⚙️
🌟 Najlepsze dopasowanie
Wpisz szukaną frazę, aby zobaczyć propozycje AI.

How to collect, analyze, and present data in logistics

How to collect, analyze, and present data in logistics
Photo: Depositphotos

Modern supply chains operate under conditions of permanent uncertainty, increasing market dynamics, cost pressure, and constant disruptions in transport networks. Traditional logistics management methods, based on managerial intuition, historical spreadsheets, and reactive operational firefighting, have become a major source of inefficiency. Transforming raw operational events into accurate, real-time business decisions requires a structured and rigorous approach to data engineering.

 

Despite the massive volume of data generated every day by warehouse scanners, telematics sensors, and transaction systems, most enterprises struggle with the paradox of a knowledge deficit amidst an information surplus. The primary barriers are information silos, poor master data quality, lack of definitional consistency, and architectural fragmentation. According to Gartner market research, only 7% of organizations in the supply chain sector have the capability to execute and enforce logistics decisions in real-time. The remaining 93% of enterprises are forced to operate in a reactive mode, incurring high costs of delays, contractual penalties, excessive buffer stocks, and suboptimal fleet utilization.

This article discusses the full information lifecycle: from modern architectures for collecting data from ERP, WMS, TMS, and Internet of Things (IoT) systems, through multi-level analytical methodologies and mathematical formalization of KPIs, to advanced visualization techniques, digital twins, and autonomous artificial intelligence agents.

Architecture for collecting and integrating logistics data

An effective analytical process in logistics requires building a stable, scalable, and flexible integration architecture capable of capturing and harmonizing data from technologically diverse sources. Traditionally, logistics data remains locked in separate departmental transaction systems, which prevents full end-to-end visibility of processes. Modernizing analytical infrastructure requires removing these barriers by implementing modern information flow patterns.

Ecosystem of data sources and domain relationships

In a complex supply chain, data is generated continuously at every stage of the order fulfillment cycle. Proper management of this environment requires clearly defining which system serves as the System of Record / Source of Truth for individual business entities:

  1. ERP (Enterprise Resource Planning) System: Serves as the organization’s financial and commercial foundation. It is the source of truth for SKU Master files, supplier and customer files, Sales Orders (SO), Purchase Orders (PO), accounting inventory valuation, and main financial flows.
  2. WMS (Warehouse Management System): Exercises direct supervision over operations within the warehouse facility. It is the source of truth for the physical location of inventory (zones, racks, bins), stock availability statuses, inbound processes, picking, packing, outbound, and cycle counting.
  3. TMS (Transportation Management System): Oversees the processes of moving cargo between network nodes. It is the source of truth for route planning, shipment consolidation, carrier selection and evaluation (carrier scoring), freight rates, transport notifications, and transport cost settlements.
  4. Telematics and IoT (Internet of Things) sensors: Constitute the edge measurement layer mounted on transport and warehouse assets. On-board telematics record continuous data streams from the vehicle’s CAN bus (fuel consumption, engine run time, speed, axle load) and GPS geolocation signals. Wireless IoT sensors monitor environmental parameters of transported cargo, including temperature, humidity, shocks, or cargo space opening, which is critical in the logistics of pharmaceutical products and fresh food (Cold Chain).

Table 1. Data systems and their data collection and analysis capabilities

System / Data Source Main operational domain System of Truth for… Dominant mode and frequency of data generation
ERP Finance, procurement, sales, inventory accounting Material files, supplier/customer master-data, finance Transactional / Batch
WMS Intra-warehouse operations Physical stock and zonal location, picking status Real-time / Event-driven
TMS Transport, shipping, carrier evaluation Routes, freight rating, delivery statuses, transport settlements Real-time / Streaming
Telematics / IoT Fleet and shipment tracking in transit GPS location, vehicle technical condition, cargo conditions High-frequency streaming

Integration patterns: Event-Driven Architecture and Data Lakehouse

The answer to information delay challenges is moving away from traditional, periodic file exchange (e.g., nightly batch transfers in EDI standard or flat CSV files) towards Event-Driven Architecture (EDA) and RESTful API interfaces. In the EDA model, any change in the state of a business object (e.g., changing an order status to “picked,” a trailer leaving a geofence zone, or exceeding a temperature threshold in a cold store) immediately generates a message published on a data bus (e.g., Apache Kafka). Interested domain systems consume this event in real-time, which allows for the automatic triggering of subsequent process steps – e.g., immediately recalculating the Estimated Time of Arrival (ETA) in the TMS and sending a notification to the customer.

