Equity research software for the whole cycle
Initiation, model maintenance, earnings, and the note at the end of it. The parts of coverage that are genuinely analytical are a fraction of the time an analyst spends; the rest is assembly. This attacks the assembly.
Free plan. No card required.
What does equity research software need to do?
Four things, in order of how much time they consume. Get the filed numbers into a model without retyping them. Keep the model current as the company reports. Surface what changed in the language as well as in the figures. And produce something publishable at the end without rebuilding it in another application. Most tools do the first well and leave the other three to the analyst.
The specifics
Ten years of filed fundamentals
Read from the company's own tagging, so a model starts from the statements rather than from a transcription of them.
More on thisValuation with a defensible discount rate
Cost of equity from a live risk-free rate and a measured, mean-reversion-adjusted beta, weighted against an after-tax cost of debt, with the basis written out so it can go in a footnote.
More on thisComparables with counts
Peer multiples with the number of peers behind each median stated, because a median computed from two companies is a midpoint and should not be presented as a market view.
Earnings and transcript analysis
History against expectations, and a comparison of what management said this quarter against last, aligned so a restated guidance figure is reported rather than lost as churn.
Filing comparison
This year's annual report against last year's, so a new risk factor or a quietly reworded revenue policy surfaces without a manual read.
Output that leaves the tool
Memos, committee packs and decks in portable document, presentation and spreadsheet formats, built from the same numbers rather than retyped.
How the work actually goes
- 01
Initiate
Pull ten years of statements, build the model on real inputs, set the peer set with the counts visible.
- 02
Maintain
When the company reports, the filed figures update. What you re-examine is the assumptions, not the data entry.
- 03
Interrogate the language
Compare the call and the filing against the prior period. The sentences that were rewritten are where guidance actually moves.
- 04
Publish
The note, the pack or the deck, exported from the same workspace the model lives in.
There are no sell-side estimates here
No consensus numbers, no analyst ratings, no price targets. Every free source of that data is licensed for personal, non-commercial use, because the vendors licence it themselves and pay to redistribute. Rather than display something we cannot licence, the feature does not exist. If your process depends on consensus revisions, you will need that from elsewhere, and it is better to know now than after a subscription.
Frequently asked questions
Can I build and maintain a model in Khan Terminal?
Yes, for valuation. The discounted cash flow, comparables, simulation and sensitivity work live in the workbench and load with filed fundamentals already populated. It is not a general-purpose spreadsheet replacement, and it exports to one where you need that.
Does it have consensus estimates?
No, and that is a licensing constraint rather than an oversight. Every free analyst data feed we tested is restricted to personal, non-commercial use, and the commercial tiers are priced well beyond this product. We would rather have an absent feature than an unlicensed one.
Can I compare earnings call transcripts?
Yes. The comparison is built specifically for transcripts rather than reusing filing diffing, because two calls are independent utterances rather than revisions of one document. Numbers are normalised for matching and preserved for reporting, so a guidance figure restated from mid single digit to low single digit is reported as a change rather than counted as boilerplate.
What formats can I export to?
Portable document format, PowerPoint and Excel. The deck is composed by a deterministic layout engine from the same content, so it is consistent across decks rather than reformatted each time.
Initiate on a name this afternoon
Ten years of statements, a model with real inputs, and a note at the end of it.