Sentiment in Bigdata.com: The Trend and the Story

July 20, 2026

Market Data, MCP, Sentiment/7 minutes read

TL;DR

Bigdata.com ships two sentiment tools from RavenPack’s research stack: a daily trend signal (REST API: backtesting, monitoring, alerts) and a real-time tearsheet (zero code, inside Claude, ChatGPT, and Copilot).
They answer different questions, what’s the trend? versus what’s driving it right now?: two complementary horizons on the same name, built to be used together.
Every chart and number below comes from live API calls you can reproduce today.

At RavenPack, we use multiple sentiment metrics. Our EDGE Analytics engine computes a whole family of them, each tuned to a different research question. Bigdata.com adds its own native layer: chunk-level sentiment, a score on every document chunk from a transformer trained on financial news and corporate event coverage. It measures likely market impact, not linguistic tone, from -1 to +1. This post covers the two tools built on those foundations: a daily trend signal for backtesting and monitoring, and a real-time read on what’s moving a name today. Here’s the trend lens on Meta over the trailing year: sentiment_pressure converted into a volatility-scaled band around price:

Meta price with sentiment-pressure ribbon, trailing 12 months. Sentiment regime flipped negative on Oct 29, one session before an 11% drop

Green is positive sentiment pressure, red is negative, and the band’s width scales with the stock’s own volatility. Meta’s sentiment_pressure turned negative on October 29, one session before an 11% drop, and stayed red through the entire drawdown. One flip doesn’t make a strategy, but the signal is daily, point-in-time, and years deep: whether it repeats across your universe is a testable question.

Two tools, two questions

	Entity Sentiment API	Sentiment Tearsheet
Question it answers	”What’s the trend over time?"	"What’s driving sentiment right now?”
Update cadence	Daily, historical depth	Real-time, 90-day rolling window
Data pipeline	EDGE Analytics: events, relevance, novelty, source ranking	Bigdata.com chunk-level sentiment: news-only, recency-decayed evidence
Best for	Backtesting, portfolio monitoring, programmatic alerts	Morning briefings, “why did this move,” analyst research
How you use it	A few lines of Python against a REST endpoint	Ask Claude, ChatGPT, or Copilot in plain language
Audience	Developers, quants, data engineers	Analysts, PMs, anyone who wants an answer without writing code

A trend signal that reacted to every headline would be too noisy to backtest. A real-time read refreshed once a day would miss the story as it develops. Two jobs, two signals, one research stack.

Tool 1: Entity Sentiment API, the trend

The /v1/entity-sentiment/ endpoint returns three signals per entity per day: daily_sentiment (today’s weighted tone), sentiment_pressure (is the 90-day trend unusual for this company), and abnormal_media_attention (is coverage volume unusual). Reach for it to look back over months or years, or to backtest against price. Daily sentiment is the day’s weighted average of event sentiment, weighted by (relevance/100)² × (novelty/90)²: a highly relevant, novel event counts far more than a marginal or repetitive one. Sentiment pressure z-scores the 90-day moving average against the company’s own trailing year, normalized to [-1, 1]. Beyond ±0.9 is statistically unusual. It answers: is the recent trend better or worse than this company’s norm? Abnormal media attention applies the same treatment to coverage volume instead of tone: a bad story spreading and a product launch both spike it, for opposite reasons. The full EDGE Analytics pipeline (relevance and novelty weighting, bias correction, subsidiary rollup, z-scoring, normalization) ships as one endpoint. None of it is yours to maintain.

import httpx
import pandas as pd

response = httpx.post(
    "https://api.bigdata.com/v1/entity-sentiment/",
    headers={"X-API-KEY": "YOUR_API_KEY"},
    json={
        "identifier": {"type": "rp_entity_id", "value": "4A6F00"},   # Alphabet Inc.
        "timestamp": {"start": "2025-07-20", "end": "2026-07-20"},
    },
)

df = pd.DataFrame(response.json()["results"]["values"])

One year of daily signals in a DataFrame. The tail, as returned today:

date	daily_sentiment	sentiment_pressure	abnormal_media_attention
2026-07-17	0.121	0.086	0.794
2026-07-18	0.150	0.063	0.792
2026-07-19	0.218	0.071	0.777

identifier.value also accepts a list: one call covers your whole watchlist.