For analytical purposes and building advanced decision models, the Data Lakehouse architecture is used. This solution combines the flexibility and low cost of storing raw, unstructured data streams (characteristic of a Data Lake) with the structural rigor, ACID transactions, and high performance of SQL queries known from traditional data warehouses. Storing data in the lakehouse layer protects production transaction systems (ERP, WMS) from performance degradation during resource-intensive analytical and machine learning queries.

Master Data Management (MDM) and elimination of semantic drift

Lack of definitional consistency is the main cause of failure for analytical projects in logistics. A common phenomenon in large organizations is so-called Semantic Drift, where different departments of the company define the same business concepts in different ways. An example could be the concept of “on-time delivery”: the transport logistics department may consider a delivery on time when the truck enters the facility, while the customer service department measures timeliness from the moment of physical unloading and signing of the delivery document without quality reservations.

The solution to this problem is implementing Master Data Management (MDM) procedures and creating a Golden Record for Critical Data Elements (CDE). Using a business rigor test based on risk analysis, revenue impact, and regulatory requirements (Risk, Revenue, Regulation test), the organization designates named Data Owners responsible for their uniform definition, quality, and validation rules across all IT systems.

How to collect, analyze, and present data in logistics
Photo: Depositphotos

Methodology and levels of logistics data analysis

The analytical process in logistics is characterized by a hierarchical structure. Building advanced predictive and autonomous systems is only possible after first stabilizing the lower reporting layers.

Levels of analytical maturity in logistics

In supply chain management practice, there are five distinct levels of analytical maturity:

The first stage is Descriptive Analytics (What happened?), responsible for traditional historical reporting. Its task is to aggregate operational events and compile key indicators, such as total freight cost, inventory turnover ratio, or average order cycle time. Descriptive analysis informs about the past, serving as a starting point for evaluating plan execution.

The next step is Diagnostic Analytics (Why did it happen?), focused on Root-Cause Analysis. It uses drill-down techniques, cohort breakdowns, and multi-dimensional data cubes. It allows for precisely determining the factors that led to a drop in customer service levels in a selected region or isolating carriers generating the most delays on specific routes.

The third level is Predictive Analytics (What will happen?), shifting the focus to imagining future states of the system. It uses advanced statistical models, time series analysis, and machine learning algorithms to forecast demand, identify the risk of stockouts, or predict spikes in spot rates in transport with a designated confidence interval.

The fourth stage is Prescriptive Analytics (What should be done?), providing active decision support by directly generating operational recommendations. Prescriptive algorithms combine forecast results with optimization technologies (e.g., linear programming and heuristics). They allow for determining optimal travel routes considering time windows, recommending dynamic inventory rebalancing between warehouses, or suggesting the optimal choice of freight supplier.

The highest level of maturity is Cognitive Analytics (Autonomous reasoning and agentic action). It is based on a virtual workforce composed of autonomous artificial intelligence agents capable of independent reasoning in complex environments. Cognitive agents can autonomously confront demand signals with inventory levels in ERP and WMS, diagnose the causes of deviations, and take corrective actions (e.g., ordering inter-warehouse transfers) within the scope of their granted business authorities.

 

Table 2. Levels of analytical maturity

Analytical maturity level Dominant business question Main product / Analytical output Prerequisites and limitations
1. Descriptive What happened? Monthly KPI reports, spend reviews, inventory turnover Informs about the past; no insight into causes
2. Diagnostic Why did it happen? Root-cause analysis, drill-down, data cubes Exhausts itself in explaining history; no prediction
3. Predictive What will happen? Demand forecasts, supplier risks, freight rate prediction Requires historically clean time series
4. Prescriptive What should be done? Network optimization algorithms, inventory rebalancing recommendations Requires maturity of levels 1–3 and tight integration
5. Cognitive How to solve the problem autonomously? AI agents performing autonomous replanning and process corrections Requires a consistent semantic layer and high data quality

SCOR-DS standard and formalization of operational indicators

Proper analysis of logistics data requires basing operational metrics on recognized industry standards. The most popular reference framework is the SCOR-DS (Supply Chain Operations Reference Digital Standard) model developed by the ASCM organization. SCOR-DS describes logistics processes through seven main domains: Orchestrate, Plan, Order, Source, Transform, Fulfill, and Return. Performance metrics are classified within five basic attributes: Reliability, Responsiveness, Agility, Cost, and Asset Management.

