Build the company that deserves the number.
A valuation is an output. The inputs are revenue, growth and retention, and the multiple is what the market decides to pay for them. This page is the plan for the inputs — including the parts that are uncomfortable to write down.
How to read the numbers here
- Benchmarks are other companies’ figures, cited. AlphaSense’s ARR is a fact about AlphaSense, not evidence about this one.
- Targets are targets. Every composition below models a business that does not exist yet, and is labelled that way rather than left to look like a result.
- Current prices live on the pricing page. Nothing here is an offer.
The problem is documented, and the incumbents just confirmed it.
Two incumbents moving in the same direction in the same year is not a warning that the category is taken. It is the market saying out loud that manual, email-driven corporate actions are no longer acceptable — while both moves stop short of carrying a single manager’s event from notice to settlement.
of surveyed hedge funds processing the relevant corporate-action workflows manually, with a further 11% only partly automated
identifying the lack of a consolidated view across asset classes as the primary challenge
estimated annual cost of inefficiency in corporate-action processing, with only 4 in 10 voluntary actions reaching straight-through processing
S&P launches Corporate Actions Core, positioned as a lighter SaaS deployment for hedge funds and asset managers that do not want a large enterprise implementation
DTCC AnnounceDirect first phase, limited to mandatory redemption events, replacing fragmented email and manual announcement submission with standardised workflow, status tracking and auditability
Seven stages, and the first three are the hard ones.
Each stage has one objective. Capital and hiring can accelerate everything after product-market fit; nothing accelerates finding it.
- Stage 0$0
Prove the workflow
One event, carried end to end, for one manager, with the operations work actually removed rather than re-presented. Nothing is generalised until that has happened once.
- Stage 1$1M
Own voluntary corporate actions
Embarrassingly narrow on purpose. The sentence is “we automate voluntary corporate actions for investment managers” — not “AI platform for the global financial ecosystem”.
- Tender offers
- Exchange offers
- Merger elections
- Rights offerings
- Consent solicitations
- Bond tender and exchange events
- Stage 2$5M
Own special situations
The same event engine, widened to the situations an event-driven desk underwrites: mergers, activism, going-private transactions, spin-offs, distressed exchanges.
- Stage 3$10M
The event graph, and agents on top of it
Stop being a website people visit and become something that watches the market for them. The agent has memory, state, triggers, tools, permissions and an audit trail.
- Stage 4$25M
The data business
Expose the normalised graph: REST, webhooks, streaming, SDKs, MCP, warehouse shares. Humans use the product; machines depend on the data. API revenue is exceptionally sticky.
- Stage 5$50M
Large institutions
Custodians, prime brokers, large asset managers, wealth platforms, fund administrators, transfer agents, broker-dealers, banks. Contracts stop being $75K and start being seven figures.
- Stage 6$100M
The securities event network
Sell both sides of the same workflow — the manager receiving the action and the issuer, agent or custodian originating it. At that point it stops being SaaS and becomes a network, and networks are extraordinarily hard to replace.
One workflow, owned completely.
The product is the completion of the workflow. The AI is an implementation detail, and saying so out loud is what keeps the roadmap from drifting into demos.
The lifecycle, end to end
- 01Notice
- 02Event extraction
- 03Notice reconciliation
- 04Portfolio impact
- 05Economics
- 06Decision
- 07Approval
- 08Instruction
- 09Acknowledgement
- 10Settlement
- 11Reconciliation
The first five steps are intelligence and exist today. The last six are operations, and they are the half that produces data no public source contains.
What V1 accepts
- EmailForwarded from the shared mailbox, exactly as it arrives
- PDFThe custodian notice, however it was scanned
- EDGAR filingThe primary source, already parsed
- CSV positionsAccounts, quantities, custodians. No integration required
Not Bloomberg, not DTCC, not a custodian integration. Six months spent on connectors is six months without a customer, and every one of those integrations is easier to sell after the workflow has already proven itself on a forwarded email.
Who it is sold to
- Event-driven hedge funds
- Merger-arbitrage funds
- Credit and distressed funds
- Multi-strategy hedge funds
- Lean asset managers
- Family offices with institutional operations
$500M to $20B AUM. Not the largest banks — they buy slowly, and they buy from incumbents.
The first million
| Customers | Average ACV | ARR |
|---|---|---|
| 5 | $40K | $200K |
| 10 | $60K | $600K |
| 15 | $75K | $1.125M |
Thirty institutional customers is past $2M. That is a more tractable problem than acquiring a hundred thousand consumers, and it is the reason the free surfaces are distribution rather than the business.
The multiple is not the lever.
