About this revised edition
Revised September 23, 2026 for clarity and source accuracy. Historical McKinsey and Kearney figures now include attribution, dates and scope; an unverified market forecast and unsupported performance promises were removed.
Supply-chain visibility has value when it changes a decision: which order to expedite, where to replenish stock or when to revise a delivery promise. Technology can supply better signals. People still need clear responsibility for acting on them.
The technologies discussed in this article's original 2024 edition remain useful categories for evaluating an investment: automation, connected sensors, predictive analytics, shared ledgers and cloud applications. They address different problems. Choosing among them should begin with a specific operating constraint.
Put the historical statistics in context
In its 2016 assessment of digital supply chains, McKinsey reported that predictive analytics in demand planning could often reduce forecasting error by 30 to 50 percent. The discussion concerned models using internal and external demand variables. That range is McKinsey's historical assessment, not an Assuras result or a forecast of what a particular implementation will achieve.
Kearney's 2019 Views from the C-Suite survey found that 50 percent of executives who already identified technology adoption as a leading opportunity selected AI and machine learning as their top opportunity. The denominator matters: this was a subgroup of respondents, not all global executives. It records expectations at that time, rather than demonstrated supply-chain performance.
Five technologies, five investment questions
Automation and robotics: is the task ready?
Warehouse automation can support activities such as moving goods, sorting and picking. Assess the task's volume, variation and exceptions before selecting equipment. A proposed change also needs a safe working method, maintenance ownership and a way to keep work moving when equipment is unavailable. Automating an unstable process can leave the underlying constraint unresolved.
IoT and sensors: who acts on the signal?
Connected sensors can supply information about location, condition or equipment status. Start with the action that information should trigger. A temperature alert, for example, needs a responsible recipient, an agreed response and a record of what happened. Check coverage, data freshness and sensor reliability before describing visibility as complete or real time.
Predictive analytics: is the forecast useful enough?
A demand forecast is an estimate. Evaluate a proposed model against a relevant baseline and the decision it supports. For inventory planning, agree the product level, planning horizon and error measure. Examine whether performance differs across products or operating conditions, and record when a planner should review the model's recommendation.
Blockchain: is a shared ledger the right tool?
Blockchain can support a shared, tamper-evident record, but it does not establish that the original input was true. NIST's 2018 Blockchain Technology Overview explains that inaccurate sensor readings or false human inputs can enter a blockchain. For supply-chain use, assess the need for shared records alongside partner participation, access, verification and correction processes. A conventional shared database may meet the requirement.
Cloud computing: can partners use the information?
Cloud applications can provide shared access to planning and operational information. The implementation still needs agreed identifiers, permissions, interfaces and support responsibilities. Assess availability, data portability and the cost of continued operation as well as the initial configuration.
Define the operating benefit before the pilot
Choose a bounded process and establish a baseline. If the problem is an unreliable delivery promise, trace the information that shaped the promise and the handoffs that followed it. If the issue is inventory, distinguish inaccurate records, demand uncertainty and replenishment delays. Each calls for a different response.
Set a small set of measures tied to the problem: forecast error at the chosen horizon, inventory-record accuracy, the age of open exceptions or delivery against the agreed promise. Include service quality and employee feedback so an apparent improvement in one measure does not hide work moved elsewhere.
At the pilot review, decide whether to extend, revise or stop. Make that decision using observed results, unresolved dependencies and the team's ability to operate the change. Broader technology adoption is a commitment to a working system of people, information and responsibilities.
Explore Assuras's business practice efficiency and technology solutions approaches, or read how to connect analytics with a business decision.