The overlay, across a basket

The same ribbon across eight names over the trailing 12 months, each panel annotated with the correlation between sentiment_pressure and price:

Price with sentiment-pressure ribbon for an 8-company basket, trailing 12 months, with per-company correlation

band(t)    = price(t) × (1 + scaling(t) × sentiment_pressure(t))
scaling(t) = VOL_MULTIPLIER × ewma_volatility(t, halflife=10 days)

Volatility scaling keeps the band proportionate whether a stock is quiet or choppy. Meta tracks tightest (r = +0.81); Netflix, Oracle, and Tesla follow (+0.60, +0.55, +0.55). Oracle’s ribbon flips red at its September peak and stays red through the slide; Microsoft sits at +0.23. The relationship is company-specific: measure it per name before you lean on it. Daily history is what makes that possible.

Continuously monitoring private companies

Private companies in the Bigdata.com knowledge graph get the same signals, driven by funding rounds, launches, and regulatory scrutiny instead of earnings cycles. Watch abnormal media attention around SpaceX’s IPO:

SpaceX Abnormal Media Attention, year to date 2026, with the June 12 IPO marked

Coverage ran well below SpaceX’s norm in January, crossed above normal in April, and hit the abnormal band before the June 12 listing. The build-up preceded the IPO by weeks, and the signal ran straight through it: the entity ID doesn’t change when a company lists. Same call as any public name:

json={"identifier": {"type": "rp_entity_id", "value": "335CF4"},   # SpaceX
      "timestamp": {"start": "2026-01-01", "end": "2026-07-20"}}

Twenty years of one company’s sentiment

One request covers up to 3,650 data points, about 10 years per entity; stitch calls together for more. Apple’s price with its sentiment-pressure ribbon, weekly, 2006 through today:

Apple Inc. price with sentiment-pressure ribbon, weekly, 2006 to 2026, log scale

The red troughs line up with the real drawdowns: 2008, the early iPhone-era dips, 2013, 2016, the 2018 selloff, COVID, the 2022 bear market. That depth makes the signal testable rather than anecdotal: backtest across regimes, don’t eyeball one good call.

Tool 2: Sentiment Tearsheet, what’s driving it right now

The trend API looks back. When you need to know why a stock is moving today, you want the news itself, ranked by what’s driving the current reading, with a narrative that ties it together. That’s the Sentiment Tearsheet. It lives inside Claude, ChatGPT, and Microsoft Copilot through the Bigdata.com MCP connector. No API key, no code: ask “what’s driving sentiment on SpaceX right now?” and it answers. Its building block is the chunk-level sentiment introduced above, aggregated into a company-level signal. The same scores come back on every Search API result, so anything the tearsheet summarizes you can drill into chunk by chunk. Three design choices keep the signal clean. SEC filings and transcripts are excluded: boilerplate carries systematic sentiment bias, not news. Near-neutral coverage is dropped and strong coverage is weighted by squared magnitude, so the score is decisive. And evidence is ranked by contribution to today’s reading (relevance, magnitude, recency), not raw relevance.

SpaceX, called just now

We saw SpaceX through the trend lens earlier. Here it is through the real-time lens, today. Everything below is the tool’s actual output.

Space Exploration Technologies Corp.SPCX · 90-day lookback

Sentiment −0.30 · slightly negativemomentum −0.13 · strong downwardsentiment z (1mo/1qt) −1.9 / −2.2 · unusually lowattention −45.8% · z −1.4 / −0.2

Summary

SpaceX’s post-IPO equity narrative is dominated by near-term Starship execution setbacks and the resulting market repricing. Multiple reports tie the stock slide to the postponed and aborted Flight 13 and engine ignition issues. Coverage also carries a genuine counterweight in AI compute and supply-chain themes, but the current regime is negative and deteriorating: -0.303 against a -0.175 baseline, momentum -0.129, and a quarterly z-score of -2.2, beyond the ±2 threshold the tool flags as statistically unusual.

Bullish

Foxconn reported to secure AI server manufacturing order from SpaceX (about USD 52B)
Quiet build-out of an AI compute rental business: Google, Anthropic, Pentagon deals
Dip-buying framing emerging: “down 45% from its high, history suggests…”

Risk / Watchlist

Starship Flight 13 aborted after four engines failed to ignite; retry set for Jul 23
Stock below its USD 135 IPO price; about USD 1T of market value lost from the post-IPO peak
FT: traders raising bets against SpaceX; ~30% of the public float shorted

Outlook

The unresolved tension: can Starship execution recover fast enough to restore the post-IPO valuation trajectory, or does the market keep discounting execution variability? Evidence supports both sides; no specific forward view.

