Nielsen Says Audit the Outcome Meter With 50 Billed Cases. The Dispute Queue Already Pulled the Sample.

Jakob Nielsen wants designers to pull 50 cases billed as resolved. A 48-hour silence rule produced 200-plus contested cases a week. Read those first.

Falk Gottlob · · 7 min read

Execution Falkster cover on teal: a large cream meter with one navy needle, and a long row of cream paper slips feeding out of its side into a navy tray, with a single orange slip at the front.

Jakob Nielsen's second edition of Service as a Software gives designers a new object to work on: the meter that decides what gets billed. This is a Rewrite of his first instruction, because the sample he asks for is already sitting in a queue most designers never open.

The short version

  • On November 18, 2026, Jakob Nielsen republished Service as a Software with an argument that designing and auditing the outcome meter is now a core UX job.
  • Per-resolution vendors can bill customer silence as success: 24 hours at Fin, 72 at HubSpot, and 2 to 72 at Zendesk, as he reports them.
  • He tells UX professionals to pull 50 cases billed as resolved and count how many customers would agree. In my field report on killing a per-seat tier, a resolved ticket meant no escalation within 48 hours.
  • The rule held for 98% of cases and still produced more than 200 contested cases a week, with disputes at 600 a month against 50 projected. Moving the window to 7 days cut disputes 60% and cost revenue.
  • So start with the dispute queue, design the dispute path and its capacity, and spend the 50-case sample on the silent cases nobody contested.

What Nielsen is asking for

The essay is long and I read all of it. The piece that concerns designers is the definition of "resolved."

He lists three published meters and how each treats silence. Fin bills an assumed resolution when the customer goes quiet for 24 hours after the agent's last answer. HubSpot counts a conversation resolved if its agent shared a help article or took an action and the customer didn't ask for a human within 72 hours. Zendesk waits out an inactivity window of 2 to 72 hours by channel, then has an AI model read the transcript. He gives credit where he thinks it is due: Zendesk added that check in May 2026, and Fin drops the charge if the customer returns to the same conversation.

His objection is that silence may mean the customer gave up. He names the practice outcomewashing and draws the meter as a detector with two kinds of error. The one the warranty covers is work that failed and wasn't billed. The one nobody counts is work that was billed and didn't resolve anything, because the vendor got paid and the buyer's dashboard shows a resolution.

Then the instructions. Guideline 7: contesting a charge for a phantom resolution should take one click. Guideline 9: never let silence pass as satisfaction, itemize the invoice into confirmed and assumed outcomes, and have the assumed ones sampled by someone who isn't paid per resolution. And for UX professionals: pull 50 cases billed as resolved and count how many of those customers would agree.

The sample that already exists

I'd keep all three. I'd change the order.

In Field Report: What Broke When We Killed Our Per-Seat Tier, which is an anonymized composite of sunsets I have led or advised on, the unit was a resolved ticket, and resolved meant the customer didn't escalate within 48 hours. That is a silence meter. It was written by pricing and product, and no designer was in the room, which is the situation Nielsen describes.

It worked for 98% of cases.

The 2% were customers who reopened after 60 to 72 hours with a follow-up that was technically a new issue and, to them, the same one. The system had already counted and billed the first. At low volume those were handled one at a time. At sunset volume they were more than 200 contested cases a week. Total disputes hit 600 a month against a projection of 50, and the analyst team went from 4 to 9 to 14 while resolution quality sagged and NPS on the new product dipped 8 points.

200+ a week

contested cases from a 48-hour silence rule that was right 98% of the time, in my field report on killing a per-seat tier.

Look at that as research. Every one of those cases is a billed-as-resolved outcome that a customer looked at and rejected, with the reason attached. It is Nielsen's audit, run by the people with the most standing to run it, at a sample size no study would get funded for.

So the first move for a designer on an outcome-priced product is to read the dispute queue. Tag each case by the rule that failed: the window was too short, the follow-up was the same issue, or the answer was wrong and the customer went quiet anyway.

The window is a price

The fix in the report was to rewrite the definition at month 20: no follow-up in 7 days. Disputes fell 60%. Some revenue was lost.

Both halves of that sentence matter. A longer window bills fewer silences and takes money off the invoice. A shorter one bills more and sends the difference to the dispute team. Fin's 24 hours, HubSpot's 72, and the 48 in my report are three positions on one dial.

Nielsen predicts, at 60% confidence, that before 2028 one of the three big support-agent vendors will stop billing silence at full price. I'd guess the dispute queue gets them there before a designer's audit does, because it has a cost line.

