About IPOSignal
We write IPO research that answers the questions investors actually ask — not just lists of data with no explanation of why it matters.
The gap we kept seeing
Plenty of IPO sites share filings, calendars, and headline numbers. Useful, but rarely enough. They often skip the questions that matter most: Does this factor actually predict returns? Is that popular strategy still working, or has the market already priced it in? What happens around lock-up expiry — and is the sample large enough to trust the answer?
Raw data without context is easy to find. Honest, tested explanations are not. That is why we built IPOSignal around research — articles, charts, and tables that show not just what the data says, but why you should care.
How we work
Question first
We start with a real investor question, not a data dump. Every piece of research should change how you think or act.
Fact-check the narrative
Common beliefs about IPOs deserve scrutiny. We test assumptions instead of repeating them — what sounds sensible is not always what works.
Respect the data
Financial series are noisy, skewed, and often dominated by a few extreme outcomes. We account for sample size, heavy tails, and outliers — not just averages.
Show the proof
Claims come with evidence: charts, tables, and clear methodology. If we cannot support it, we do not publish it as fact.
Strategies that seem intuitive on paper can fail in practice because edge gets arbitraged away. Our job is to measure what is still there — and to be upfront when the evidence is weak, mixed, or counter-intuitive at first glance but sensible once you dig into the details.
How we treat the numbers
IPO returns and valuation multiples are heavy-tailed. A handful of extreme pops, blow-ups, or one-off re-ratings can dominate a simple average or tilt a trendline through the wrong part of the chart. That is not a bug in the data — it is how markets behave — but it does mean naive summaries can overstate a signal or make a random event look like a repeatable pattern.
On Analytics, we use industry-standard robust techniques so charts and tables reflect durable structure rather than a few outliers:
- 5% winsorization — extreme values in the top and bottom 5% are clamped to the boundary, not dropped. Tail events still count, but they cannot dominate means or scatter coordinates. This matches the winsorized mean columns marked with * in factor and event tables.
- Medians and win rates — alongside means, we show medians and the share of positive outcomes. A factor can have a positive median but a negative winsorized mean (or vice versa) when the distribution is skewed; seeing both reduces false confidence.
- Robust trendlines (Huber regression) — scatter plots default to a Huber fit after winsorization, which locks the line onto the dense cluster (“core mass”) instead of being pulled by remaining high-leverage points. You can switch to winsorized OLS for a broader market-direction read.
- Non-parametric tests — where we compare groups (e.g. bucket vs bucket), we prefer Wilcoxon and Mann–Whitney over t-tests when sample sizes are modest, because IPO return series are skewed and far from normal.
- Sample size upfront — every breakdown shows N. Small buckets and thin factors are labeled as such; we would rather show “not enough data” than imply precision from a handful of names.
These choices are deliberately conservative. They make it harder to “discover” a factor that only worked because of one memorable IPO — and easier to spot relationships that still show up after tails are tamed. When we publish research on the Insights Hub, the same philosophy applies: tested claims, visible methodology, and honest limits.
What you will find here
On the Insights Hub, we publish research grounded in that approach: borrow-rate buckets vs forward returns, event windows around lock-up and quiet period, factor performance, and other questions IPO investors run into. On Analytics, you can explore the underlying return and win-rate data yourself.
Some results will confirm what you suspected. Others will surprise you — until you see the breakdown by sample, time period, or tail behavior. That is the point. Better decisions come from tested ideas, not recycled conventional wisdom.
