Choose it as an operational decision about a repeatable service, not as a feature comparison. The tests that matter across a client base are ledger coverage, what adding the thirtieth company looks like, how much of each cycle goes on building rather than reviewing, whether someone who did not build the forecast can maintain and explain it, permissions that keep client data separate, and the cost of the standardised state. An in-house finance team running several entities faces the same evaluation with "client" swapped for "entity".
Forecasting for clients is a different job from forecasting one business
A finance team forecasting its own business builds one model, learns its quirks and lives in it. A practice builds a comparable forecast twenty or fifty times, hands it to people who did not build it, and maintains every copy on a fee. That changes what good looks like. A tool that rewards deep configuration suits the first job and punishes the second, because configuration time repeats per company. The 2024 CPA.com and AICPA PCPS CAS Benchmark Survey of 206 US firms found only 55% charge a separate fee for the technology and connection setup each new client needs, so for roughly half of firms that repeated work is absorbed rather than billed.
The same shape appears inside companies. Where scaling actually comes from is worth naming precisely: it is the number of separate forecasting units, data sources and exceptions, not headcount or revenue. HP's Asia-Pacific treasury reported spending three to four hours per legal entity on cash positioning and forecasting before automation, in an HSBC customer case study. A finance team of four running six entities across two currencies is managing a portfolio on the same logic as a practice managing clients.
How to evaluate a tool across a client base
Coverage, and the exceptions you will keep. The tool must connect to the ledgers the portfolio actually runs, and the honest question is not whether coverage is complete but which companies will sit outside the standard and what happens to them. Count the client list by ledger before shortlisting anything, then write the exception policy rather than pretending there will be none.
What company thirty looks like. Multiply the setup time you measure on a real client file, not demo data, by the size of the list. Forty minutes of mapping per company is a day and a half across fifty. Ask the same question about every recurring task the tool creates.
The review-to-build ratio. In the automated forecasting implementations documented by the Association for Financial Professionals, the pattern that works is consistent: systems generate and update the underlying numbers, and finance time goes on reviewing assumptions and exceptions. So ask how much of each forecast cycle a tool spends building versus reviewing. Software that makes model construction impressive but still demands heavy monthly maintenance has worse economics across fifty companies than software producing a simpler forecast that mostly maintains itself.
Handover. Not only whether a client can read the output, but whether a colleague who did not build the forecast can maintain and explain it. A sophisticated model that depends on one person's knowledge cannot be productised as a recurring service, and this is one of the sharpest differences between choosing for yourself and choosing for a practice.
Permissions. Practice staff need edit access across companies; a client needs sight of their own company and nothing else. Check what the roles actually are, not whether a permissions page exists.
The cost of the standardised state. Licence cost per company is the smallest part. The real figure is operational cost per supported company: onboarding, staff training, integration maintenance, client training, exception handling, review time, support and eventual export. No published dataset puts money against those components, so they are criteria to price yourself rather than numbers to look up.
Forecasting as a billable service
The deliverable is not a file. It is a standing answer to the question owner-managed clients ask eventually: are we going to be fine for the next quarter? ICAEW's IT Faculty guide to the virtual FD describes three delivery models UK practices use, from basic compliance with ad hoc advice, through scheduled monthly board-level meetings, to a hybrid of scheduled advice and technology. The economics support the shift: in the CAS benchmark, hourly billing has fallen from the primary model at 53% of advisory practices in 2018 to 10%, and participating practices reported 17% median growth.
Two cautions belong here, and neither is a tooling problem. First, in the UK, preparing cash flows and budgets is a regulated accountancy service under ICAEW's statement on members engaging in public practice, effective January 2025, which requires a practising certificate where the work is charged for. Check your own position before packaging the service. Second, ICAEW's guide observes that practitioners comfortable billing hourly often find higher-value strategic advice a confusing place and risk giving away their value. Automating the data collection does not fix that. Pricing and positioning are a separate skill from tooling, and the practices that struggle usually struggle there.
What remains knowable in advance is the cost side: tool cost per company, plus review time, per month. If those two numbers leave no margin at the fee a client will pay, the service does not work whatever the software.
Which question do you need answered?
Cash flow forecasting, three-way modelling and management reporting answer different questions, though the products increasingly overlap and several tools do more than one. A cash flow forecast projects money in and out over a near horizon, commonly thirteen weeks, to answer whether there is enough cash to meet obligations. A three-way model links profit and loss, balance sheet and cash flow so a change in one flows through the others. Management reporting explains what already happened, with variance analysis and commentary.
Decide by the question, not the product label. Practitioner guidance names four triggers for genuinely needing three-way modelling rather than a cash forecast: bank covenants or a facility application, planning beyond twelve months with growing stock and debtors, board or investor requirements, and a capital expenditure programme. Outside those, building a three-way model to answer a near-term cash question is over-engineering. The failure mode is buying one category and expecting another's answer, and vendors blur the line, which is exactly why the question comes first. We compare the categories in our guide to comparing cash flow forecasting tools.
Worth knowing: the thirteen-week horizon is not a software marketing invention. The Association of Corporate Treasurers documented it as the normal territory of operational cash forecasting in The Treasurer as far back as 2000.