What a billion-dollar outcome requires depends entirely on the quality of the revenue underneath it. The uncomfortable row is the last one.
| Quality of company | Multiple | ARR for ~$1B |
|---|---|---|
| Ordinary software | 4–5× | $200M–$250M |
| Excellent vertical SaaS or data | 8–12× | $83M–$125M |
| Exceptional | 20× | ~$50M |
| AI hype on tiny revenueNot a strategy | — | Possible temporarily |
So the target is not a valuation. It is $100M ARR growing above 50%, with enterprise net revenue retention above 120% and strong gross margins. Design for the company that gets there at 10× and the 20× outcome is upside rather than a requirement.
What the market is actually paying
Median public software company
As of 31 July 2026. Growth remains the biggest driver of the spread.
~4.1× implied ARR
Top ten public software companies
The same date. Investors still pay these multiples for exceptional growth.
~20.3× implied ARR
AlphaSense
June 2026 round, roughly 12.5× ARR — the closest comparable for financial intelligence at scale.
$7.5B on $600M+ ARR
Allvue
Reported potential sale, with >15% growth and >30% margins — financial infrastructure, not hype.
Up to $3B on $200M+ ARR
Broadridge recurring revenue
$305M of closed sales in the year. The scale of recurring workflow spend inside financial institutions.
$4.878B in FY2026
Quartr
More than 14,500 companies covered, $18M raised, triple-digit growth — data distributed by API, webhook and Snowflake.
~120% NRR
Four revenue lines, and only one of them creates the valuation.
The consumer and prosumer tiers are not the business. They are distribution, brand, and a feedback loop on extraction quality. The enterprise line is what creates the valuation — which is why the free surfaces have to lead somewhere rather than exist for their own sake. What is on the pricing page today is the current, real price list; this is the shape the business is being built toward.
| Segment | Customers | Average | Revenue |
|---|---|---|---|
| Asset managers and hedge funds | 250 | $180K | $45M |
| Banks, custodians and brokers | 30 | $750K | $22.5M |
| Data and API customers | 100 | $200K | $20M |
| Professional and team subscriptions | ~12,000 | ~$85/mo blended | ~$12.2M |
| Total | ~$100M |
A model, not a forecast, and not a price list.
Revenue mix at scale
- Enterprise workflow45–55%
- Financial data and API20–25%
- Institutional network and platform10–20%
- Self-serve professional10–15%
Diversification of this shape is what makes the company hard to kill, and interesting to more than one kind of acquirer. But acquisition is not the plan — the plan is to make acquisition expensive.
What has to be true at $30M–$50M ARR
These matter far more than the feature count. A premium multiple is paid for growth and retention, and the last three lines are the ones specific to this product.
70%+
Year-on-year growth
120%+
Enterprise net revenue retention
95%+
Gross retention
75%+
Gross margin
Eventually
>$100K
Enterprise ACV
<18 months
CAC payback
<10%
Revenue from the largest customer
>80%
Workflow automation rate
~100%
Source-grounded fields
Near zero
Material operational errors
Share of an event lifecycle completed without manual operations work
A better North Star than queries answered or documents processed, because it is the only number that goes up when the product does the job rather than describes it.
Eight things, in order.
Everything else waits.
- 01
ActionDesk V1
Ingest a corporate-action notice and a positions file, and produce an actionable event workspace: citations, deadlines, affected holdings, the decision, and the instruction workflow.
- 02
The event graph, in parallel
Do not hard-code the product around PDFs. Every notice extracted creates or updates a canonical event — otherwise there is a demo and no company.
- 03
Four public surfaces, not forty
M&A and tender tracker, activist and 13D radar, insider tracker, company event timeline. Enough to prove the distribution thesis and no more.
- 04
Five hundred named prospects
Fund, AUM, strategy, COO, head of operations, middle-office staff, OMS/PMS where discoverable, prime brokers where discoverable.
- 05
Twenty to thirty conversations a day
Founder sales, not marketing automation. Every weekday.
- 06
Five design partners before going broad
Breadth before depth is how a product ends up shallow in twelve directions.
- 07
Charge early
Free proof of concept, then a $10K pilot, then a $50K–$100K annual agreement.
- 08
Track one number
The share of an event lifecycle completed autonomously. Everything else is a proxy.
The opening line
“We built an AI operations agent that manages voluntary corporate actions from the incoming notice through election and confirmation. I am looking for funds still handling part of this in email and Excel. I will process ten of your historical events free and show you exactly what it would have automated.”
Trying to look like a billion-dollar company before becoming a one-million-dollar one.
AlphaSense already has hundreds of millions of documents, financials, screeners, M&A intelligence and private-company data, on more than $600M of ARR. S&P has enormous licensed datasets and has wired them straight into ChatGPT through an MCP server. You do not beat either of them by being a smaller version of them. You beat an incumbent by owning a workflow they do not own end to end, and then expanding from it.
Not being built
- Screeners
- Earnings calls
- News
- Stock pages
- Options
- Crypto
- Portfolio analytics
- General chatbots
- Voice
- Social feeds
- Mobile apps
- Fifty EDGAR tools