Evidence, ranked by contribution to today’s reading (relevance x |sentiment| x recency decay)

Headline	Source	Date	Sent	Contrib
SpaceX stock wobbles after Starship launch postponement	InvestingCube	Jul 20	−0.79	−0.53
Weeks after record IPO, SpaceX loses USD 1 trillion in market value	Times of India	Jul 19	−0.80	−0.47
SpaceX stock is down 45% from its high; history suggests…	Nasdaq	Jul 20	+0.83	+0.44
Foxconn secures AI server contract manufacturing order from SpaceX	Wallstreetcn	Jul 20	+0.82	+0.38
Four engines fail to ignite, forcing SpaceX to abort Starship launch	MSN	Jul 19	−0.70	−0.33
Beyond rockets, SpaceX is quietly renting AI compute	Fortune	Jul 19	+0.69	+0.32
SpaceX stock falls below its USD 135 IPO price	Wealth Professional	Jul 20	−0.55	−0.26
Traders are increasingly betting against SpaceX	Financial Times	Jul 18	−0.61	−0.20

Contrib: contribution to the current reading, computed as relevance x |sentiment| x recency decay. A month-old FT story about the USD 400B value drop scores near zero here; this weekend’s launch abort dominates.

Read the two horizons together. The trend lens says the 90-day regime is solidly positive: months of pre-IPO enthusiasm are still in that window, and one bad weekend doesn’t undo a trend. The real-time lens says the news cycle turned sharply this weekend, and shows you the exact articles. The trend tells you where the name stands; the tearsheet tells you what’s moving it now. Before the open, you want both.

How to try it

No setup beyond connecting Bigdata.com to your assistant of choice:

Tool reference: Sentiment Tearsheet
Connect it to Claude: Claude MCP integration guide
Connect it to ChatGPT: ChatGPT MCP integration guide
Microsoft Copilot and other agentic platforms: MCP introduction

Once connected, just ask. Developers can call it programmatically (find_securities, then bigdata_sentiment_tearsheet), but the point is that you don’t have to.

Choosing between them

Backtest or screen a factor

Use the Entity Sentiment API. Pull sentiment_pressure historically across your universe, join it to price, and test it like any other signal.

Monitor a book programmatically

Use the Entity Sentiment API. Daily batch pulls, rank by sentiment_pressure change, flag |sentiment_pressure| > 0.9 for review.

Why did this stock move today?

Use the Sentiment Tearsheet, inside Claude, ChatGPT, or Copilot. Ask in plain language, get the narrative and the sources behind it in seconds.

Morning briefing on a watchlist

Use the Sentiment Tearsheet. No code, no API key: just ask your assistant what’s driving sentiment on each name before the market opens.

Try it on your own book

Everything above came from live API calls. The fastest way to evaluate the signals is against names you actually cover.

Zero-code start: connect the Bigdata.com MCP to Claude or ChatGPT and ask for a sentiment tearsheet on any name in your portfolio. Two minutes, no key.
API access: request an API key and pull sentiment_pressure history for your universe; the Entity Sentiment API reference has the full schema.
For bulk historical pulls across large universes, see the Smart Batching guide.

This content is for informational purposes only and does not constitute investment advice.

Francisco Gomez

Manager of Applied Intelligence (VP)

Andres Diaz

Senior Structured Data Manager

August - 2026

July - 2026

June - 2026

May - 2026

April - 2026

March - 2026

Two tools, two questions

Tool 1: Entity Sentiment API, the trend

The overlay, across a basket

Continuously monitoring private companies

Twenty years of one company’s sentiment

Tool 2: Sentiment Tearsheet, what’s driving it right now

SpaceX, called just now

How to try it

Choosing between them

Backtest or screen a factor

Monitor a book programmatically

Why did this stock move today?

Morning briefing on a watchlist

Try it on your own book

​Two tools, two questions

​Tool 1: Entity Sentiment API, the trend

​The overlay, across a basket

​Continuously monitoring private companies

​Twenty years of one company’s sentiment

​Tool 2: Sentiment Tearsheet, what’s driving it right now

​SpaceX, called just now

​How to try it

​Choosing between them

Backtest or screen a factor

Monitor a book programmatically

Why did this stock move today?

Morning briefing on a watchlist

​Try it on your own book

Two tools, two questions

Tool 1: Entity Sentiment API, the trend

The overlay, across a basket

Continuously monitoring private companies

Twenty years of one company’s sentiment

Tool 2: Sentiment Tearsheet, what’s driving it right now

SpaceX, called just now

How to try it

Choosing between them

Try it on your own book