Season 3 · Episode 3

Two Per Cent Sell Outcomes. Here Is What Happened to One of Them.

HubSpot moved its support agent from a price per conversation to a price per resolution, published the rate card, and let customers convert committed seat spend into credits. The market took a fifth of the company on the announcement, and another fifth on the quarter that beat every headline number. Eight days after those results the chief executive was gone, and the shares rose six per cent.

The Agentic Commercial Model Newsletter · Fessal Rahman · November 16, 2026 · 11 min read

From FR Advisory.

The author spent twenty years running commercial organisations inside software and private equity portfolios. He no longer works in the technology industry, holds no position in any company named in this series, has no clients in it, and is selling nothing to anyone in it.

Start with the seller, because the seller felt this first.

A HubSpot account executive went into 2026 with a number. During that year their company cut the entry price of its AI agents, introduced 28-day free trials, moved two agents onto outcome-based pricing, and allowed customers to convert contracted seat spend into credits without waiting for a renewal.

Every one of those decisions was correct. Every one of them made that seller's year harder.

Sales cycles lengthened as buying committees got larger. Net new customers came in around 7,000 against a company target of 9,000 to 10,000. Net revenue retention fell to 102%. And the quota, the commission timing and the definition of a good quarter did not move an inch to reflect any of it.

That is the whole of Episode Three. The commercial model moved. The things that measure and pay for it did not. And the market, holding the same unchanged instruments, read the entire thing as decline.

What they actually did

This matters because almost nobody else has done it.

In Episode Two the numbers were these. Seventy per cent of buyers want to pay for what they consume or what they get. Two per cent of vendors sell outcomes. Sixty nine per cent answered agentic AI with a higher tier or an embedded feature, because a tier requires one meeting and a meter requires the operating model.

HubSpot built the meter.

The route took two years. A data enrichment charge in 2024. A pooled currency at about a cent a credit in 2025. Credits extended across agents through late 2025. Then, on 14 April 2026, the decision that separates them from the other ninety eight per cent.

The Customer Agent moved from $1 per conversation to $0.50 per resolution. The Prospecting Agent moved to $1 per recommended lead. In June the Content Agent was repriced at $10 per piece.

Read that first change again. They stopped charging for the attempt and started charging for the result, and in doing so they halved their own price per unit. A conversation that fails now earns them nothing. That is not a pricing tweak. That is the company accepting the risk of its own product not working, which is the thing every buyer has wanted for thirty years and almost no vendor will do.

They also published the rate card. No bundle, no contact sales, no opaque conversion. Fifty credits to resolve a conversation, a hundred to recommend a lead, at a cent a credit. Anyone could do the arithmetic, including their competitors.

What it cost them

8 May 2026. The pricing news lands. The shares fall 19% in a day, to $197.35.

6 August 2026. Second quarter results. Revenue $911.7m, up 20%. GAAP operating income $43.3m against a loss the year before. Adjusted earnings of $3.26 a share against $3.02 expected. A beat on revenue, a beat on profit, a beat on earnings.

The shares fall 19%, having been down 20.2% intraday. The largest single-day drop in the company's twelve years on the New York Stock Exchange.

Full-year revenue guidance had been trimmed by about $22m, which is six tenths of one per cent. Net revenue retention had slipped a point, to 102%.

11 November 2026. The stock sits at $196.99, down 50.91% on the year.

Now the honest part, because the easy version of this argument is wrong. That 51% is not all the pricing model. 2026 has been brutal for application software generally, the Nasdaq 100 has come well off its October peak, and peers with entirely different pricing have been hit too. Anyone telling you a single decision cost HubSpot half its value is selling you something.

The two single-day falls are a different matter. Those are specific, dated, and attributed by the reporting to the pricing transition itself. A fifth of the company on the day they announced it. A fifth again on the day they beat their numbers.

The mechanism, which is the actual point

Here is the sentence that should stop any executive considering this move.

Net revenue retention fell to 102%, and the Chief Financial Officer explained that part of the reason was customers swapping seats for credits.

Sit with that. The company built a mechanism to let customers convert fixed committed spend into variable consumption. Customers used it. That is the mechanism working precisely as designed, on the first attempt, at scale.

And it arrived in the accounts as a decline.

Net revenue retention was built for a world where the only directions were up and out. More seats, higher tier, or churn. It has no way to represent a customer who spent the same money through a different meter, and no way at all to represent one who will spend more through it next year. A customer converting seats to credits looks identical to a customer cutting back.

This is the same defect this series has been naming all year, in a third costume. Net revenue retention could never separate efficiency from abandonment. Contracted backlog cannot separate conversion from decline. And now we can add that none of them can separate a customer moving onto your new meter from a customer walking away from your old one.

The instruments cannot see the transition. So the transition reads as deterioration, every time, for everyone who attempts it.

What the chief executive was actually saying

Yamini Rangan was unusually direct about the model, and her description of it is worth quoting because it was the clearest statement of the hybrid position from anyone running a company this size.

"A hybrid model of seats and credits."

Not a replacement. Not an abandonment of the seat. Both, deliberately. Seats carry access to the platform, the data and the capabilities. Credits carry the work the agents do. Her rule for which is which is the cleanest I have seen anyone state publicly: if it makes a go-to-market team member more effective, it belongs to the seat; if it is not tied to a person at all, it belongs on credits.