What to check before standardising
Standardising is where the payoff and the risk both sit, and the evidence on it is thinner than the enthusiasm. The CAS benchmark names a standardised stack as an efficiency lever, but we found no independent study measuring realised gains or switching costs, and the widely repeated standardisation statistics all trace to a single Intuit survey of 700 US accountants rather than to several sources.
That survey also carries the cost. In it, 61% of respondents said firms committed to standardisation should sever ties with clients unwilling to adopt their preferred platforms. Treat that as the price on the ticket rather than a footnote. The profession has not settled the question either: practice technology commentators continue to argue that best-of-breed selection beats suite standardisation, citing switching costs and vendor lock-in.
The workable position is a standard default with a written exception policy. Which clients cannot use the tool, and which should not use any tool at all. A cash-stable business with one entity, modest transaction volume and no borrowing may be well served by a spreadsheet or by the forecasting already in its ledger. ICAEW's own guidance says a forecast can live in a spreadsheet, an app or an accounting system, and that what matters more than the system is the information going into it. We have written honestly about when Xero's built-in forecasting is enough and when QuickBooks Online's native options are reasonable. Then check data freshness precisely, security in terms an IT manager will accept, and whether forecasts export cleanly if the practice changes course.
How Float fits
Float is a cash flow forecasting layer that connects to Xero and QuickBooks Online, with a direct Sage Intacct connection in development on a public waitlist. It maintains a rolling 13-week forecast for operational decisions and a monthly forecast extending up to three years, refreshing from the ledger roughly every 24 hours with a manual sync available. Float does not connect directly to banks. It reads the reconciled bank data already in the ledger, so it works with whatever banks the ledger supports and inherits the ledger's reconciliation discipline.
For portfolio work, whether the portfolio is clients or entities, multiple companies connect to one dashboard, and consolidation shows a combined position across any selected set with currency conversion where entities differ. Presentation mode strips the interface for walking a client or a board through the numbers live, and summaries export to CSV and PDF. Permissions run on three roles, Admin, Editor and Viewer, with two-factor authentication mandatory for Xero-connected users. Plans and per-entity pricing are on the pricing page; Float asks practices and fractional finance leaders to talk about specific pricing rather than publishing a practice rate.
Frequently asked questions
Which cash flow software do accountants actually recommend?
Recommendations cluster around whichever tools connect to the ledger the accountant already runs, so the useful question for a business is what works with its own Xero or QuickBooks file rather than what accountants recommend in general. Treat published "best of" roundups carefully, since many are written by vendors that appear in them. Ask who publishes a list before trusting its ranking.
Does automated cash flow forecasting actually save time for accountants?
We could not find an independent study measuring time saved specifically by automated cash flow forecasting for accounting practices or smaller finance teams, and the 50% to 90% figures in circulation are vendor-authored. Published corporate case studies do report real savings, including six days of manual effort a month at SIG plc, but these are vendor and customer accounts of large treasury operations rather than independent research. The nearest peer-reviewed evidence is a Stanford and MIT study in the Journal of Accounting Research covering 277 accountants and field data from 79 firms, which found accountants using generative AI reallocated about 8.5% of their working time from data entry towards client work and closed the month 7.5 days faster; its field firms were all one software vendor's clients. The mechanism is better established than any headline number: research by the Association for Financial Professionals found finance teams spend about half of forecast time gathering and preparing data rather than analysing it.
What should a small accounting practice look for in a forecasting tool?
Coverage of the ledgers its clients actually run, setup measured in minutes per client rather than hours, output a colleague or client can maintain and read without the builder present, and pricing that still works across the full client list. A small practice has less slack to absorb a tool that fits only some clients, so coverage and the exception policy matter most.
Can Float consolidate multiple clients or entities into one view?
Yes. Multiple companies connect to one Float dashboard, and the consolidation view combines the cash position of any selected set, with currency conversion where companies use different currencies, alongside each company's own breakdown. The architecture is the same whether the companies are a practice's clients or a group's entities.
How does Float handle permissions and security for client data?
Float uses three roles: Admin, with full edit access and user management, Editor, with full edit access, and Viewer, read-only. Two-factor authentication is mandatory for Xero-connected users and strongly recommended for everyone. Float has no direct bank connection, reading bank data through the authorised ledger integration instead, so client bank credentials never reach it.
Do all clients need a dedicated forecasting tool?
No. A single-entity business with stable cash, modest transaction volume and no debt covenants is often well served by a spreadsheet or its ledger's built-in forecasting, and ICAEW guidance takes the same view that the system matters less than the inputs. A dedicated tool earns its cost where cash is genuinely uncertain, entities or clients multiply, or someone wants a standing forward view rather than a periodic snapshot.
Is cash flow forecasting a regulated service in the UK?
Preparing cash flows and budgets is listed as an accountancy service in ICAEW's statement on members engaging in public practice, effective January 2025, which means an ICAEW member charging for that work generally needs a practising certificate. Requirements differ by professional body, so check your own body's rules before packaging forecasting as a paid service.
How much does Float cost for an accounting practice?
Float publishes plans with per-entity pricing on its pricing page and asks practices, bookkeepers and fractional finance leaders advising clients to get in touch about specific pricing. The published figures answer the in-house multi-entity case directly, and the practice case begins as a conversation from a known starting point.
If the next step is testing this against a real file, start a free 14-day trial, connect a client's or an entity's Xero or QuickBooks Online ledger, and Float will build the first rolling forecast from the live data. No credit card required.
