Almost every CPA firm outsourcing cost savings case study ends in one percentage. That number is real for somebody, and it still cannot go into your budget, because a percentage without a denominator is not evidence.
In the profession's own benchmarking survey, most responding firms report outsourcing only a thin slice of their work. So a headline saving is usually a saving on the slice, not on the firm.
Two moves make the number usable: a disclosure test you can run on anyone's published result, and a measurement design that produces a defensible case inside your own practice.
What a CPA Firm Outsourcing Cost Savings Case Study Is Really Claiming
A saving percentage is a ratio, and a ratio means nothing until you know its base. The same dollar saved reads as a large percentage against one seat, a moderate one against a service line, and a rounding error against the firm's total cost. A case study that never names its base has not told you which of those three it measured.
Survey evidence says the base is usually small. In the 2025 National MAP Survey, 59% of respondents said only 1% to 5% of their work is outsourced or completed by a global team, and 25% put the figure between 6% and 15%.
The firm-level ceiling sits in the same survey's size-band exhibits. Salary expenses excluding owners ran 34.4% of net client fees for firms with $1.5 million to $5 million in annual net client fees, and 41.1% for firms above $10 million (National MAP Survey size-band exhibits).
Put those two facts side by side and the ceiling is arithmetic. Outsourcing can only move the part of the salary line it actually touches, so a firm that sends out a twentieth of its work will not cut firm cost by anything near the percentage a vendor page prints. A big saving on the work that moved and a small change in what the firm spends are both true at the same time, which is why the headline alone decides nothing.
The Six Things a Case Study Has to Disclose
A case study that discloses these six is worth reading closely. One that discloses none of them is an advertisement with a number in it.
The baseline, and when it was taken. A before number captured after the decision was made is a reconstruction, and reconstructions drift toward the answer the author wants. Ask what the firm's cost per file was in the season before anything moved, how it was calculated, and where the data came from. If the answer is "we estimated it later", the saving is an estimate too.
The denominator. Ask what the percentage is a percentage of: one seat's loaded cost, one service line, or total firm cost. That one question resolves most of the distance between a modest claim and a spectacular one, because the two are often the same dollars measured against different bases.
The review that stayed behind. Nobody outsources the partner's signature, so somebody in the firm still reviews the work. A saving figure that never subtracts those retained hours is a gross number wearing a net number's clothes, and that gap tends to be largest while the arrangement is new and review is heaviest.
The measurement window. Two months of a slow summer is not a season. A credible case names the window, and the window is long enough to include a filing deadline and to be compared with the same calendar months of the prior year.
The one-time costs. Ramp weeks, extra software seats, process documentation and the consent administration all land in year one. A case study that reports year-two economics and calls it the result of the change has quietly removed the price of admission.
Who measured it, and who paid for it. A case study on a provider's site is that provider's own advertising, whatever the layout suggests. That does not make it false. It does mean the disclosure standard for advertising is the right lens to read it through.
What the FTC's Endorsement Guides Say About One Firm's Result
The Federal Trade Commission's Endorsement Guides describe how the Commission reads a testimonial. Under section 255.1(a), endorsements must reflect the honest opinions, findings, beliefs or experience of the endorser, and an endorsement "may not convey any express or implied representation that would be deceptive if made directly by the advertiser" (eCFR, Endorsement Guides, general considerations).
The substantiation rule is the sharper one. An advertisement relating one customer's experience on a central attribute of the product "will likely be interpreted as representing that the endorser's experience is representative of what consumers will generally achieve", and where the advertiser lacks substantiation for that, the advertisement "should clearly and conspicuously disclose the generally expected performance in the depicted circumstances" (eCFR, consumer endorsements).
The Guides close the obvious escape hatch in their worked examples. A disclosure such as "Results not typical" is called insufficient to prevent an advertisement from being deceptive, and the Commission reports testing that wording alongside a stronger version and finding that "neither disclosure adequately reduced the communication that the experiences depicted are generally representative" (eCFR, consumer endorsements).
