How RealQuant Credits Work
Why we meter at all
A 12-page teaser and a 200-page OM with a rent roll and T-12 are not the same compute job. When we mapped the cost variance across document types during beta, the spread was around 15x. A flat subscription collapses that into one monthly rate, which means someone absorbs the difference on every large document you run.
In practice that someone is you. When a vendor is losing margin on complex deals, they recover it somewhere. A smaller model. Truncated context on long documents. A monthly document cap that only surfaces when you hit it. You rarely see it happening. The extraction results just look a little worse on the 80-page packages.
For a $50 million acquisition, that degradation has real downstream cost. A missed rent concession. A pro forma assumption that never got challenged. A rent roll silently cut off at page eight because the context window ran out.
We bill per unit of compute instead. Each credit maps to actual work: pages parsed, tokens processed, output generated. A large deal costs more credits than a small one. It also gets the same model and the same context window. The permanent reference at realquant.ai/credits has the full operation taxonomy and formula breakdown.
What consumes credits
Four things consume credits. Navigating the interface, reviewing extracted data, and managing your deal pipeline do not.
1. Document parsing
When you upload an OM, rent roll, or T-12, we send it to Azure Document Intelligence to extract the text and layout structure from the raw PDF. The billing unit is pages. A 50-page OM costs more than a 12-page teaser because there is more document to process. This is the first step; nothing else runs until it is done.
2. AI extraction
After parsing, a language model reads the extracted text and pulls deal fields: unit mix, in-place rents, opex line items, pro forma assumptions, debt terms. Input tokens and output tokens are billed separately because they hit different cost curves. Both show up as distinct line items in your usage history.
3. Comps and market data
When the add-in fetches comparable sales or rent comps, we run a search-augmented query against market data sources. The billing unit is per query, scaled by output length. A comps pull on a 280-unit multifamily deal in Atlanta with a two-mile radius costs more than a quick cap rate lookup.
4. LOI and report generation
LOI drafts, investment memos, and executive summaries are billed on output tokens. A standard LOI is shorter than a full IC memo. Both are shorter than anything you would write from scratch.
Three deal scenarios
The easiest way to understand credits is to run the numbers on actual deals.
Quick Screen — 12-page OM, no rent roll
~20 creditsScenario: A broker sends a teaser on a 48-unit garden-style deal in Phoenix. You want a quick pass before deciding whether to request more materials.
- ›Parse 12 pages = 12 credits (1 credit/page, exact)
- ›AI orchestration + field extraction = ~7 credits
- ›No comps pull, no LOI
On a Starter plan (500 credits), you could run 20+ quick screens per month. Parse cost is exact at 1 credit/page; AI token costs vary by document density.
Full Underwriting — 52-page OM + rent roll
~95 creditsScenario: You are under LOI on a 280-unit value-add deal in Atlanta. Full pipeline — OM receipt through LOI execution.
- ›Parse 52-page OM + 8-page rent roll = 60 credits
- ›AI orchestration + field extraction = ~18 credits
- ›Comps pull (3 queries + market data) = ~7 credits
- ›LOI draft generated = ~7 credits
Full pipeline from OM receipt to LOI. Parse is the largest single cost at 1 credit/page. On a Growth plan (1,500 credits), you can run 12–15 of these per month.
Portfolio Screen — 5 OMs, same market
~95 creditsScenario: A broker sends a five-deal package — all garden-style in the same submarket. You need quick screens on all five before your investment committee call Friday.
- ›5× parse (12 pages each) = 60 credits
- ›5× AI extraction = ~28 credits
- ›1 shared comps pull (same submarket, pulled once) = ~7 credits
Credits scale linearly. No batch penalty. Five short screens cost roughly the same as one full pipeline run because the documents are smaller. The shared comps pull saves versus five separate queries.
The formula
The math, if you want it:
# Credits charged per operation
Credits = (Unit Cost × Quantity) ÷ $0.01
# Where:
Unit Cost = per-page rate (parsing) or per-token rate (LLM)
Quantity = pages processed or tokens consumed
$0.01 = credit denomination (one cent per credit)
One note on caching: when we can reuse context from a prior request in the same session, cached token rates apply. For a long document you query multiple times in a session — running different extraction passes, asking follow-up questions — the second and subsequent reads are meaningfully cheaper than the first.
What is included in each plan
| Plan | Credits / seat / mo | Est. full pipeline runs | Price | Overage rate |
|---|---|---|---|---|
| Starter | 500 | ~3–5 | $199/seat/mo | $0.05/credit |
| Growth | 1,500 | ~8–12 | $349/seat/mo | $0.04/credit |
| Professional | 5,000 | ~30+ | $799/seat/mo | $0.03/credit |
| Enterprise | Custom | Custom | Contact us | By contract |
Credits are account-pooled, not per-seat. Your whole team draws from one balance. Analysts, associates, and PMs all work the same pipeline. That is how deal teams actually operate, so that is how we bill.
The rest of what you should know
- ›Credits reset each billing month. Unused credits from the prior month do not carry over. Your allotment replenishes on your renewal date.
- ›You can see your balance in real time. Your dashboard shows current balance, per-deal usage, and a full transaction history with operation-level detail.
- ›Extra usage is on by default. If you exceed your monthly allotment, overages are billed at month-end at your plan rate: Starter $0.05/credit, Growth $0.04/credit, Professional $0.03/credit. You can turn it off in your account settings — when credits run out, new AI jobs pause until the next cycle or you buy a credit pack. Either way, you get email alerts at 80% and 100%. We never cut off access mid-workflow on accounts with extra usage enabled.
- ›Enterprise credits are by contract. If you are running high deal volume or want dedicated compute, reach out to discuss your pipeline.
The reference page at realquant.ai/credits has the full operation taxonomy, formula breakdown, and tier table. Bookmark it if you want to estimate costs before uploading a deal package.
Zachary Shapiro
Co-Founder & CEO, RealQuant
Before building AI tools for CRE, Zachary spent over a decade as a principal at Blackstone-backed platforms and other institutional REPE firms, deploying $11B+ across 150+ transactions. He writes from the perspective of someone who has sat in the analyst seat, the IC seat, and now the founder seat.
Ready to automate your CRE underwriting?
RealQuant populates your Excel model from your deal documents in minutes, with every number cited to its source.