Market data
Stock social sentiment: investigating mention spikes
Read the baseline, sample and duplicated posts behind a social activity alert before connecting it with a company or price move.
A count describes the collected sample
A mentions alert says something changed in the posts a service collected. Before interpreting it, ask which sources are included, which language is counted and how company names are matched to tickers. A familiar word can also be a ticker symbol. Counts produced by different collection rules should not be treated as measurements of the same population just because they share a label.
Inspect the denominator behind the multiple
Suppose a fictional feed counted 20 matching posts per day on average over seven prior days, then counted 80 in the latest comparable 24-hour window. The ratio is 80 / 20 = 4×. If the average had been only two, eight posts would also give 4×. Both the raw count and baseline therefore belong beside the multiplier. With a zero baseline, the ratio is undefined; display the counts instead of inventing a finite multiplier.
Repeated posts are not independent observations
Imagine those 80 collected posts contain 50 reposts of one original headline. A raw activity total may still be useful, but it is different from 80 independently authored reports. Keep the raw count, deduplication rule and unique count distinct. Removing identical text does not prove that the remaining accounts are independent or human. An alert should explain what was measured rather than silently implying that a large crowd investigated the company.
Positive text is not an outcome probability
If a classifier labels 90% of a selected set of posts positive, the percentage describes those labels under that model and sampling rule. It is not a 90% probability that the share price will rise. Sarcasm, quotations and a noisy ticker match can complicate interpretation. Read a few underlying examples and compare the time window with the original company announcement before attaching a causal story to a price chart.
Check the underlying claim and its source
The SEC warns that social-media stock recommendations can be part of investment scams. That does not mean every mentions spike is manipulation; the count alone cannot establish either authenticity or fraud. Open the original material, identify what can be independently checked and separate a company disclosure from a promotional claim. Popularity is a reason to inspect the evidence, not evidence that the proposed trade will work. SEC Investor.gov: Social Media and Stock Tip Scams (February 6, 2026)
A compact note for a GSDAQ alert
Record the ticker match, collected sources, observation window, raw count, comparison baseline and original link. Mark fields that the available alert does not provide as unknown. Then write the research question the alert raised. Keeping that note alongside the company’s filings and news gives the activity number a useful role without making it carry a forecast it was never designed to establish.