Reliable analysis starts with data lineage: identify provider, venue, unit, window, timestamp, and transformation before interpreting a number.
Trace every number to its source
Market data is not self-explanatory. A useful checklist is provider, venue coverage, quote currency, observation window, update time, and calculation. The market indicators endpoint requests CoinGecko's coins-markets data in USD, ordered by market capitalization, with one-hour, 24-hour, and seven-day percentage changes. It also uses current price, market cap, reported total volume, and the 24-hour high and low.
CoinGecko aggregates market information under its own methodology, while the DeepBook indexer describes specific on-chain order books and pool summaries. Two values bearing the same name can differ because their market coverage is different. Preserve the source label rather than blending an aggregate price, one venue's depth, and another provider's volume into a supposedly seamless measurement.
Understand derived fields
Several useful fields are transformations. Range position is (price − daily low) / (daily high − daily low) × 100. Volume intensity is total volume / market cap × 100. The latter is an activity proxy, not TVL and not a guaranteed comparable turnover ratio across every token. Missing or zero values also need care: a default can keep a screen usable while reducing analytical certainty.
Home pressure is another derived field. Conditions involving momentum, range location, and volume modify an internal score; the display is clamp(50 + score × 8). It is a heuristic scale, not a percentage of aggressive buys, not exchange order flow, and not a probability. A clear data dictionary should keep raw observations separate from these labels.
- Raw field: provider-supplied current price or reported volume.
- Derived field: range position, volume/market-cap ratio, or pressure.
- Interpretive label: bullish, bearish, compressed, or another rule output.
Manage latency, missingness, and scope
APIs can rate-limit requests, return missing fields, or change between polls. The CoinGecko market endpoint caches responses for five minutes, with an additional 30-second client freshness window. DeepBook retrieval can fall back to a last-known-good snapshot with no maximum age, or default values when no snapshot exists. Those values may be stale or synthetic rather than live. Verify source timestamps and status independently when needed; an unchanged screen or successful request does not establish fresh market data.
Order-book data has additional limits. The DeepBook calculation sums price × quantity for visible levels, so it measures displayed bid and ask notional, not long/short positions. It does not capture hidden intent, future cancellations, or liquidity on other venues. Likewise, a pool's 24-hour quote volume is transactional flow, not actual TVL. Accurate naming is part of accurate analysis.
Illustrative data audit
Suppose a hypothetical asset record has the following USD fields. Audit the transformation before drawing a conclusion.
- Raw market fields
- Price $2.40; low $2.00; high $2.50
- Range position
- ($2.40 − $2.00) / ($2.50 − $2.00) = 80%
- Activity inputs
- Volume $30m; market cap $200m
- Volume intensity
- $30m / $200m = 15%; not TVL
The asset is high in its daily range with substantial reported activity relative to market cap. That is a bounded description of the snapshot, not evidence of future direction or capital locked.
Common interpretation traps
- Combining data from different timestamps or venues without disclosing the mismatch.
- Treating volume divided by market cap as TVL or exact economic turnover.
- Presenting cached, missing, or transformed values as direct live measurements.
Frequently asked questions
Why does CoinGecko data differ from an exchange screen?+
CoinGecko aggregates markets according to its methodology, while an exchange shows its own instruments and liquidity. Quote currency, timing, and coverage can all differ.
What should I do with a missing field?+
Mark it unavailable and lower confidence. Do not silently convert missing data into a meaningful zero unless zero is genuinely defined by the source.
Is total volume comparable for every asset?+
Not perfectly. Coverage, pair composition, liquidity quality, and provider methodology vary, so use it as context and inspect outliers before ranking assets.