Our last post explained the mechanics behind Bigdata’s token-based pricing for content: what a token is, why grounding tokens exist, and why content is priced in tiers rather than one flat rate. This post is about the payoff of understanding that: what real financial research workflows actually cost, and why.
The value isn’t simply in giving an AI agent more tokens. It comes from giving the agent the right knowledge. High-quality financial content supports serious research without forcing the model to piece together answers from noisy, inconsistently sourced web data. Better grounding makes research more precise, more efficient, and less prone to hallucination.
We ran a set of research, monitoring, and enablement workflows through Bigdata MCP, from a post-earnings flash to a full morning portfolio briefing, and broke down not just the total cost but which content it actually paid for, line by line. If you’re trying to understand what high-quality, AI-driven financial research costs in practice, this is the real, unrounded answer and a starting point for budgeting your own.
The headline number
# | Use case | Workflow | Output type | Cost |
|---|---|---|---|---|
1 | NVIDIA initiation report | NVIDIA initiation report | ~2,000-word institutional report | $1.79 |
2 | RH post-earnings flash | RH post-earnings flash | ~400-word earnings note | $0.68 |
3 | Tesla long thesis memo | Tesla long thesis memo | ~1,500-word IC memo | $1.93 |
4 | Delta short thesis pressure test | Delta short thesis pressure test | ~1,200-word short thesis memo | $1.85 |
5 | Weekly market commentary | Weekly market commentary | ~800-word macro note | $0.83 |
6 | Salesforce competitive report | Salesforce competitive report | Multi-company comparison table | $2.69 |
7 | Starbucks pre-call sales brief | Starbucks pre-call sales brief | 5-minute sales enablement brief | $1.15 |
8 | NVIDIA company deep dive | NVIDIA company deep dive | ~2,000-word company research report | $0.99 |
9 | Portfolio sentiment scan | Portfolio sentiment scan | Multi-name portfolio sentiment report | $1.56 |
10 | Morning portfolio briefing | Morning portfolio briefing | Daily portfolio briefing | $1.66 |
Total | $15.13 |
Across all ten workflows, total usage came to $15.13, with individual runs ranging from $0.68 to $2.69.
Why a 400-word note can cost as much as a 2,000-word report
The clearest illustration is use case 2 versus use case 8. The RH post-earnings flash is a ~400-word note, and it cost $0.68. The NVIDIA company deep dive is roughly five times the length at ~2,000 words, and it cost $0.99, barely more.
The difference isn’t length, it’s what each prompt demanded as evidence:
The RH flash required side-by-side transcript quotes and consensus framing after every claim: a short output, but one that had to pull earnings context, compare transcripts, and support exact quotes. Its content mix: Premium News ($0.42), Web Content ($0.11), Earnings Transcripts ($0.07), Regulatory Filings ($0.04), Corporate Communications ($0.03), Corporate Fundamentals ($0.01).
The NVIDIA deep dive needed a broader but shallower evidence base: Premium News ($0.51), Fund Holdings ($0.18), Corporate Fundamentals ($0.08), Web Access ($0.07), Earnings Transcripts ($0.05), ESG Scores ($0.04), Jobs ($0.04), Regulatory Filings ($0.01), Corporate Communications ($0.01).
Same idea shows up again at the top of the range. Use case 4, the Delta short-thesis pressure test, is a ~1,200-word memo that cost $1.85, nearly as much as the 2,000-word NVIDIA initiation report ($1.79), because pressure-testing a thesis from multiple angles (skeptical coverage, filings, smart-money exit signals, operational deterioration, sentiment regime, and a bull rebuttal) pulls a wider set of evidence than simply writing up a company.
And the single most expensive workflow in the set, the Salesforce competitive report at $2.69, wasn’t the longest output at all: it was a comparison table. It cost the most because it expanded horizontally, repeating the same evidence requirements (financials, valuation, margins, hiring trends, strategic moves, citations for every figure) across six companies instead of one.
