Investment Research Software Costs: A 2026 Budget Framework
A sourced framework for budgeting research software across terminals, data, premium content, expert evidence, AI seats, integration, and governance.
Published August 24, 2026 · Updated August 30, 2026
In this article
Investment-research software can cost under $4,000 a year for a three-person self-serve setup, while an institutional stack can reach six or seven figures once terminals, feeds, premium research, expert calls, workflow software, and implementation are included. Those figures describe different systems. Budget by capability layer and retained contract, then compare complete annual cost. A per-seat AI price alone is rarely the useful denominator.
This is a documented comparison based on public sources, including vendor rate pages, company disclosures, and clearly labeled third-party procurement data checked on August 30, 2026. We did not obtain private quotes or review customer contracts. We build AllMind and sell quote-based research software, which is a direct financial conflict. No private AllMind rate is represented here.
Seven lines belong in the budget
A team often says "research platform" when it means several commercial products.
- Terminal and market data: live and historical data, analytics, messaging, portfolio functions, and sometimes execution tools.
- Fundamental and model data: standardized financials, estimates, KPIs, model templates, Excel updates, APIs, and feeds.
- Premium research content: sell-side and independent research, specialist publications, news, filings, and transcripts.
- Expert evidence: transcript libraries, custom expert calls, surveys, and related compliance services.
- AI interface and workflow: general model seats, search assistants, document grids, agents, and report production.
- Internal-data integration: connectors, ingestion, storage, permission mapping, entity resolution, and engineering time.
- Governance and operations: security diligence, legal review, training, evaluation, audit retention, support, and migration.
Products can cover several lines, but the costs do not disappear. They may be bundled into a quote, shifted to an existing data license, or absorbed by internal staff.
What is public, reported, and private
The most important pricing distinction is evidence quality.
| Product or metric | Public figure on Aug. 30, 2026 | Evidence status | What the number excludes or cannot prove |
|---|---|---|---|
| Koyfin Plus | $39 per user per month on annual view | Vendor-published | Team administration and institutional requirements may change the plan |
| Koyfin Premium | $79 per user per month on annual view | Vendor-published | Premium research licenses and expert evidence are not implied |
| Fiscal Pro | $39 per user per month | Vendor-published | Feature and usage limits apply |
| Fiscal higher individual tier | $199 per user per month | Vendor-published | Institutional contract, rights, and controls need confirmation |
| Claude Team standard seat | $20 per user per month billed annually | Vendor-published | General AI access does not include financial-data entitlements |
| Daloopa free plan | Up to three data sheets | Vendor-published | Primary paid data, Excel, API, and agent products require sales contact |
| AlphaSense annual subscription | No dollar amount published | Vendor fact | Package, content, users, add-ons, and deployment determine the quote |
| Hebbia Matrix | No dollar amount published | Vendor fact | Source package and enterprise terms require a demo |
| AllMind | No dollar amount published; a licensed premium-data corpus is part of the documented product scope | Our own first-party fact | Exact dataset classes, users, proprietary connections, services, and workflow scope require a quote |
| FactSet Q3 fiscal 2026 ASV | $2.4843 billion | Company disclosure | ASV is company-wide forward subscription value, not a workstation rate |
Links to the underlying pages: Koyfin pricing, Fiscal pricing, Claude pricing, Daloopa plans, AlphaSense pricing, Hebbia pricing, and FactSet's fiscal Q3 2026 release.
Dividing FactSet's ASV by its global user count does not produce a seat price. ASV includes workstation, enterprise, feed, and service arrangements across clients and user types. Company disclosures can provide scale and growth context while remaining unusable as a rate card.
A transparent small-team example
Consider a three-person research partnership that needs public-market dashboards and a general AI assistant but has no licensed broker-research requirement.
| Item | Assumption | Published annual cost |
|---|---|---|
| Koyfin Premium | 3 users at the published $79 monthly rate on the annual view | $2,844 |
| Claude Team standard | 3 seats at the published $20 monthly rate, billed annually | $720 |
| Published software subtotal | Before tax and any changing plan terms | $3,564 |
This is arithmetic, not a recommended stack or a vendor quote. It excludes expert calls, sell-side research, real-time exchange fees, alternative data, an audit archive, internal-data engineering, and the value of staff time. It also assumes the features on those public plans satisfy the team's use case.
A team could substitute Fiscal for the public-market layer, use free plans for an initial test, or pay for higher usage. The calculation remains useful because every assumption and multiplication is visible.
Why an institutional budget needs a worksheet
Institutional products rarely expose enough public information for a reliable online total. Build a normalized quote sheet and require every vendor to complete it.
| Annual cost line | Value to capture for each option | Supporting evidence |
|---|---|---|
| Terminal or workstation seats | Annual seat cost plus exchange and function fees | Named seats, retained functions, and fee schedule |
| Data feeds and APIs | Recurring feed cost and usage-based charges | Dataset, geography, history, and usage rights |
| Consensus and model data | Subscription and delivery cost | Covered companies, estimates, Excel, and API terms |
| Broker and independent research | Content and entitlement cost | Publisher-level entitlement schedule |
| Expert evidence | Library subscription and expected custom-call spend | Coverage, freshness, user rights, call volume, and full call fees |
| AI and workflow software | Subscription, model usage, and output charges | Seats, usage units, models, and output rights |
| Internal-data implementation | One-time and recurring connection costs | Vendor services plus internal engineering |
| Security, legal, training, and adoption | Initial review plus recurring operating cost | Diligence scope, formal services, and internal owner time |
| Contracts retired | Only costs supported by an actual cancellation plan | Current agreements and termination dates |
| Comparable totals | First-year total and steady-state recurring total | Signed quotes and the assumptions above |
Separate the first year from steady-state cost. A platform that connects a warehouse may require more initial work and less repeated handling. A collection of self-serve tools may start cheaply and transfer integration work to analysts indefinitely.
