Network congestion and fee spikes on public blockchains are most reliably forecast by observing a combination of on-chain supply and demand indicators. Multiple academic and industry sources converge on a handful of metrics that show strong predictive power when tracked together.
Core predictive metrics
The size of the mempool or pending transaction backlog is the clearest short-term predictor because it directly measures unmet demand for block space. Gas price percentiles such as the 50th and 95th percentile gas price show how much users are bidding to get included and reveal upward pressure before an extreme spike. Gas used versus gas limit or block fullness signals utilization and how close the network is to capacity; persistent fullness often precedes prolonged fee elevation. Transactions per second and change in active addresses capture demand growth across use cases, while sudden jumps in smart contract events, decentralized exchange volume, or NFT mint transactions often coincide with acute congestion. Individually these metrics can be noisy, but together they form a robust signal.
Causes and consequences
Fee spikes are typically caused by transient demand surges from coordinated activity such as token launches, airdrops, or popular NFT drops, and by sustained demand increases from DeFi activity and high-frequency bot behavior. Vitalik Buterin at the Ethereum Foundation has explained how fee market mechanics, including the EIP-1559 base fee algorithm, tie base fees to demand and can amplify visible fee changes during volatility. Arvind Narayanan at Princeton University and other researchers studying blockchain data emphasize that fee dynamics reflect not only technical load but also economic incentives and behavior patterns among users and miners. Consequences include reduced accessibility for small-value users, shifts toward layer 2 solutions, displacement of low-margin services, and regional equity impacts where users with limited means or high on-chain exposure bear disproportionate cost.
Interpreting metrics for action
For monitoring, combine real-time mempool depth with percentile fee curves and block utilization trends to detect both imminent spikes and longer-term stress. Watch for rising 95th percentile fees paired with growing mempool age as an indicator of sustained congestion rather than a momentary blip. Operational responses range from delaying noncritical transactions to adopting fee estimation tools and using layer 2 or alternative chains when on-chain costs become prohibitive.