Data Science

Optimising marketplaces using machine learning

Evolve your platform. Every day.

Your marketplace doesn’t stand still – and neither do we. Our product roadmap is constantly moving forward to ensure that your investment is future-proof. B2B marketplaces are a valuable source of data, which is why our platform collects and tracks every click and bid.

Our machine learning capability ensures that your platform is continuously improving.

Data protection

NovaFori data science ecosystem

Data Science - Data lake
Data Science - Data analytics
Next step
Data Science Atom
Data Science Atom
Next step
Data Science Atom
Step 1

Understand what’s happening in the market

Understanding who is doing what on a multi-sided trading platform is hard. We help our clients understand what marketplace information is most instructive for their specific markets. Measuring marketplace performance to ensure that all segments of the market are thick (attracts optimal number of both sell-side and buy-side participants) and functioning optimally is key to growth, long-term sustainability and value optimisation.

Understand market with data science
Step 2

Predict future trends and performance

We help you predict your market's direction of travel by calculating probabilities and forecasting if the market will be ‘trending-up’ or ‘trending down’. Providing this information to buyers and sellers guides them in their pricing decisions. We also use this information to match buyers and sellers depending on the market’s current state.

Next step


Our machine learning models capture buyer preferences and adapt to current product availability.

Our models isolate patterns in historical supply and demand data. Trained models yield the probability that a bidder will place a bid, based on current product supply.

Next step


Our algorithms estimate private valuations from historic auction data.

We use information on successful and unsuccessful bids to estimate a buyer’s private valuation for their preferred product types.


We apply auction theory to simulate future scenarios, predict prices and marketplace performance.

We know the probability that a buyer will submit a bid, plus their likely maximum bid. Therefore, we can use auction theory to evaluate potential future scenarios.

Step 3

Optimise marketplace performance

Marketplace performance check

Price discovery

We use all of our capabilities to suggest buyers, sellers and inventory to optimise transaction prices and volumes.

Marketplace performance check

Recommendation engine

We embed auction theory in our recommendation engine to maximise price. Preference is given to recommended listings for highest possible bids. The objective is to maximise revenue across the entire marketplace and not just for a minority of listings.

Marketplace performance check

Buyer and seller matching

We use three methods to optimise segmentation:
micro - the activity and behavior of bidders in the current period;
mining - the most similar metadata attributes (e.g. reserve price);
macro - significant bid behaviour patterns in all available historical auction data.

Data Science inventory marching