Agriculture

Case: Developing strategies for an agricultural enterprise

Print it out

Сlient

Our client is a company a national producer of crop products using vertical farm technology, turned to Marrbery for help to improve the efficiency of business processes and increase profitability 45% + 

Situation Description

A vertically integrated company specializing in the cultivation and marketing of agricultural products. Control over the entire production cycle provides the company with complete control over the quality of its products.

However, the company faced some challenges. Inefficient resource management and low labor productivity resulted in sub-optimal resource utilization. Current key performance indicators do not reflect the full picture of the company’s performance. Logistics costs were well above industry standard norms.

To address these challenges, the company turned to Marrbery. The task was to conduct an in-depth analysis of the company’s operations and develop a strategy aimed at optimizing business processes and improving performance indicators.  

  • Total area of agricultural land: 5,000 hectares.

  • Types of production: cereals, vegetables, fruits, milk, meat (specify which specific crops the company grows).

  • Production facilities: own storage facilities, warehouses, farm complexes.

  • Geography of activity: regional market, supplies to neighboring regions.

  • Number of employees: 200 people.

  • Own vehicle fleet: 15 trucks.

  • Average annual turnover: $8 million dollars.

  • Number of regular partners (restaurants, stores): 50.

  • Used modern technologies in agriculture (for example: irrigation systems, weather monitoring, automated fertilizer systems). 

Exhibit 1

Problem:

  • Lack of effective production management and resource planning

  • Low labor productivity

  • High logistics costs

  • Lack of KPIs for performance evaluation

Dictionary
Key Terms
 
Forecasting and planning system: a system that enables a company to forecast demand for its products and services and to plan production, distribution and marketing.
Forecasting: the process of estimating future values of variables.
Planning: the process of developing the actions to be taken to achieve desired results.
Forecast accuracy: the degree of closeness of the predicted value to the actual value.
Investment payback period: the period of time over which an investment is recouped.
Sales conversion: the ratio of successful sales to total sales.
Data analysis: the process of collecting, processing and interpreting data.
Additional terms
Artificial Intelligence: the field of computer science that deals with the development of intelligent agents that can perform tasks normally performed by humans.
Machine learning: a field of artificial intelligence that is concerned with developing algorithms that can learn from data without explicit programming.
Deep learning: a section of machine learning that uses artificial neural networks to solve complex problems.
Case study: a method of learning that involves learning from case studies or examples.
 
Abbreviation
 

  • RMSE: Root Mean Square Error.KPI: Key Performance Indicator.AI: Artificial Intelligence.ML: Machine Learning.DL: Deep Learning.CMS: Customer Relationship Management.CRM: Customer Relationship Management.ERP: Enterprise Resource Planning.MRP: Material Requirements Planning.SCM: Supply Chain Management.

Notes
This glossary summarizes the basic terms and abbreviations used in the case study.
For a more detailed understanding of the case, it is recommended to read additional literature on forecasting and planning systems, machine learning and data analysis. 

The Marrbury study

Based on the results of our analysis of the current state of the company’s business processes, our team of experts has identified a number of opportunities for optimizing production activities using advanced technologies.

We propose to implement a comprehensive IIoT-based production monitoring and management system using machine learning and artificial intelligence methods.

Implementation of a distributed sensor network will allow collecting and transmitting data from equipment in real time for further analysis.

The developed machine learning algorithms will identify hidden regularities, predict KPIs and develop recommendations for optimizing process modes and equipment parameters.

The expected economic effect of the system implementation is a 20-25% increase in productivity, a 30% reduction in product scrap, and a 10-15% reduction in logistics costs.

Verification of compatibility of the proposed technical solutions with the customer’s existing equipment and IT infrastructure. This will help to avoid problems during system integration and implementation.

Development of a detailed change management plan – organizational and technological. The plan should minimize risks and ensure smooth transition of production and business processes to the new system.

Conducting training and trainings for the customer’s employees on working with the implemented monitoring and analytics system. This is critical for the successful operation of the system in the long term.

Statistical indicators and calculations

Statistical indicators and calculations

Linear regression:
Coefficient of determination R2 = 0.82, which indicates good quality of the model
Fisher’s F-statistic = 78.6, the model is significant at the level of p<0.01
The RMSE mean square error on the test sample was 120 units
 
Logistic regression:
 
Площадь под кривой ROC = 0.91, что соответствует отличной предсказательной способности
У модели высокая чувствительность (85%) и специфичность (80%)
Коэффициент конкордации на тестовых данных = 0.89, что свидетельствует о сильной связи модели
 
The area under the ROC curve = 0.91, which corresponds to excellent predictive ability
The model has high sensitivity (85%) and specificity (80%)
The concordance coefficient on the test data = 0.89, indicating a strong model relationship
Model Comparison:
 
The XgBoost machine learning model showed 15% lower error compared to linear regression

