1. Define the universe and observation window
Write down the ticker or ticker list, session date, time zone, expirations and query cutoff. Record whether the source is a trade feed, an aggregated alert, or a contract snapshot. A result capped at a fixed number of rows is not a complete market ranking.
Use a premium threshold as a declared filter, not a rule that separates professionals from retail traders. Note the excluded rows and missing values. If a source does not supply a prior-period baseline, report that absence instead of calling the activity three times normal.
2. Validate the fields and calculate like-for-like measures
For each result, retain its contract identifier, call or put, strike, expiration, timestamp, price, size and quote reference when available. Check whether a row represents a single execution or an aggregate. Adding both an aggregate alert and its component prints can double-count the same activity.
For standard 100-share contracts, premium equals price times contracts times 100. Check the multiplier for adjusted contracts. Compare contract volume with contract volume and premium with premium; a dollar threshold and a contract-count threshold select different observations.
3. Treat direction as a hypothesis
Ask-side or bid-side labels estimate the execution’s relation to a quote. They do not establish the buyer’s full portfolio, whether the trade opens or closes exposure, or whether it belongs to a spread. A midpoint execution can remain unclassified.
Pineify’s Market Insights flow view uses reported ask-side and bid-side premium when available. Its sentiment label uses a greater-than-60% share of total premium: ask-side calls and bid-side puts are labeled bullish; the reverse combinations are labeled bearish. Other cases remain neutral. That is a classification rule, not a 60% probability of profit. Read the underlying fields before turning the label into a conclusion.
4. Add context without inventing corroboration
Check the earnings calendar, expiration, contract liquidity and available price history for the same observation time. A stock print in a separate dark-pool feed does not confirm that the same investor made the options trade. It also does not independently establish a buy or sell intent.
Use comparable historical windows only when the source supplies them. A short interval before earnings and a full quiet session are not interchangeable baselines. Open interest is a clearing-based outstanding count, so it cannot by itself classify each intraday execution as opening or closing.
5. Write a conclusion that can be checked later
A useful note states the observed activity, sample limits, interpretation and alternative explanations. For example: “In the returned sample, call premium was concentrated in one expiry; the feed labels most of it ask-side. Opening status and other strategy legs are unknown.” This is an illustrative research note, not a live result.
Save the observation time before reviewing subsequent prices. If testing a rule, define the entry, exit, available-data delay and transaction costs before measuring results. Include rejected and losing observations. A favorable screenshot selected afterward does not measure the rule’s performance.