The Best ROI Number Is the One Your Client Argued With

"I don't think it would spend 13 hours."

That is what the managing attorney said when I put my first labor estimate in front of him. I was walking a personal-injury law firm through the business case for a new tech stack, and the whole argument rested on one number: 13 hours per demand packet. A demand packet is the bundle of medical records, liability narrative, and settlement math a firm assembles to open a claim with an insurance carrier. I had counted the time it takes to build one. Thirteen hours.

Multiply those 13 hours by the firm's case volume and their loaded labor rate, and the current state comes to roughly $87,000 a year in human hours. Bring in an AI drafting tool, cut each packet to about six hours, and that figure drops to roughly $40,000. The gap is the pitch: $47,000 a year in saved labor.

He looked at the 13 and pushed back. "I don't think it would spend 13 hours."

That pushback was the moment the number stopped being mine and started being his.

The number nobody argues with is the number nobody trusts

Here is how most ROI cases die. Someone opens a spreadsheet, or a vendor's calculator, and produces a number. The number arrives fully formed. The person asking for the money presents it. The person holding the money receives it the way they receive a sales brochure: politely, with a discount already applied.

They do this for good reason. Every automation vendor ships a calculator that asks for your headcount and your average salary and spits out six figures of "savings." Every consultant has an incentive to make the current state look expensive and the future state look cheap. The person approving the spend has been burned by enough of these numbers that their default posture is skepticism. They do not audit your math. They quietly assume it is inflated and say no, or say yes to a smaller number than you asked for, or table it.

The operator then does the worst possible thing. They add more numbers. More tabs in the spreadsheet. More caveats. They defend the estimate instead of opening it up. The harder they defend, the less the number gets believed. Defending a number reads as attachment to an outcome, and the moment the other party smells attachment, they assume the number was built to justify the outcome rather than the reverse.

The pushback test

There is a better way to tell whether a number is real. Call it the pushback test.

A number is defensible when the person who would write the check has tried to break it and failed.

That is the whole test. It is about whether the number survives contact with someone who has a reason to want it to be wrong. A number nobody has pushed on is just an assertion. A number that survived being pushed on is evidence.

The best place to run this test is in a room, in real time, with the assumptions out on the table where everyone can reach them.

How to run the argument

Put the number in front of the person who does the work

The first mistake is presenting the ROI case to the wrong person. You pitch the CFO or the owner, because they sign. But the person who can break your assumptions is the person who does the work. In the law firm meeting, the person who said "I don't think it would spend 13 hours" was the attorney who assembles the packets. He had a lived sense of how long one takes, and it did not match my estimate.

That is the person you want arguing with you. The signer asks questions about totals. The doer asks questions about inputs, and inputs are where numbers live or die.

Concede the moment they push back

When he said 13 was too high, I did not defend it. I said, "We could lower the numbers. We'll do 10." Then I re-ran the calculation at the lower number.

This feels backwards. You built the number. You want to protect it. But conceding is what makes the number theirs. When you lower an estimate at their request, they start treating the result as their own conservative figure rather than your optimistic one. A number they helped build down is a number they will defend to someone else later.

The case got stronger when I dropped 13 to 10, because the number that remained was one the attorney had already vouched for.

Separate the dollars from the hours

Labor savings matter, but they undersell the argument. In that meeting the clean dollar figure was a little over $100 saved per demand packet. Useful, but not the thing that moved the room.

The thing that moved the room was the hours. Thirteen hours to six. Across a year of cases, that is the difference between roughly 800 hours and roughly 135 hours of work. Eight hundred hours is a person. That is where the real conversation happened: "It's the time you really can't compare. One hour versus six hours. Think how much more you could do."

Dollars are what you justify the spend with. Hours are what you justify the change with. Lead with the hours and let the dollars follow.

Use the loaded rate, not the salary

The other number that quietly destroys ROI cases is the labor rate. People price an employee at salary divided by 2,000 hours, and they are off by a wide margin. The Bureau of Labor Statistics puts benefit costs at close to 30 percent of total private-sector compensation. A $75,000 salary carries another $30,000 or so in benefits before the person logs a single productive hour, and the effective rate climbs once you account for the hours that disappear to meetings, admin, and rework.

If your rate is too low, your savings are too low, and you have already lost the argument before anyone pushes back.

Let the client deliver the punchline

If you run the argument well, you do not need to close it. The client will close it for you. Near the end of that meeting, after the numbers had been pushed on and rebuilt, the attorney said it himself: "It's literally a no-brainer."

That sentence is worth more than any slide. When the person who will write the check states the conclusion in their own words, the case is closed. Your job is to set the table and then get out of the way.

The limits of the argument

A negotiated number can still be wrong. It is only as good as the inputs, and the meeting itself said so: "It's only as good as the person putting in the information. If they're leaving out critical pieces, it's not going to do a fantastic job. You still have to have the monkey on the keyboard."

The pushback test also leaves out the things labor math misses. Quality. Compliance. The risk you remove when fewer hands touch a document. A process improvement that prevents one careless error is worth something real, and it will not show up in a per-demand labor calculation. The numbers argument works for the things numbers capture and leaves the rest on the table.

And a number that survives one person's pushback has not survived a real audit. If you only argue with the stakeholder who is already on your side, you have rehearsed, not tested. The test is only as strong as the skeptic you run it against.

The number you built together is the one that holds

The 13 hours never came back. The final case ran on a number the attorney had already agreed was fair, because he was the one who talked it down to it. That is the difference between a number you present and a number they own.

The next time you build a business case, stop trying to make the number bulletproof. Hand it to the person who does the work and invite them to break it. The moment they say "I don't think it would spend 13 hours," you are negotiating. A negotiation both sides have had a hand in is one both sides will defend.

---

Puzzle is the visual platform where teams document and continuously improve their operations. Its process costing tools attach time, frequency, and rate to every step, so the current state and future state sit side by side and the ROI conversation happens against real, documented numbers rather than a vendor's estimate. See how it works at puzzleapp.io.