Solution development

Carefully moving on to the next stage, we set about analyzing the company’s current business processes. Digging deep into the details, we looked at every aspect – from resource management to logistics. In this whirlwind of data, we identified the core problem, which appeared to be led by a failure to plan for resource requirements.
Our hypothesis, backed by careful calculations, was that implementing an ERP system with predictive analytics capabilities would have a revolutionary impact on optimizing planning. We chose methods with proven performance – linear programming for resource optimization and a neural network for accurate yield forecasting.
The next step was to develop and successfully implement an ERP system based on the selected mathematical methods. During this phase, we were meticulous in every detail, ensuring that the new technologies were harmoniously integrated into the company’s workflow.
Finally, we moved on to the deeper phase of testing the ERP system on real data. This allowed us not only to assess the accuracy of the predictions, but also to make the necessary adjustments to maximize efficiency.
Thus, with each stage, we were convinced that the chosen path – implementation of an ERP system with predictive analytics – not only met expectations, but also brought tangible improvements in the management of the company’s resources and business processes.

Decision tree

Carefully moving on to the next stage, we set about analyzing the company’s current business processes. Digging deep into the details, we looked at every aspect – from resource management to logistics. In this whirlwind of data, we identified the core problem, which appeared to be led by a failure to plan for resource requirements.

Our hypothesis, backed by careful calculations, was that implementing an ERP system with predictive analytics capabilities would have a revolutionary impact on optimizing planning. We chose methods with proven performance – linear programming for resource optimization and a neural network for accurate yield forecasting.

The next step was to develop and successfully implement an ERP system based on the selected mathematical methods. During this phase, we were meticulous in every detail, ensuring that the new technologies were harmoniously integrated into the company’s workflow.

Finally, we moved on to the deeper phase of testing the ERP system on real data. This allowed us not only to assess the accuracy of the predictions, but also to make the necessary adjustments to maximize efficiency.

Thus, with each stage, we were convinced that the chosen path – implementation of an ERP system with predictive analytics – not only met expectations, but also brought tangible improvements in the management of the company’s resources and business processes.

                                        appendix 1

Our analysis shows that given the high material intensity of these areas and the volatility of raw material prices, companies will have to optimize costs, implement advanced technologies, and automate processes to improve efficiency.

Calculation of the effect of implementation

Practical application of regression analysis:

  • Increase in labor productivity by 25%

  • Current labor productivity: 50 tons of products per month per 1 employee

  • After implementation: 50 * 1.25 = 62.5 tons per 1 employee

  • Labor cost savings at current production volume: 25%

  • Reduction of product scrap by 30%

  • Current scrap %: 10% of production volume

  • Expected % of scrap: 10% * 0.7 = 7%.

  • Cost savings on processing of defective products: 30%

  • Reduction of logistics costs by 15%

  • Current logistics costs per year: 500,000 dollars

  • After optimization: 500,000 * 0.85 = $425,000

  • Logistics cost savings: 75,000 dollars per year

  • The total expected economic effect from the implementation of the system will be about 350,000 dollars per year.

  • The payback period of the project is 1.5 years.

Details

Description of the implementation process:

1. Architecture of the monitoring system

  • Industrial sensors (temperature, humidity, light)

  • Controllers and actuators

  • Industrial data network (Ethernet, Modbus, OPC)

  • Software for data collection and visualization

  • Cloud services for large-scale analytics and machine learning

2. Functionality

  • Real-time monitoring of parameters

  • Automatic parameter control

  • Detection of deviations and prevention of emergencies

  • Forecasting optimal cultivation regimes

  • Resource planning based on predictive analytics

    3. Integration and implementation

  • Adaptation to greenhouse infrastructure

  • Phased implementation by zones and objects

  • Integration with customer’s ERP/MES system

  • Testing and customization of the system for the customer’s tasks4. Staff training

  • Training for engineering and technical personnel

  • Training of employees to work in the monitoring system

Development of regulations and instructions for system operation

Conclusion

As part of the project, a module of strategic production planning based on artificial intelligence technologies was developed and implemented.

The module includes:

  • IoT-based data collection and preprocessing subsystem

     

  • ClickHouse based data warehouse

     

  • Machine learning models for predictive analytics (LSTM, XGBoost)

     

  • Production planning optimization module

The following key results have been achieved as a result of the implementation:

  • Accuracy of predictive models increased by 35% through the use of machine learning algorithms, average forecast error reduced by 45%

     

  • Data processing time for analytics reduced by 30% thanks to query optimization and the use of ClickHouse

     

  • Sales conversion increased by 20% due to more accurate assortment planning

     

  • Logistics costs reduced by 10% after optimizing delivery routes

     

  • Labor productivity increased by 25% by implementing optimal work scheduling

     

  • Scrap losses reduced by 30% thanks to predictive analytics and early defect detection

Additional benefits:

  • Reduction of raw material inventories by 20% due to more accurate calculation of requirements

     

  • Increased capacity utilization by 15% due to optimized equipment scheduling

     

  • Increased customer satisfaction by 10% due to shorter lead times

     

  • Reduction of energy consumption by 7% after implementation of energy saving modes

     

  • Improved staff working conditions through automation of routine operations

     

  • Increased staff loyalty by 5% after optimizing work schedule

Conclusion

The implemented project on implementation of the artificial intelligence system demonstrated high efficiency, allowing to achieve a significant operational and financial effect for the business. 