That is the answer. It is also, almost exactly, what I argued on Friday. Keep the predictable line, put a variable line beside it, and let the variable line grow with usage rather than with hiring. Monthly credit consumption at HubSpot has roughly doubled over the year. The variable line is doing precisely what the variable line is supposed to do.

And the company reporting it is down a half.

Then the board changed the chief executive

On 12 November, eight days after those third quarter results, Yamini Rangan stepped down as chief executive and left the board on the same day. Brian Halligan, who co-founded the company and ran it until 2021, returned as chief executive with immediate effect. The chief financial officer is leaving too, in February, in a retirement the company describes as long planned.

The shares rose about six per cent.

Put the three dates next to each other, because the sequence is the entire argument of this episode.

8 May. She announces the meter. Down 19%.

6 August. She delivers it, and beats revenue, operating income and earnings while doing so. Down 20%.

12 November. She goes. Up 6%.

In five years she took the company from roughly 135,000 customers to more than 311,000, and revenue from $1.3bn to about $3.7bn. The market's verdict on her departure was a gain.

Now the part that is easy to get wrong, and most people writing about this will get it wrong this week.

The cheap reading is that the market cheered the removal of the person who built the meter, and that the lesson is therefore not to build meters. That reading is tidy, it is quotable, and it is incomplete.

Look at what Halligan has actually been handed. The company calls it a refounding, across product, go to market and culture. The chair says the board is united behind it. He has told Bloomberg that software companies need to become "very, very AI-pilled", and that he came back to rebuild the business rather than watch it shrink or be taken private. The fourth quarter and full year guidance was reaffirmed on the same day, unchanged.

That is not a board deciding the meter was a mistake. That is a board deciding that changing the pricing and the metering was never going to be enough on its own.

Rangan moved two layers, and moved them further than almost anyone in enterprise software. Pricing and packaging. Metering and entitlement. What she did not move, because no chief executive moves it in eighteen months against a falling share price, was everything underneath: the compensation design, the revenue architecture, the numbers handed to the market, the organisation and the go-to-market motion built for seats.

A partial crossing is the most exposed position available. The old model's instruments still measure you. The new model's economics have not arrived yet. You are carrying the cost of the transition and none of its benefit, in public, every ninety days.

She got two layers across. The board has now given the founder a mandate for all nine.

Why this is the bill for Episode Two

Episode Two asked why sixty nine per cent of vendors answered agentic AI with a higher tier, and answered that a tier is a price change while a meter is an operating model change.

That was the cost side. This is the consequence side, and it is worse than expensive.

A chief executive watching HubSpot this year has learned something very specific. Announcing outcome pricing cost a fifth of the company in a day. Delivering it, and beating every number while doing so, cost another fifth. Then it cost the chief executive her job, and the shares went up when it did. Meanwhile the vendor down the road who put the AI in a higher tier and raised the renewal by seven and a half per cent had a far quieter year, and still has a chief executive.

That is not an argument for the tier. It is an explanation of why the tier keeps winning, and it is a sharper one than cost or laziness. The tier wins because the meter is punished, on a schedule, by instruments that cannot read it, and no chief revenue officer or chief financial officer gets to opt out of being measured by them.

Two per cent sell outcomes. Now you know why it is two.

What an operator should take from this

Three things, and none of them is "do not do it".

The transition has a reporting problem before it has a revenue problem. HubSpot's revenue grew 17% in the third quarter and its margins expanded. The damage was done by a retention metric that fell one point for a reason that was structurally good news. If you are planning this move, the number you need to prepare is not the revenue forecast. It is the explanation of why your existing metrics are about to look worse, delivered before they do, not afterwards.

Build the second number yourself. Nobody is coming to invent consumption-era disclosure for you. If seats converting to credits will dent your retention figure, then publish converted spend separately, or report retention excluding conversions alongside the standard calculation, or show credit consumption growth as its own line with its own history. Hand the market something that can tell conversion from decline, because the standard set cannot and will not.

Sequence it so the sales floor is not the shock absorber. HubSpot cut agent prices, added free trials and changed the meter, and the people carrying a quota absorbed longer cycles and lower average prices while the comp plan stayed where it was. If the model moves and compensation does not, you have not run a transition. You have run an unfunded tax on your best people, and they leave.

The position

HubSpot did the thing this entire series argues for. They built a real meter, moved to a price per result, published the rate card, and let customers move their own money from the fixed line to the variable one.

The market took a fifth for saying it, a fifth again for doing it well, and then it took the chief executive.

The lesson is not that they were wrong. Credit consumption doubled, revenue grew 17%, margins expanded, and the company could describe its target model in a single sentence, which is more than most of this industry can manage.

The lesson is that a partial crossing is the worst place to stand. Two layers moved and seven did not, so the old instruments kept scoring her and the new economics had not arrived yet. There is no version of that which looks good for eighteen months, and eighteen months is longer than most boards will wait.

Which leaves a question for everyone still deciding.

The founder is back, talking about a refounding across product, go to market and culture. That is the right scope. It is also the scope that was available to the last chief executive only in theory, because nobody gets a mandate that wide until the share price has already halved.

You can change your pricing in a quarter. You can change your metering in a year.

Changing all of it apparently requires a crisis, a founder, and somebody else's career.

Next: Season 3, Episode 4. The Capital Turns. Marks not taken, exits cancelled, and what happens when half a trillion of credit meets a revenue line nobody can forecast.

The Agentic Commercial Model. agenticcommercialmodel.com

Sources

All Season 3 episodes →Subscribe on LinkedIn →