Three questions stand out. Each one uses a dimension the data actually has: 152 tickers, daily open/high/low/close/volume, and a span from 2012-01-02 to 2025-08-04. ## 1. Did the AI surge change price level and risk, and did it do so evenly across companies? - **Why it's interesting:** The dataset is framed around the "AI surge", and 99,984 of the 422,729 rows (23.7%) fall after ChatGPT's launch. That is a large post-period for a before/after comparison. - **How to answer it:** For each company, compare price growth, volatility and volume after versus before late 2022. Then rank companies by the change. The gold tables `semi_conductor_se_by_company_name_date__yyyy` and `..._yyyy_mm` already hold these aggregates per company and year or month. - **What it could show:** Whether the gains were concentrated in a handful of AI-exposed names or spread across the sector. That matters for deciding what to overweight, and for which companies belong in a sector basket. - **Caveat:** A date split shows association, not cause. The dataset has no AI-exposure field, so you would have to define exposure yourself. ## 2. How does volatility differ across companies, and is it rising? - **Why it's interesting:** The average daily high-low range is about 3.69% of price across the whole dataset. That is a useful baseline, and a per-company, per-year range against it would show which names are riskier and whether risk has shifted over time. The result could feed position sizing. - **How to answer it:** Compute (high − low)/close by company and year, then compare each company's range to the 3.69% sector average. - **Caveat:** Daily range is a rough proxy for volatility. Return-based measures (standard deviation of daily returns) would be stricter, and both can be computed from this table. ## 3. When does trading volume signal unusual interest, and what happens next? - **Why it's interesting:** Volume is extremely skewed. The median is 898,500 shares, the mean is about 8.47 million, and the maximum is about 3.69 billion. Spikes are therefore rare and stand out, and they can be tested as timing signals. For example, you could ask whether days with volume several times a stock's own norm lead or lag large price moves. - **How to answer it:** Flag days where volume is well above a stock's own rolling median. Then look at the returns over the following days and compare them with normal days. - **Data-quality caveat:** 4,888 rows have zero volume. These are probably halted or illiquid days, or missing data, and should be excluded or handled separately before building any signal. ## Caveats across all three - The card lists about 149 distinct company names, but my profiling counted 152 companies and 152 tickers. Check how names map to tickers before ranking by company. - The `company_name` and `stock_name` columns differ, and the grain is one row per stock per day. - This is observational price and volume data only. It has no fundamentals, AI revenue or exchange fields, so the questions can show what happened, not why.