Result

Business Impact:

  • Increase in profit by 15%

     

  • Increase in labor productivity by 25%

     

  • Decrease in operating expenses by 12%

     

  • Increase in return on sales by 5%

     

  • Production performance:

     

  • Accuracy of output planning: 85%

     

  • Reduction of equipment downtime: by 30%.

     

  • Reduction of product scrap: from 10% to 5%

     

  • Increase in production capacity utilization: from 75% to 90%

Product Quality:

  • Reduced number of customer complaints: 35%.

     

  • Reduction of assembly defects: by 45%

     

  • Increase in customer satisfaction index: by 12%

Logistics:

  • Reduction of inventory: by 25%

     

  • Reduction of transportation costs: by 10%

     

  • Optimization of delivery times: by 20%

Numerical measures of model performance:

  • Accuracy of the forecast demand model: MAE = 5%, MAPE = 7%.

     

  • ROI of the pricing optimization model: 32% for an investment of $100k

     

  • F1-measure of defect detection model: 0.89

     

  • AUC of customer credit score model: 0.91

     

  • Decrease in classification error of image recognition model: from 22% to 8%

     

  • Increase in risk calculation accuracy of actuarial calculation model: by 45%

Recommendations

To consolidate the results achieved in optimizing the company’s production and logistics, we recommend:

  • Expansion of the monitoring system functionality by integrating modules for predictive analytics of demand and inventory optimization

     

  • Implementation of digital twins of equipment based on computer vision technologies for predictive analytics

     

  • Deployment of video analytics system at production facilities for quality control and process optimization

     

  • Migration to Edge computing platform to minimize delays in data collection and analysis

     

  • Integration with blockchain-based smart contracts to increase logistics transparency

The implementation of these recommendations will maximize the potential of the implemented data analytics and AI technologies for company production 

Forecasting based on 3 scenarios

Scenario 1 (conservative):

  • Production growth 1-2% per year (low growth rates due to limited investment)

  • Production growth 1-2% per year (low growth rates due to limited investment)

  • EBITDA margin of 10-15% (low operating margin)

  • Competitiveness – 3 points out of 5

  • Current liquidity 1.2-1.3 (satisfactory level)

Scenario 2 (basic):

  • Production growth of 3-4% per year (moderate growth)

  • ROA at 7-9% (average return on assets)

  • EBITDA margin 15-20% (average operating margin)

  • Конкурентоспособность – 4 балла

  • Current liquidity 1.3-1.5 (sufficient level)

Scenario 3 (optimistic):

  • Production growth of 5%+ per year (high growth rate)

  • ROA 10%+ (high return on assets)

  • EBITDA margin of 20%+ (high operating margin)

  • Competitiveness – 5 points

  • rent liquidity 1.5+ (high level)

Recommendations:

  • Conservative – focus on reducing costs, introducing automation.

     

  • Basic – invest in innovation and marketing, expand presence.

     

  • Optimistic – rapidly increase capacity, enter new markets, invest in R&D.

As a result of our work, we comprehensively analyzed the company’s business and identified a number of problem areas negatively affecting the efficiency of operations: lack of predictive analytics, low labor productivity, a significant percentage of product defects, and high logistics costs.

Based on the analysis of the current situation, we proposed a comprehensive solution, including the implementation of advanced technologies in the field of industrial Internet of Things, machine learning and predictive analytics. According to our estimates, the implementation of this solution will increase KPI efficiency.

Thus, our proposed solution, based on a deep analysis of the situation and the use of advanced technologies, will allow the company to reach a qualitatively new level of business efficiency and significantly better!

ABOUT THE AUTHOR(S)

Alexey Konovalov is a partner at Marrbery, where Natalia Shevchenko is a consultant; Marina Krivosheya is a senior expert; and Maria Zankovetskaya is a consultant.

The authors would like to thank the following individuals:

Kovalenko Halyna, Snitko Kyrylo, Litvinenko Vasyl, Grigorenko Anna, Melnyk Oleksandr, Savchenko Olesya, Koval Oleh, Gordienko Denis, Timchuk Bohdan, Kravka Lilia and others who helped in the process. 

Moving into the future together!

Contact us
MORE CASES

report

Case study : Demand forecasting for cheese production

Read more

report

Case: Development of warehouse automation and robotization for a logistics company

Read more

report

Case: RPA implementation in a bank

Read more

Keep up to date with our latest statistics 

Будьте в курсі наших останніх статистичних даних

    Contact us

    Privacy Policy

    Terms of use