Thread by @FabiusMercurius

🤖What "alternative data" do hedge funds use?🤖

Dataminr just raised $475M at $4.1B valuation. 👀

For the last 10 yrs, hedge funds swarmed at "alternative data" like pigeons at bread crumbs. Why? What exactly constitutes alt data & does it actually generate investment alpha?
👇

1/ What is it?

"Alternative data" is just a fancy term for data that doesn't appear on a 10-K/Q or earnings transcript.
Things like foot traffic at a retail store or credit card orders at a restaurant. Such data helps hedge funds predict earnings better (supposedly, at least).

2/ What are some common data sources?

App usage-- # of downloads over time is used to predict MAUs/DAUs & forward adoption rate; reviews also used to gauge product quality

Supply chain & logistics-- used to predict inventory & sales bottlenecks, pricing/bargaining power, etc.

Consumer transactions-- (e.g. credit card & email receipts) used to predict store-level & product-level revenue growth before quarterly earnings releases

Parking lot data-- to predict revenue growth at hotel, food, & entertainment chains

ESG-- used as early indicators of risks

Sentiment & NLP on wallstreetbets-- they wanna know the next 🦍🔥🚀 meme stonks

Satellite & drone imagery-- used to track industrial production (esp. in data-scarce 3rd world), oil inventories, agriculture harvests ...

Web crawlers-- used to track pricing & product distribution

3/ What data is underrated & under-tracked that hedge funds absolutely (should) use?

Cyber.
Especially short funds.

All these intrusion detection systems (Crowdstrike, Darktrace, Dragos) & automated pentesting tools (Metasploit, AttackIQ) & scanners (Qualys, Rapid7) etc... 🤔🤔

4/ Who provides alt data?

See diagram👇 for market map.
Data flow looks like this:
You+me/weather/some cornfield ➡️ aggregator➡️ cleanser/modeler➡️trader

Data providers are always aggregators & maybe cleansers/modelers.

Most hedge funds do their own (re)-cleansing/modeling.

5/ How much do hedge funds spend on alt data?

2020: Hedge funds spent $1.7B (in aggregate) on alt data (i.e. NOT THAT MUCH)
2019: 1/3 of funds spent >$1M/yr
Annual subscriptions range from $25K-$500K+
30% of transaction datasets cost >$150K/yr
40% of ESG datasets cost <$25K/yr

6/ Does it work?

So does alt data actually help generate investment alpha?
I wish I had stats to compare <returns of funds using alternative data> vs. <returns of funds NOT using>.
Sadly I don't. (Does anyone?)

The quick answer is that **some** datasets are truly useful.
Which?

1. A good wand needs to meet the right wizard. Data is only useful when an investor knows how to interpret it & knows its major sources of bias.
2. Alpha-rich data is not buyable by literally every bloke on the Street!
3. ESG is low demand, i.e. some low-signal charity donation.

4. Reread 1 & 2. It means do ur own data collection. Build ur own scrapers. That way u understand all sources of bias, and when things go wrong u can identify those hindsight-obvious factors you omitted.

5. If you can't build sh*t, at the very least sign an exclusive contract.