Our GEN*4 AI will blow your neural network

PROS created the Profit Optimization Software category, and we remain the undisputed leader in the space. With the launch of the PROS Platform, we now have a flexible and modular platform which future proofs customers who want to build organic margin by optimizing costs and revenue in existing operations.

PROS TODAY

GEN*4

2022

Neural Network

Leveraging the last AI advances in ML and neural networks to improve modeling techniques.

HOW IT WORKS

Price Prediction: Neural network eliminated and replaced segmentation approach that use all available attributes, not a subset as in segmentation models.

Price Optimization: Provides customer-specific, market-aware, and win-rate based pricing.

ADVANTAGES

  • Peers are predicted dynamically based on transaction attributes (which includes customer and product features)
  • Improved prediction accuracy of the model
  • Ability to use categorical features with huge number of values
  • Ability to generate optimal price recommendations using loss information
  • Automatic Seasonality and Trend
  • Ability to use unconventional co-variates (market indices, competitor data)
  • Sparse data handling — many features, categorical features with many values
  • Profit/Revenue Optimization to determine target price
  • Uses explainable AI model to provide transparency
  • Extensible to Non-Negotiation Guidance

GEN*3

2018

Dynamic Segmentation-Based Pricing

Major move to user-driven workflows so that the science moves out of the backroom and into the application itself.

HOW IT WORKS

Segmentation: Enhanced to SKU-centric symmetric tree that utilizes dynamic attribute roll-up that is done online.

Price Optimization: Customer-level analysis with benchmarking and guidance against peers with price change aggressiveness levers changes controlled by the user.

ADVANTAGES

  • Product centric
  • Embedded gradual correction of underperformers
  • Easy results validation by the user within the UI
  • Simulation capabilities that include a benefit estimate (for ROI estimation)
  • Enable users to create and manage all aspects of segmentation and pricing guidance
  • Further improvement on sparsity handling with rolled up ability to re-segment if needed
  • More meaningful floor and expert price to aid negotiation

DISADVANTAGES

  • Limited number of features and feature values
  • Limited ability to use market indices that informs potential changes in historical data. History can be misleading during volatile market conditions
  • Inability to use loss data (indirect use of win/loss data)
  • Segmentation model is limited to pre-selected fixed set of attributes

COMPETITION TODAY

GEN*2

2014

Decision Tree

AI advances continue as we move to the use of data-science driven segmentation to reduce the Cartesian data sparsity problem.

HOW IT WORKS

Segmentation: Supervised machine learning asymetric binary tree model (based on CART) with advanced model fit statistic uses Bayesian Information Criteria to prune the decision tree.

Price Optimization: No change.

ADVANTAGES

  • Wights recent transactions more heavily
  • Dynamic attribute use
  • Flexibility of the tree generation process improves predictability
  • Improved data sparsity handling (not enough transactions and not enough customer diversity)

DISADVANTAGES

  • Segmentation was difficult or infeasible to visualize, which can hurt adoption
  • Business rules to ensure target pricing achievable
  • Attribute selection and segmentation were determined in offline manner

GEN*1

2009

Cartesian Model

AI begins in its basic form. Customer-specific willingness-to-pay price guidance to provide the right price to the customer.

HOW IT WORKS

Segmentation: Symmetric cartesian cross-product algorithm. (Bucket feature values and then create cross-product from the bucket) uses Bayesian Information Criteria.

Price Optimization: Percentile of historical price paid within the segment.

ADVANTAGES

  • Consistent peer groups
  • Placeholder segments
  • Easy to visualize and explain
  • Customer specific pricing takes into account low and high performers

DISADVANTAGES

  • Static attribute groupings
  • Rigid data structure leading to sparse data conditions
  • Require business rules to ensure target pricing is achievable (otherwise could result in large price changes) 
  • Attribute selection and segmentation determined in an offline manner
  • No explanation for price recommendations in UI

GEN. 0

2005

Digitization

Foundational price management technology. Replace Excel and integrate into back-office systems.

ADVANTAGES

  • More efficient and automated
  • One central platform for all pricing and cost information
  • Price transparency and visibility
  • Granularity — able to dive into more detail (price at lower-level components and rolled up to the whole producct)
  • Exception-based review relying on alerts and thresholds
  • Streamlined workflow
  • Ability to see the impact of price change (added 2019)
  • Ability to use business rules based on deader-follower relationships on product portfolio 
  • Flexibility to build complex business rules

DISADVANTAGES

  • Business-rules becomes too numerous and hard to maintain and kept updated
  • Rely on human decision
  • Relies on broad (business/marketing) customer classifications/segments
  • Matrix-based discounting structures