A key metric in the Reliability area is the OTIF (On-Time In-Full) indicator, measuring the percentage of orders delivered to the customer exactly within the agreed time window, in the full ordered quantity, and without damage. The mathematical formula for calculating the OTIF indicator takes the form:

In the above equation, N denotes the total number of completed orders in the analyzed period. The binary variable Oi takes the value 1 if order i was delivered within the designated time window, or 0 in case of lack of timeliness. The variable Fi takes the value 1 if the order was delivered with 100% SKU quantity compliance, or 0 in case of shortages. The variable Qi takes the value 1 when the shipment showed no quality damage or documentation errors, and 0 in case of defects.

For benchmarking purposes and building synthetic scorecards for suppliers or logistics network nodes, methods of normalizing indicators with different units of measure are used. A commonly used standard is the Snorm de Boera normalization, transforming raw KPI values to a uniform scale from 0 (worst possible performance) to 100 (optimal performance).

In the case of stimulating indicators, where a higher value means a better operational result (e.g., OTIF, inventory turnover), the formula is used:

In the case of destimulating indicators, where the lowest possible value is preferred (e.g., order cycle time, freight cost per unit, damage rate), the formula takes the form:

The value Si represents the achieved actual result, while Smax and Smin designate the best and worst boundary values in the analyzed research sample, respectively.

Presentation and visualization of logistics data

The way analytical results are presented determines their comprehensibility and real impact on business decisions. A common mistake of organizations is designing so-called Orphaned Dashboards – reports created on a one-time request that eventually stop being maintained and generate information noise.

Designing operational, tactical, and strategic dashboards

Visual presentation of logistics data must be strictly tailored to the specifics of the recipient and the time horizon of the decisions being made.

Support for the operational level (route dispatchers, warehouse shift managers) requires designing dashboards focused on immediate identification of deviations through an Exception Reporting mechanism. Instead of presenting a full database of hundreds of executed transports, the operational interface should expose only those shipments for which the forecasted delay exceeds the assumed critical threshold or in which the cold chain has been broken.

The tactical level (supply chain analysts, transport procurement specialists) requires dashboards enabling multi-dimensional analysis and drill-down. These tools are used to assess the degree of SLA fulfillment by individual carriers, analyze warehouse cost structures, or identify seasonal bottlenecks in the distribution network.

In turn, the presentation for the strategic level (chief operating officers, board members) must aggregate complex indicators, linking logistics performance with the company’s financial results. Key indicators here include Total Logistics Cost expressed as a percentage of sales revenue, Cash-to-Cash Cycle Time, and sustainability indicators including the annual fleet carbon footprint (Scope 3 Emissions).

Advanced forms of visualization: Graph databases and Digital Twin simulations

Traditional column charts and pivot tables prove insufficient in situations where it is necessary to present network relationships and complex spatial dependencies. In modern logistics analytics, the following play an increasingly important role:

Visualizations based on Graph Analytics, which represent the logistics network as a set of nodes (warehouses, suppliers, ports) and edges connected to them (transport routes, supply dependencies). This way of presenting data allows analysts to immediately locate Single Points of Failure and optimize complex intermodal flows.

Supply Chain Digital Twins connected to simulation engines simulate real objects and networks based on a continuous stream of data. Visualizing a digital twin allows decision-makers to conduct interactive What-If Analysis scenario tests in real-time. This allows for directly observing the predicted effects of opening a new distribution center, blocking a selected maritime node, or a sudden increase in fuel prices without incurring operational risk.

