Most research desks tracking on-chain liquidity are primarily watching 24-hour trading volume. It is the most visible metric, easy to obtain from any block explorer or aggregator API, and it has clear intuitive meaning: a pool with higher volume is more active. The problem is that volume is a lagging indicator of liquidity quality, and for execution cost assessment it is close to useless. Depth is the metric that matters, and the two can diverge dramatically.
This is worth being specific about because the conflation is not careless. Volume and depth are often correlated over long time horizons, so using volume as a liquidity proxy does not produce obviously wrong answers most of the time. The divergence becomes a real problem at specific points: after large volume events that drain one side of a pool, during LP withdrawal periods, and for long-tail assets where a single liquidity provider contributes the majority of depth.
What volume measures and what it does not
Trading volume on a DEX pool records the aggregate notional of all swaps that settled in a given period. It is a historical count. A $10M 24-hour volume figure means $10M worth of tokens changed hands in that pool over the past day. It says nothing about whether the pool can absorb a $500k trade tomorrow at an acceptable price impact.
Depth, in the sense we are using here, is the amount of liquidity available at a specific price impact threshold. Depth at 1% means: how large a trade can the pool absorb before price moves by 1%? This is a forward-looking execution constraint. It is what a treasury analyst needs before deciding whether to execute on-chain versus through a CEX. It is what a routing algorithm needs to decide between two routing paths. Volume provides neither of these things.
The clearest illustration of the gap: a pool that had $8M in volume on a single high-volatility day may now have significantly reduced depth if that volume was directional. If a large fraction of that $8M was one-sided (all selling, for instance), the reserve ratio has shifted and the pool depth on the sell side is now much thinner than it was before that session. The volume figure remains high, the depth figure has declined. Researchers using volume as a liquidity proxy will overestimate how much the pool can handle tomorrow.
The concentration mechanics behind the depth-volume gap
For concentrated liquidity AMMs like Uniswap v3, the gap between volume and depth can be especially pronounced. In a v3 pool, liquidity providers concentrate their positions around the current price range. When the price is within a high-concentration band, depth is very high relative to TVL. When the price moves outside that band, depth drops sharply even if volume on the day was high, because the high volume was generated within the old price range while most LP positions have not been repositioned to follow the price.
This matters for any asset that has experienced a significant price move recently. High-volume pools that traded heavily during a price trend may now be sitting outside the concentrated liquidity band of their largest LPs. The TVL is still there (the tokens are still in the pool), but effective depth at the current price is a fraction of what it was at peak concentration. Volume data does not reveal this. Only direct depth measurement does.
Where the research error actually shows up
The practical error mode is in execution cost estimation for moderately sized trades. Consider a research desk benchmarking on-chain execution quality for a set of token pairs across chains. If they rank pools by volume and assume higher-volume pools have better execution costs, they will produce a ranking that is right on average but wrong in the cases that matter most: unusual market conditions, post-stress periods, and tail assets.
We see this pattern specifically in how research desks assess routing quality for cross-chain transfers. A common approach is to look at which pool has the highest 7-day volume and assume that pool has the best execution depth. For major pairs during stable periods, this is usually approximately correct. For anything involving a bridge-adjacent pool or a chain with concentrated LP behavior, the volume-depth correlation weakens substantially.
The cases where this produces materially wrong analysis tend to cluster: assets that recently experienced high volatility (volume went up as depth went down on one side), newly launched pools that have not yet accumulated stable LP behavior, and any pool with a dominant single LP who may have added or withdrawn position recently.
How to use depth measurements correctly
Depth measurement for execution quality assessment should specify the notional threshold you care about. Depth at 0.5% price impact and depth at 3% price impact are different numbers and serve different questions. For most institutional execution contexts, depth at 1% and 2% impact thresholds are the relevant benchmarks: they correspond to the execution quality range where a real trade of moderate size would actually land.
The frequency of depth observation matters too. A single depth snapshot at the start of the trading day gives you a point estimate that can be stale within hours. For research purposes, a time-series of depth at regular intervals over a trading session gives you something more useful: a sense of depth stability, which itself is a distinct quality dimension. A pool that has consistently high depth is more reliable than a pool whose depth oscillates around the same average, even if the averages are equal.
For comparative analysis across pools or chains, depth-to-TVL ratio is more informative than volume-to-TVL. It reveals how efficiently the pool is deploying its capital toward executable depth. Concentrated liquidity pools can score highly on this ratio. Passive full-range liquidity pools tend to score lower. Stableswap pools have a structurally different ratio altogether because the math of the curve concentrates effectively depth near the peg. Comparing these without accounting for pool type is another common source of analytical error.
What volume is still useful for
This is not an argument that volume is worthless. Volume is the right metric for protocol health, fee revenue analysis, market participation trend identification, and historical activity tracking. If you want to understand whether a protocol is growing or declining, volume time-series is the appropriate input. If you want to understand whether a specific pool can absorb a specific trade at an acceptable cost, it is not.
The more useful framing is to treat volume and depth as answering different questions: volume answers "what happened in this pool," and depth answers "what can this pool do right now." For research desks focused on execution quality, the second question is the operative one, and it requires a different data workflow to answer properly. Volume is available from almost every block explorer and aggregator for free. Depth at specific impact thresholds is a computed metric that requires continuous pool state indexing, which is precisely why so many desks still default to volume. The data workflow is harder, not the concept.