The scope needs a caveat. That section addresses consumer endorsements, and a CPA firm buying staffing is a business buyer, so treat the standard as the reasoning a careful reader applies rather than as a rule that automatically governs the page in front of you. The question it hands you is still the right one: is this result what a firm like mine should generally expect, and what backs that up?
Build the Case for Your Own Firm
Set the Baseline Before Anything Moves
Pick one service line, not the whole practice. Individual returns, business returns and monthly bookkeeping have different review depths, and a blended firm-wide number hides the line where the saving is real.
Then record four things for a full comparable period before any work moves: the count of files completed, the preparer hours per file, the review hours per file, and the fully loaded cost of the people doing both. Cost per reviewed file is the unit that survives a change in volume. Total cost is not.
Capture it from your time system rather than from memory. If your firm does not track review time separately from preparation time, start there, because the review line is where an outsourcing decision either pays off or quietly does not.
Price the Seat With Public Inputs, Then Swap In Your Own
National data gives you a sanity check while you assemble your own numbers. In the May 2025 Occupational Employment and Wage Statistics survey, accountants and auditors had a median hourly wage of $40.23 (BLS, Occupational Employment and Wage Statistics).
Pay is not the whole seat. Benefit costs averaged 30.1% of total employer compensation for private industry workers in March 2026, with wages and salaries making up the other 69.9%, so read pay as a share of employer cost and gross it up rather than adding a benefit percentage on top of it (BLS, Employer Costs for Employee Compensation).
Unemployment insurance is the input people forget because it is small federally and variable locally. The federal unemployment tax rate is 6.0% on the first $7,000 of each employee's wages, and an employer that pays its state unemployment taxes in full, by the due date of its Form 940, and on all the same wages subject to FUTA tax is entitled to the maximum credit of 5.4%, which leaves a federal rate of 0.6%. In a credit reduction state that credit is reduced, resulting in a greater amount of federal unemployment tax due (IRS, Topic no. 759).
Your state rate is the one that moves. State unemployment tax is set by your state against a state wage base and generally varies with your own claims history, so it is a number you look up for your own account rather than borrow from a national table. Turnover belongs here too: recruiting fees, the vacancy months and the ramp for a replacement are real costs of the in-house side, and they show up in your ledger, not in a survey.
The Consent Work Comes First
You cannot run a measured pilot on live client files until the disclosure paperwork is done. If the preparer receiving the file is located outside the United States, the taxpayer's consent under the Treasury regulations at section 7216 is required before any disclosure, and that consent must be knowing and voluntary (eCFR, Treasury regulations under section 7216).
Treat that as a scheduling constraint on the experiment. Consent language belongs in the engagement letter cycle, months ahead of the season you plan to measure, or the pilot slips a year.
Five Things That Move Your Number Besides Outsourcing
A naive before-and-after credits the decision with everything that changed. These five change on their own.
Pay drift. Private industry compensation costs were 3.3% higher over the 12 months ending in June 2026, with wages and salaries 3.1% higher and benefit costs 3.8% higher (BLS, Employment Cost Index). Part of a year-over-year saving can be pay drift you would have absorbed anyway, so restate the baseline at current pay bands before you compare.
Volume drift. National filing volume barely moved between the last two seasons: the IRS had received 140,633,000 individual income tax returns by April 18, 2025 and 140,222,000 by April 17, 2026, a decline of 0.3% (IRS, filing season statistics). A single firm's count usually moves more than the national aggregate does, which is why the comparison has to be per file rather than per season.
Mix and complexity. Sending out the simplest returns first makes the average outsourced file look cheap and leaves your in-house average looking expensive. Compare like files, or segment the measurement by complexity band.
Ramp landing in year one. Training weeks, documentation and early rework are concentrated at the start. Report the first year and the steady state separately instead of averaging them into a number that describes neither.
The seat you could not fill anyway. This one changes the comparison rather than the arithmetic. Employment of accountants and auditors is projected to grow 5% from 2024 to 2034, with about 124,200 openings projected each year on average over the decade, many of them expected to result from the need to replace workers who transfer to different occupations or exit the labor force (BLS, Occupational Outlook Handbook). Those are projected openings across the occupation, not a queue of candidates waiting for your posting. If your last search ran past a season, the honest comparison is capacity against no capacity, and the cost on the other side is the work you turned away.