Design the dispute first

Guideline 7 asks for one-click contesting. I agree, and I'd size it before shipping it. Disputes ran 12x projection in the report without a one-click path. A designer who makes contesting easy and doesn't tell anyone what it does to volume has moved the failure from the invoice to the queue.

Three things to hand over with the button:

  1. The reason on the form. Same issue, wrong answer, or never resolved. One tap. It is the tagging above, done by the customer.
  2. The capacity number. Disputes per thousand billed outcomes last month, and who answers them.
  3. What an upheld dispute changes. A credit, and a row in the list of cases the definition gets tested against next time.

The service design for the moment after a wrong charge is close to what the Design handbook covers in Recovery and Trust Repair.

Where the 50 go

The dispute queue has one blind side. A dispute comes from the buyer who reads the invoice. The end customer who gave up at hour three and went to a competitor files nothing, and the buyer's dashboard says resolved.

That is the quiet cell in Nielsen's table, and it is where his sample earns its keep. Pull the 50 from outcomes closed by silence where nobody came back and nobody contested. Call or message those customers. In the report's terms, the band between 48 hours and 7 days is where I'd look first, since the tightened rule shows that is where the definition was wrong.

The order, then: the queue, the dispute path, and the sample. This sits in SaaS to AI Business Models because the meter is the price, and it carries the Design tag because the meter is now something a designer can be handed. For the board-level version of the definition problem, see monday.com Counted 75M Agent Requests. The Board Needs One Outcome Number..

Related answer: How do you audit a per-resolution pricing meter?

Sources: Jakob Nielsen, "Service as a Software: The New SaaS Is Ahead of Schedule," Jakob Nielsen on UX, November 18, 2026 (second edition of a December 2024 article), read in full from the feed: the three vendors' silence windows, outcomewashing, the detector table, guidelines 7 and 9, the 50-case audit, and the prediction. I did not open the vendors' own pricing pages; their terms are as he reports them. The 48 hours, 98%, 200-plus a week, 600 against 50, 4 to 9 to 14, 8 points, 7 days, and 60% are from my field report linked above.

Share this post

Frequently asked

What does Jakob Nielsen mean by auditing the meter?+

In the November 18, 2026 second edition of Service as a Software, Nielsen argues that under outcome pricing the billing unit is a measurement artifact and therefore UX territory. His instruction to UX professionals is to pull 50 cases billed as resolved and count how many of those customers would agree, and to design the intake, checkpoints, and redress of an outcome-priced service.

What is an assumed resolution?+

A resolution billed because the customer went quiet. As Nielsen reports it, Fin bills one when the customer is silent for 24 hours after the agent's last answer, HubSpot counts a conversation resolved if the customer doesn't ask for a human within 72 hours, and Zendesk waits 2 to 72 hours depending on channel and then has an AI model verify the transcript. He notes Fin drops the charge if the customer returns to the same conversation.

Why start with the dispute queue instead of a 50-case sample?+

Because it already exists and it is larger. In my field report on killing a per-seat tier, a 48-hour silence rule was right 98% of the time and still produced more than 200 contested cases a week at scale, and disputes ran 600 a month against a projection of 50. Each one arrives labeled with the charge the customer rejected.

What does the dispute queue miss?+

Everything nobody fought. A dispute comes from the buyer who reads the invoice. An end customer who gave up and left does not file one. That is where Nielsen's sample belongs: cases closed by silence where the customer never came back.

What happened when the silence window was tightened?+

In the field report the rule moved from no escalation in 48 hours to no follow-up in 7 days. Disputes fell 60% and some revenue was lost. The window is a pricing decision with a number on each side.

Where do the field report numbers come from?+

From Field Report: What Broke When We Killed Our Per-Seat Tier, which I describe there as an anonymized composite of pricing-tier sunsets I have led or advised on, with numbers that are illustrative but representative of the actual ranges.

DesignJakob NielsenUX Tigersoutcome pricingper-resolution pricingassumed resolutiondisputesFin
Helpful?

Keep reading

The short answer

Pricing Migration: The 18-Month Quarterly PlaybookSix quarters, four mandatory CFO/CRO/Board/Lead-customer conversations, the gross-margin recovery curve, and the contract terms that defend outcome billing.Read the chapter

Part of

SaaS to AI Business Models

About the author

Falk Gottlob

Falk Gottlob

Product Executive · Founder, Falkster.AI

Thirty years shipping product, from Microsoft Research and Adobe to Salesforce, where he grew Quip into what became Slack Canvas. Four startup exits, including a $6.5B healthcare merger and a company Microsoft bought. Four-time Chief Product Officer. Now founder of Falkster.AI, building Heidi, the company brain for AI-native businesses. This notebook is written from inside the build, not above it.

Sign in with LinkedIn to leave a comment.

Sign in with LinkedIn