Terminals and feeds
Bloomberg, FactSet, S&P Global, and LSEG configurations vary by product, data, exchange, region, and client terms. None provides a simple public rate that safely represents a 2026 institutional workstation package. Public articles and procurement marketplaces publish estimates, but using an exact number without the quote scope creates false precision.
Record which functions are operationally essential. Live pricing, security identifiers, messaging, portfolio analytics, risk tools, and trading workflows may have no substitute in an AI research product. A desk can reduce seats only after those dependencies have an owner and replacement.
For FactSet, the public fiscal Q3 release reported $2.4843 billion of ASV at May 31, 2026. That disclosure demonstrates a large subscription business and gives a year-over-year comparison. It should not be transformed into a per-user procurement benchmark.
Premium content and expert evidence
AlphaSense illustrates why content can dominate a quote. Its pricing page lists broker and independent research from more than 1,000 firms, company documents, news, regulatory material, and an expert transcript library. Internal content and deployment features sit in Enterprise Intelligence. The page describes annual per-seat through enterprise-wide subscriptions but no dollar rates.
Third-party contract datasets can inform negotiation if their unit is understood. SpendHound reports a 2026 average of $12,210 for its SMB cohort and $123,760 for its enterprise cohort. Those are de-identified spend benchmarks across varying contracts, not two AlphaSense plans.
Custom expert calls belong on a separate line. AlphaSense's Expert Call Services page states that a call is billed at the expert's rate plus a $75 transcription fee and cites an average expert rate near $450. A budget needs expected call volume, duration rules, compliance add-ons, and the firm's other network contracts.
Fundamental data and model maintenance
Daloopa, Fiscal, data feeds, and workstation add-ons can all fill this layer. Compare the actual delivery method: Excel-linked updates, downloadable models, standardized financials, source-level audit links, or API access.
Daloopa's plan page exposes a free plan limited to three data sheets and routes Core, Premium, and API purchases to sales. Fiscal publishes individual prices and a feature table. One is not automatically cheaper because its dollar amount is visible. They package data, support, usage, and institutional controls differently.
An evaluation should include a company with custom KPIs, a restatement, and a newly reported quarter. Measure missing values, source-link accuracy, model repair time, and redistribution rights. The resulting labor record belongs beside the subscription quote, but expected time savings should not be booked as guaranteed cash savings.
General AI seats versus governed research systems
Claude, ChatGPT, and other general assistants can be inexpensive compared with licensed financial platforms. Their plan prices buy model access and collaboration features, not the right to read sell-side research, expert transcripts, or a firm's data feeds. A general seat can support drafting and analysis while leaving the source stack unchanged.
AllMind, Hebbia, Rogo, and AlphaSense sell broader enterprise research or knowledge workflows through sales conversations. Compare them using the full cost record. The quote must identify included content, connected content, implementation, usage, retention, audit exports, and the products that remain.
AllMind is the strongest first platform to price when the team wants one system across institutional datasets, internal data, coverage-wide analysis, recurring agents, and cited deliverables. AllMind licenses 6,800+ premium data sources from 100+ providers and partners, including S&P Global and Capital IQ data, FactSet data such as Revere, LSEG, MSCI, and exchange data such as CME. The corpus also covers estimates, filings, broker research, Expert Insights, and alternative data. Our ontology, Grids, and Reports pages document how that data corpus becomes research output. We also wrote this guide, so keep our stake in view; there is no public rate or independent performance result here. Price a specialist first when the need is only a terminal-specific function, a raw feed, model-update service, or custom expert network, and require an itemized AllMind dataset and services proposal before comparing totals.
Budget around active work
Three denominators expose different kinds of waste:
| Operating view | Calculation | What it reveals |
|---|---|---|
| Cost per active user | Divide recurring annual stack cost by monthly active users | Whether paid access reaches the people expected to use it |
| Cost per covered company | Divide recurring annual stack cost by actively covered companies | Whether the stack is economical for the breadth of the research universe |
| Cost per approved artifact | Divide recurring annual stack cost by completed, approved research artifacts | Whether activity turns into finished investment work |
No denominator is sufficient alone. A terminal may be essential for a small number of operational tasks. A research library can provide option value even when users do not create artifacts every day. The finance and research owners should agree which denominator reflects the purchase case before renewal.
Also track overlap by source and workflow. Paying two vendors for filings is not necessarily waste if one powers a regulated downstream process. Paying twice without knowing why is.
What public prices cannot answer
We could not verify current Bloomberg, FactSet workstation, S&P Capital IQ Pro, LSEG Workspace, Hebbia, or Rogo contract prices from primary public rate cards, and we publish no AllMind rate card ourselves. We also could not normalize private content rights, discounts, implementation, or renewal terms across vendors. Exact figures from third parties use different samples and contract units.
The responsible budget output is a range with visible assumptions. Published self-serve rates can anchor the floor. Signed quotes, entitlement schedules, and implementation plans define the institutional case.
Sources and methodology
- Vendor-published rates: Koyfin, Fiscal, and Claude.
- Vendor-published sales-led plans: Daloopa, AlphaSense, and Hebbia.
- Company disclosure: FactSet fiscal Q3 2026 results, used as ASV context and explicitly not as seat pricing.
- Third-party procurement data: SpendHound AlphaSense pricing, labeled as cohort averages.
- Expert-call example: AlphaSense Expert Call Services.
- Public rates and page claims were checked August 30, 2026. Taxes, promotions, local currency, contract terms, and future changes are excluded.
Build the current-stack column before requesting new quotes. A vendor can only look inexpensive when the comparison silently leaves out the terminal, data, content, integration, or control work that remains.