AI Pricing Optimisation Made Simple: 3 Use Cases

In transport freight and logistics, pricing optimisation is crucial to achieving business objectives like increasing margins, market share, customer retention and revenue. Effectively setting prices requires a nuanced understanding of market dynamics, customer preferences and competitive landscapes, within which AI can offer invaluable insights.

Implementing AI technology might seem complex; it requires organising data, model development, training and extensive fine-tuning. Yet, focusing AI on select pricing areas can yield benefits over a shorter period, requiring less time and fewer resources. 

Here, we run through three quick and simple AI-driven price optimisation use cases, each one broken down with a step-by-step implementation guide. These demonstrate how the time vs. investment vs. payoff weighs up, offering a cost-effective solution to AI-powered pricing optimisation.

1. Differentiated pricing customised according to customer characteristics

The freight transport market is highly fragmented, with diverse industries seeking services from numerous carriers. However, pricing information on freight rate management, particularly on the spot market, remains somewhat opaque. This leads to significant price fluctuations for both buyers and sellers, independent of supply shortages or demand spikes. 

For instance, the European contract road freight rate index hit an all-time high of 121 points in Q2 2022, showing a 13.1-point increase year-on-year. Following this peak, road freight demand experienced a marked downturn during 2023. Carriers seeking to secure loads amidst heightened competition adopted low-price strategies, resulting in a notable decline in freight rates. In Q1 2023, these plummeted by as much as 9%. Considering this volatility, it’s never been more important to implement intelligent solutions to navigate demand fluctuation. 

This fluctuation in freight prices arises from the varying price sensitivities of customers across different industries, each with unique margins and business objectives. Companies dealing in low-value goods often prioritise cost and may favour loyalty programmes and discounts over quality. Conversely, those in high-value goods industries prioritise quality and service over price, displaying lower price sensitivity.

Understanding the dynamics of your customer base is crucial for effective business strategy. This knowledge can be efficiently gained and utilised through the implementation of AI technology.

5 steps to effective customer segmentation

1. Analyse historical and real-time data on conversion. How do shippers behave when they request a quote? Do they log in many times and generate a lot of quotes per day before logging in to order? Or do they enter once to confirm? 
2. Segment the customers. Analyse shipper behaviour to segment them based on preferences. AI could help identify how many homogenous groups are out there and in which way they are different.
3. Complement your data with general market information. This could include the size of the company, industry, life cycle, number of employees etc. 
4. Define your price strategy depending on your business objectives. For example, it’s recommended to implement a loyalty programme for price-sensitive clients, giving discounts based on volume or other conditions. Meanwhile, for the segment focused on quality, it’s less effective to apply any discounts. Instead, they probably would be more incentivised by special attention. 
5. Use the AI technology to periodically review the segmentation and to assign new customers to one or another segment. 

2. Geographical differentiation for pricing optimisation

Transportation costs vary across countries and regions, influenced by geographical factors such as pickup and drop-off locations. Effective price optimisation should reflect these differences, pricing higher for remote or challenging-to-access regions and competitively for areas with ample transport options. 

Challenges arise as disparities can occur within a given area. For example, a street's accessibility due to size, rush hour congestion, or mountainous areas accessible only via rugged terrain can significantly impact pricing dynamics.

5 steps to geographical pricing optimisation 

1. Use publicly available information to analyse a region. This analysis should include industrial areas and registered companies as well as traffic information.
2. Create a heatmap of possible pickup and dropoff points. Using the addresses of all registered companies, create a geographical visualisation or “heatmap” of demand at various points in different areas. 
3. Estimate densities. By consolidating the pick-up, drop-off and transit points, it is possible to create a “cluster”. This cluster serves as the foundation for estimating density areas, achieved through Kernel Density Estimation.
4. Assign density codes. Subsequently, assign density codes to zip codes to represent the initial categorisation of each area. These density codes – which can be elaborated upon – impact pricing on both the buy and sell side.
5. Develop a model comparing density to ease of supply. Once operational in the area, a model can be developed that considers the actual density of orders and the ease of finding supply in each area. With the help of AI, this model iteratively updated the density codes. Consequently, we can offer competitive pricing that considers the geographical specifics of each zip code.

‍

3. Seasonal pricing optimisation

It's widely acknowledged that the transportation market experiences pronounced seasonal fluctuations, significantly impacting its dynamics. As part of our Ontruck AI Tech solutions, we've devised an AI model to optimise the buy-price and sell-price year-round. This is part of our general demand forecasting model, which is explained further in our article on seasonal demand forecasting.

4 steps to seasonal pricing optimisation with an AI model

1. Analyse time-series data. Available information was utilised to conduct a time-series analysis of demand (number of orders) and supply (number of active drivers, job views, or bids). 
2. Deploy Prophet model. We implement the Prophet model developed by Meta, which is adept at handling datasets with strong seasonal patterns.
3. Consider holidays and events. Our developers then factored in holidays and events that may impact demand and supply fluctuations (e.g., Christmas).
4. Enable dynamic forecasting. The model utilises real-time data for dynamic forecasting, enabling adjustments on-the-fly, based on changing market conditions.

With the assistance of this model, logistics service providers can regulate and refine their pricing strategies throughout the year. For instance, historical data indicates that buy prices typically decrease by 10-15% in January and February, gradually rising again until reaching maximum levels in July.

Effective price optimisation at your fingertips

AI technology, combined with advanced data analysis, presents numerous opportunities to enhance critical business processes within transport companies. While the use cases discussed here provide a comprehensive overview, they only scratch the surface of AI’s potential.

If you're interested in exploring these solutions further, schedule a demo with us today. Experience firsthand how our technology is revolutionising pricing optimisation for logistics service providers, driving higher margins and maximum profits. 

‍

References:

• Facebook. (2023). Prophet,Forecasting at scale. https://facebook.github.io/prophet/

• IRU. (2022,August). Benchmark of European road freight rates Q2 2022: European road transport prices break new records. https://www.iru.org/news-resources/newsroom/benchmark-european-road-freight-rates-q2-2022-european-road-transport-prices-break-new-records
• IRU. (2024,February). European road freight rate benchmark Q4 2023: Contract market holds, spot market plummets. https://www.iru.org/news-resources/newsroom/european-road-freight-rate-benchmark-q4-2023-contract-market-holds-spot-market-plummets
• scikit-learn. (s.f.) Density estimation. https://scikit-learn.org/stable/modules/density.html#kernel-density-estimation