01
Diff your worldview
Compare actuals with your own pre-earnings estimates and evaluate thesis rules you defined in advance. Albatross does not label synthetic or user-entered estimates as market consensus.
Private beta · Built for fundamental equity teams
Your earnings workflow, completed automatically.
Albatross detects earnings filings, extracts source-linked actuals and company KPIs, compares them with your pre-earnings estimates, checks your thesis rules, and can write validated results into a copy of your Excel model.
Designed for small fundamental teams where the model, KPI sheet and research thesis already exist before earnings day.
Primary SEC sources
Explicit financial QA
Evidence on every fact
Copy-only Excel writeback
What is different
Not another earnings summary.
The product starts from the analyst's own expectations. The output is a source-linked diff between the quarter that reported and the worldview that existed before the print.
02
Evidence before interpretation
Every surfaced numerical fact retains its metric, value, unit, period, basis, source document, source URL, locator and evidence snippet. Ambiguous required facts fail closed.
03
Protect the model
Validated actuals can be written into a new copy of the analyst's workbook. The source model is never overwritten, formulas are protected by default, and every write is documented.
The workflow
Configure once. Run every earnings season.
Albatross is designed around the existing model, KPI and thesis process rather than forcing the team into a generic research template.
01PrepareSave analyst estimates, thesis rules and optional Excel mapping before the company reports.
02DetectWatch selected issuers for SEC 8-K Item 2.02 earnings events or run an accession manually.
03ExtractPreserve primary sources and extract reported financials plus confirmed company-specific KPIs.
04ValidateRun implemented accounting, period, unit and plausibility checks before downstream model actions.
05UpdateShow estimate variance and thesis results, then optionally create a validated model copy.
Earnings day
Move the analyst from collection to judgment faster.
Without Albatross
Hunt through the release, exhibits and filing for the quarter's actuals.
Reconcile units, periods, GAAP/non-GAAP labels and recurring company KPIs.
Manually compare the print against the team's model and thesis notes.
Copy values into Excel and retrace sources later if something looks wrong.
With Albatross
Primary documents are preserved and source-linked facts are extracted immediately.
Implemented financial checks run before the report is treated as ready.
Your estimate misses and thesis checks are surfaced in the same workflow.
Validated actuals can flow into a safe copy of the model with evidence attached.
Trust architecture
The LLM is not the financial database.
Albatross V1 does not require a production LLM provider. Source-of-truth financial numbers come from deterministic source extraction or deterministic calculations over source-derived facts.
What Albatross will do
Preserve source documents, retain provenance, normalize periods and units, run explicit implemented checks, fail closed on required ambiguity, and keep analyst inputs distinct from reported facts.
What Albatross will not do
Invent a missing financial value, silently overwrite a model, describe a user-entered estimate as consensus, or imply that passing implemented checks certifies every fact in the filing.
“If software is going to touch an analyst's model, ‘the AI said so’ is not a sufficient evidence trail.”
V1 acceptance coverage
Tested against different issuer shapes.
The private beta was accepted using hardened company workflows and a generic human-reviewed onboarding path.
Financial statements, membership and supplemental KPIs, margin context, estimate/thesis workflow and Excel writeback.
Data Center KPIs, GAAP/non-GAAP metrics, forward guidance, analyst inputs, source evidence and report persistence.
Generic schema path with no Microsoft-specific extractor, including segment data, Cloud revenue and commercial RPO.
Founding pilot
Run it on companies your team actually covers.
C$1,500 setup + C$499 / month
- Up to 25 companies and 3 users.
- One investment team per private deployment.
- Custom company KPI schemas and research workflow setup.
- Analyst estimates, thesis rules and optional Excel model mapping.
- Founder-led onboarding and direct support during the pilot.
Questions
What a pilot team usually wants to know.
Does Albatross use market consensus estimates?
V1 is built around the analyst team's own pre-earnings estimates. It does not require a paid consensus feed and does not represent synthetic or user-entered estimates as consensus.
Will it overwrite our Excel model?
No. The source workbook is retained as an immutable input. When writeback is enabled, Albatross creates a filing-specific copy, blocks the write if financial validation fails, protects formulas unless explicitly allowed, and adds an evidence sheet.
How do you add a new company?
Albatross proposes a reusable research schema from deterministic candidates. A human reviews and confirms the metrics once; the schema is then stored by issuer CIK and reused on later quarters. V1 does not silently auto-confirm schemas.
Is this a public SaaS product?
Not yet. V1 is a founder-led private beta with one investment team per deployment. Public signup, billing, multi-tenancy, screening and automated trading are intentionally outside the current product boundary.
Founding teams
Want to test it on a real earnings workflow?
Founding pilots are arranged directly with the founder. Email [email protected] and tell us a little about your team and the companies you cover.