systems for digital marketplaces.
Specializing in Amazon Marketplace pricing, competitive behaviour and commercial decision frameworks.
Portrait slot
images/portrait.jpg
SignalCompetitor offers, Buy Box, delivery
InterpretationBehaviour, position, intent
ConstraintCost, margin, inventory, velocity
DecisionAct, hold or withdraw
PriceThe output, not the objective
Featured research
Published work, indexed.
Each entry documents an observed problem, the method used to study it, what was found, and how the finding was applied in production.
Behavioural Analysis of Automated Repricers
How competing repricing systems react, how fast, and what their reaction latency reveals about the rules behind them.
Read the entry →Buy Box Probability Modelling
Treating the featured offer as a probability conditioned on price, delivery promise, fulfilment channel and seller state.
Read the entry →The field
A price is the end of a decision, not the beginning.
Most automation in ecommerce operates at the level of the price itself: a rule detects a competitor and changes a number. The decision that should sit behind that number — whether the position is worth holding, what it costs to hold it, what the competitor is likely to do next, whether the inventory and the margin support the move — is left to the operator, or to nobody.
My work studies that missing layer. It combines direct observation of marketplace behaviour, measurement of real costs and outcomes, and the construction of software that reasons through those variables before acting.
The research is conducted on a live marketplace business rather than in simulation, which is both its main constraint and its main value: every hypothesis is tested against real competitors, real fees and real consequences.
The framework
Five stages, and the price comes last.
Trajectory
Seven years of continuous work in one field.
The research did not begin as software. It began as an operating problem.
2020
Ecommerce operationsStart of marketplace trading and first exposure to automated competition.
2022
Decision modelsFirst structured models of when a price change is worth making.
2024
Repricer researchSystematic study of competing repricing behaviour and reaction latency.
2026
SellerFlowThe research consolidated into an operating decision platform.
Applied outcome
SellerFlow
SellerFlow is the software implementation of the Decision Intelligence Framework developed through years of applied marketplace research. It observes the marketplace continuously, establishes the real cost and margin of every unit, and decides whether a pricing action serves the objective of the business before it is taken.
It is presented here as the consequence of the research, not as a product.
Enquiries
Media, research and expert enquiries.
Available for interviews, technical commentary on marketplace pricing and competition, and collaboration with researchers and industry publications.
Contact