An Illustrative Example: What a Headline Percentage Is Worth
The point of the arithmetic below is the shape, not the answer. Three of its inputs are assumptions rather than facts: the length of a working year, the headline saving rate, and the share of the seat's hours that actually move. Every other figure is a public one cited above, and none of it describes a real engagement.
``` Illustrative only. Public inputs, no client data.
One seat, fully loaded median hourly wage, accountants and auditors $40.23 BLS OEWS, May 2025 pay is 69.9% of employer compensation cost, / 0.699 BLS ECEC, March 2026 all private industry = employer compensation cost per hour $57.55 x 2,080 hours a year $119,704 + federal unemployment tax after full credit $42 IRS Topic no. 759 ------- = about $119,746, before state unemployment tax, software seats, workspace and recruiting
Now apply a headline saving of 40 percent 40% of the loaded seat about $47,900 but only on hours that actually move; if half the seat's hours move about $23,900 less retained review, ramp and extra licenses your numbers
The same dollars, seen from the firm's P&L median net client fees, firms at $1.5M to $5M $2,620,966 MAP 2025 salary expense excluding owners, 34.4% of NCF about $901,612 $23,900 as a share of that salary line about 2.7% $23,900 as a share of net client fees about 0.9% ```
The same dollars can be described honestly as a large percentage of one seat or as a fraction of a percent of the firm, and the only thing that changed is the base. That is the whole problem with a headline number, and the reason a case study without its denominator tells you almost nothing about your own firm.
Four Questions That Get You Past the Marketing
Ask these of any provider whose results you are being shown, and ask them in writing.
"What was the baseline, and who measured it?" You are looking for a number captured before the change, from a system rather than a memory. A provider who has never asked a client for a baseline cannot produce a saving, only a rate.
"Does the figure net out the review hours the firm kept?" A provider who has thought about your review layer answers this immediately, because retained review is the number their delivery model has to survive. Gross and net are different claims, and only one of them is your cost.
"Which of your clients looks most like my firm, and what did their first year look like?" First-year economics carry the ramp. A provider who only quotes steady-state results is describing year two to a firm about to live year one.
"Will you run a fixed block of my own work, on my software and my process, and let my reviewer grade it?" This is the question that replaces belief with measurement. Graded output on your own files is evidence you own, and it is the only case study certain to be about your firm.
Frequently Asked Questions
What Should a CPA Firm Outsourcing Cost Savings Case Study Include?
A named baseline and how it was measured, the denominator behind the percentage, the treatment of retained review hours, the length and calendar position of the measurement window, the one-time ramp costs, and who performed the measurement. Results reported per file rather than per season are more portable, because they survive a change in volume.
Is a Published Saving Percentage Ever Useful?
As a prompt, yes. As a forecast, no. Use it to work out which base would have to be true for the number to hold, then ask whether your own base looks anything like it.
How Long Before a Firm Can Measure a Real Result?
Long enough to cover a ramp and one comparable season, measured against the same calendar months of the prior year. Anything shorter measures the transition rather than the arrangement.
Can I Rely on a Provider's Reference Calls?
They help when you ask the reference the same six disclosure points rather than asking whether they are happy. Ask what their baseline was, what stayed in-house, and what the first year cost them. A reference who cannot answer those has not measured either.
The Case Worth Building
The only cost savings case that can settle this decision is the one built from your own baseline, your own review depth and your own file counts, measured across a window you chose before you knew the answer. Everything a seller publishes is at best a reason to run that measurement.
Accountably has made 30+ placements across 20+ US firms since 2022. The results from those engagements are self-reported, held under confidentiality and never independently audited, so they belong in the same bucket as every other vendor's published number, not in your budget.
Produce your own instead. Accountably runs a Free 40-Hour Proof Pilot on a fixed block of your own work, prepared on your software and your process and put through full review, so your reviewer grades real output before a client file depends on it. If a placement is not the right fit in the first 30 days, we replace them free.