Case study
Smarter FX and Equity Trade Recommendations
How SquareShift built nightly FX and equity trade recommendation engines for a multinational investment bank headquartered in the EU.
Ranking trade ideas nightly beats real-time noise. Bankers don't need every signal the moment it fires — they need the best ideas already ranked and waiting when they sit down.
AI & Generative AI Practice Lead, SquareShift
A multinational investment bank headquartered in the EU, serving institutional clients across Europe and North America.
Its sales and trading desk wanted better FX and equity trade recommendations to grow trading revenue and cut research time.
- Nightly recommendation engine — FX and equity trade ideas delivered every night via Tableau.
- Unified client data lake — transactions, meeting notes, and investment goals in one place.
- Faster trade-opportunity identification — replaces manual research across a large volume of transactions.
How SquareShift delivered it.
The challenge
Sales and trading bankers were sitting on a large volume of transactions but had no systematic way to surface which FX and equity trades were actually worth recommending. Spotting the right opportunity meant manually sifting through transaction history, meeting notes, and client goals.
What we delivered
SquareShift built a data lake that unified all client-related data — transactions, meeting notes, and investment goals — in one place. On top of it, SquareShift built an FX recommendation engine using DTW clustering, cross-correlation, and PageRank, plus a separate equity recommendation engine built on the RFM model.
Recommendations generate nightly and land in Tableau dashboards and email, so bankers start each day with ranked trade ideas instead of a blank research queue.
The payoff
S&T bankers now open ranked FX and equity trade recommendations every morning instead of building them from scratch. The nightly-refreshed data lake and recommendation engines cut manual research time and give the desk a repeatable way to spot trade opportunities across a high volume of transactions.
Where this work sits