Calculating Caseload ROI: Which SSDI Cases Are Worth Taking

Every SSDI practice is a portfolio of contingent bets with a hard ceiling on the upside. The fee is capped, you only get paid if you win, and the wait between signing and getting paid is measured in months or years. That combination means the case you feel good about and the case that's actually worth taking are not always the same file — and the only way to tell them apart is to price each one the way an underwriter would.

This isn't about turning intake into a spreadsheet exercise that ignores the human in front of you. It's about being honest that a firm that signs every sympathetic claimant and a firm that signs the same claimants plus a clear-eyed view of expected value end up with very different cash flow. Here's how to build that view.

The unit economics of an SSDI case

Start with what you can actually earn. Under SSA's fee-agreement process, an approved representative's fee is the lesser of 25% of past-due benefits or the $9,200 cap (the cap that applies to agreements approved on or after 11/30/2024). That's the whole revenue line. There's no hourly upside, no premium for a hard-won case, and no fee at all on most prospective benefits — the fee comes out of the back-pay pool.

Three variables drive the economics of any single file:

• Expected fee. Bounded above by $9,200, but in practice it's 25% of accrued back pay — which is a function of monthly benefit amount times the number of months between onset (or application) and award. A case that pays out at $9,200 and a case that pays out at $3,000 cost you roughly the same effort.
• Time-to-pay. Back pay accrues while the claim sits, but so does your carrying cost and the time value of the fee. A case won at the initial level pays out far sooner than the same case won three years later at a hearing.
• Win probability. The single hardest term to estimate, and the one most firms guess at. It depends on the medical record, the stage of entry, and — at hearing — the office and judge.

Expected value = win probability × expected fee − cost-to-serve

The framework is deliberately simple:

EV = p(win) × E[fee] − cost-to-serve

Where E[fee] is bounded by the lesser of 25% of expected back pay and the $9,200 cap, p(win) is your probability of a favorable outcome at the stage you'd enter, and cost-to-serve is the fully loaded cost of developing the record, ordering evidence, and staffing the case through decision.

Two implications fall straight out of the algebra. First, because E[fee] is capped, you cannot price a weak case by charging more — the only lever on a marginal file is p(win) and your cost-to-serve. Second, EV is a per-case number, but you run a portfolio. A book of moderate-p(win), low-cost cases can outperform a handful of "perfect" cases that each soak up months of associate time.

You'll notice this framework never asks you to invent a precise win percentage. It asks you to rank and compare files on consistent terms. Even ordinal estimates — "this is a stronger record than that one" — improve portfolio selection, as long as they're applied the same way to every file.

How stage of entry changes the ROI

The same claimant has a different EV depending on where you pick up the case, because all three variables move at once.

• Initial. Fastest time-to-pay if it's allowed, lowest back-pay accrual, and the lowest grant rate of any stage. Low cost-to-serve per file, but you carry the denials too.
• Reconsideration. Slightly more back pay accrued, still a screening-level decision.
• Hearing. The longest wait — which means the largest accrued back pay (more cases hit the $9,200 cap) but the slowest payout and the highest cost-to-serve. It's also where the win probability becomes estimable from public data, because outcomes vary systematically by office and judge.
• Federal court. Highest cost-to-serve and longest horizon; a remand resets the clock rather than paying out.

The practical takeaway: a hearing-stage case is a different financial instrument than the same claimant at the initial stage. More of the fee is likely to reach the cap, but it's discounted by a multi-year wait and a heavier cost-to-serve. Model the stage you're actually entering, not the case in the abstract.

Office and judge grant-rate effects on portfolio returns

At the hearing stage, the p(win) term stops being a pure guess. SSA's ALJ Disposition Data publishes favorable, partially favorable, unfavorable, and dismissed counts per judge, and the OHO public workload files give you processing time and pending volume at the office level — your time-to-pay input.

Read these the way you'd read any base rate, with the usual caveats: dismissals aren't merits denials, small denominators are noisy, and a grant rate is the output of a docket's composition as much as a judge's tendency. (We unpack those traps in the ALJ-data piece and in reading hearing-office statistics.) Used as a base rate to anchor a per-case estimate — not as a forum-shopping tool or a number to quote a client — office and judge data sharpens the p(win) and time-to-pay terms in your EV model. Across a full caseload, even modest improvements in how you anchor those two terms compound into materially different portfolio returns.

Building an intake scorecard

The EV framework only changes behavior if it lives at intake, before you've sunk cost into a file. A workable scorecard scores each prospective case on the terms that drive EV:

1. Technical screens (pass/fail). Insured status, DLI, and earnings over SGA are gates — a file that fails these has no EV regardless of medical strength.
2. Medical-record strength. An ordinal estimate of p(win) from the longitudinal record and functional documentation.
3. Expected fee. A rough back-pay estimate (benefit amount × accrued months), capped at $9,200, to flag the low-payout files.
4. Stage and forum. The entry stage, and at hearing, the office/judge base rate and expected processing time.
5. Cost-to-serve. What it will actually take to develop this record to decision.

Combined, those produce a comparable EV ranking across every lead — which is exactly the calculation AISSDI's Lead Scorer is built to run, using the per-office and per-judge data as the p(win) and time-to-pay anchors.

None of this is a guarantee of any individual outcome, and it shouldn't replace the judgment that tells you a sympathetic, well-documented case deserves a shot even when the spreadsheet is lukewarm. But run consistently across a full intake pipeline, an EV discipline is the difference between a caseload that feels busy and one that actually pays — and it lets you say yes to the marginal-but-cheap case and no to the sympathetic-but-uneconomic one with the same clear eyes.