The workflows built to run every day
Part of what these ten runs make clear is that “financial research” on MCP isn’t just long-form writeups. Three of the cheaper, more repeatable use cases show a different shape of workflow entirely:
Weekly market commentary ($0.83). A recurring macro note covering equities, rates, currencies, commodities, and credit, with citations required for every numeric claim and every central bank reference. Its cost was spread across Premium News ($0.43), Corporate Fundamentals ($0.19), Economic Calendar ($0.11), Podcasts ($0.07), Web Access ($0.02), and Corporate Communications ($0.02), and it still came in under $1.
Portfolio sentiment scan ($1.56). Instead of writing anything long-form, this workflow scanned 15 names (NVDA, MSFT, AAPL, GOOGL, META, AMZN, TSLA, JPM, GS, XOM, COST, UNH, LLY, NFLX, CRM) for sentiment z-score breaks, media-attention momentum, recent 8-K filings, or earnings within the next 7 days, then surfaced the top 3 warranting review. Almost the entire cost here, $1.52 of $1.56, was Company Sentiment data. That’s the workflow’s job: it’s a triage tool, and its cost profile reflects that single-mindedly.
Morning portfolio briefing ($1.66). The same 15 names, but reframed as a daily “what changed since yesterday” briefing covering market backdrop, price moves, analyst changes, sentiment shifts, filings, earnings, headlines, and podcast mentions. Premium News dominated the cost here ($1.10 of $1.66), which tracks, since a same-day catch-up brief is, structurally, a request for the newest content available. Podcasts also showed up ($0.15) specifically because the prompt asked for relevant executive or employee appearances.
Where the money actually goes, in aggregate
Zooming out across all ten workflows, here’s the total spend by content type:
Content type | Aggregate cost | Approx. share |
|---|---|---|
Premium News | $5.43 | ~36% |
Fund Holdings | $2.01 | ~13% |
Company Sentiment | $1.61 | ~11% |
Corporate Fundamentals | $1.61 | ~11% |
Jobs | $1.34 | ~9% |
ESG Scores | $1.15 | ~8% |
Web / Web Access | $0.65 | ~4% |
Earnings Transcripts | $0.50 | ~3% |
Regulatory Filings | $0.26 | ~2% |
Podcasts | $0.25 | ~2% |
Corporate Communications | $0.19 | ~1% |
Economic Calendar | $0.11 | ~1% |
Knowledge Graph / Corporate Calendar | ~$0.04 | ~0% |
Total | ~$15.13 | 100% |
This is the practical payoff of the tiered pricing model we walked through last time. Premium News alone accounts for over a third of total spend across every single workflow, not because it uses the most tokens, but because current, licensed news is priced for what it is: freshness and exclusivity, not volume. Meanwhile Corporate Fundamentals shows up in nearly every use case in large token volumes (as high as 69,515 tokens in the Salesforce report) but consistently lands as a smaller share of cost, because it’s foundational, high-coverage content. The token count tells you how much the agent read. The tier tells you what it actually paid for.
What this means if you’re trying to budget
A few practical takeaways fall out of running these ten side by side:
Ask “what evidence does this need,” not “how long is this.” A short, citation-heavy note (RH flash, $0.68) and a long report with a similar evidence footprint (NVIDIA deep dive, $0.99) land close together. A short note that pressure-tests a thesis from six angles (Delta, $1.85) can cost more than either.
Horizontal breadth is the real cost multiplier. The jump from single-company research (roughly $1 to $2) to a six-company comparison ($2.69) or a 15-name portfolio scan ($1.56) is about the number of entities covered, not the prose length.
Recurring, narrow workflows are cheap and predictable. The sentiment scan and the market commentary are the two use cases here built to run on a schedule, and they’re also two of the least expensive, because their evidence requirements are narrow and repeatable by design.
The content mix is the bill. Any given run’s cost is really a blend of however many providers it drew on, each settled at that source’s own rate, which is exactly what “settlement happens per answer” meant in the last post, now with real numbers behind it.
The takeaway
Across ten genuinely different research, monitoring, and enablement workflows, every single one landed between $0.68 and $2.69. The value is in the retrieval, sourcing, and citation work the agent did to get there, run by run, source by source.
If you want to see what a specific workflow of your own would cost, the token calculator and your account’s usage dashboard will show you the same kind of breakdown used here, by content type, before you’re surprised by a bill.
Check current per-source rates at bigdata.com/pricing.