How to collect, analyze, and present data in logistics
Photo: Depositphotos

Case studies and technology trends 2026+

Implementing modern methods of collecting, analyzing, and presenting data translates into concrete operational achievements and market advantage for leading logistics enterprises.

Empirical examples of analytical transformation

A review of industrial practices confirms a step-change improvement in performance indicators after eliminating information silos.

Implementing electronic document flow and automation of analytical processes at Raben Group allowed for shortening the processing time of logistics and settlement documentation by 30–50%. This ensured continuity of operational data exchange between entities and eliminated information bottlenecks in handling cross-border processes.

In turn, at the APM Terminals port transshipment hub (Maersk Group), the challenge was the low performance of traditional Wi-Fi networks in the difficult environment of large-scale metal structures and containers. Building a dedicated, private 5G cellular network enabled a 10-fold increase in data transmission performance from IoT sensors mounted on cranes and autonomous vehicles. This translated into a direct improvement in transshipment fluidity and reduced vessel handling time.

At the Lufthansa Cargo facility at Los Angeles Airport (LAX), integrating data from real-time location sensors and automatic cargo flow analysis allowed for improving the speed of warehouse and transshipment processes by 75%.

Equally spectacular effects are brought by Master Data Management projects in the industrial and energy sector. Creating a Golden Record for over 20,000 logistics assets and materials eliminated manual data corrections, saving over 5,000 hours of analytical work annually and reducing financial reserves created for inventory valuation errors by millions of dollars.

Technology trends shaping logistics analytics

Modern logistics data management strategies are shaped by key technological trends:

Physical AI represents a direct combination of machine learning models with edge IoT sensors, robotics, and warehouse automation. This technology enables machines to autonomously analyze the environment in real-time, which translates into dynamic adaptation of packing, sorting, and intra-warehouse transport processes to variable cargo parameters.

AI agents and Collaborative Multiagent Systems (MAS) are redefining the traditional approach to decision analytics. Instead of passive managerial dashboards, organizations are deploying specialized agents based on Domain-Specific Language Models. These agents work closely together: one agent identifies a delay in a vehicle, the second verifies inventory levels in the WMS, and the third autonomously sends a request for quotation to an alternative carrier.

The development of autonomy, however, forces the parallel implementation of Decision Governance systems. In conditions where nearly 67% of digital budgets in logistics are directed at AI solutions, while at the same time over 55% of management staff report uncertainty regarding the return on investment (ROI), building transparent control frameworks becomes key. Decision Governance sets the rules of accountability, auditability, and decision boundaries for algorithms, ensuring their actions are consistent with the company’s business goals and regulatory requirements.

Conclusions

Effective collection, analysis, and presentation of logistics data requires moving beyond treating analytics as a purely technical domain of the IT department. To build a modern, resilient, and data-driven logistics organization, management should implement a structured program of actions:

  1. Adopting an Outcome-First Approach: Building an analytical architecture should not start with an attempt to consolidate all available information resources, but with selecting a specific, key operational decision that is currently made too slowly or generates errors. Reconstructing the data path necessary to make this decision allows for precisely defining the required sources and integration points of ERP, WMS, and TMS systems.
  2. Rigorous selection and protection of Critical Data Elements (CDE): The organization must identify critical fields in databases – such as SKU code, supplier identifier, location, or delivery promise date – and assign named business owners (Data Owners) to them. This establishes unambiguous semantic definitions and prevents situations where errors in master data affect operational decisions.
  3. Implementing API-first and Data Lakehouse architecture: The technical priority should be moving away from outdated batch transfers towards Event-Driven Architecture (EDA) and lakehouse technology. This ensures flexibility and low latency in information transfer between systems, enabling real-time analytics.
  4. Constructing interfaces tailored to business roles: One should move away from complex aggregate dashboards towards operational views integrated with the daily work tools of dispatchers and analysts. The presented data should be filtered according to the exception reporting principle, eliminating information noise.
  5. Preparing frameworks for agentic analytics (Decision Governance): Automating decision-making processes using artificial intelligence and AI agents must go hand in hand with ensuring full transparency, model auditability, and maintaining human oversight at points with a high level of financial and operational risk.

The article comes from Logistyka magazine no. 5/2026