Why I Stopped Billing by the Hour

The problem wasn't the hourly rate. It was using a single measure for work of different natures.

A few weeks ago a client wrote to me. The quote request form on his site, built with a plugin he had chosen some time before, works: requests come through. Except that the notification emails appear unreadable in some email programs and perfect in others, and this has been going on for months without anyone understanding where the problem originates.

I closed that job in a few hours. Not because I'm fast at pressing keys, but because I've seen that type of behavior before and I know the few causes that generate it. Someone who has never dealt with it could take much longer, spending almost all that time ruling out wrong hypotheses: the mail server, the domain, the theme, when the problem lies elsewhere.

Here's the point that made me change models, and it doesn't concern me. By the hour, the bill for the two situations would have been exactly opposite to the value delivered: higher for those who know less.

What an Hour Really Measures

There's an anecdote that's been around for a century and is worth more than any reasoning about prices.

In the 1920s Henry Ford has a generator stopped in his plant and none of his technicians understands why. He calls Charles Steinmetz, an electrical engineer. Steinmetz listens to the machine, walks around the facility for a couple of days, then takes a piece of chalk and draws a mark on one side. He says to remove sixteen coils from the winding at that point. The generator starts up again.

Then he sends the invoice: ten thousand dollars. Ford, astonished, asks for a detailed breakdown. Steinmetz sends it to him: one dollar for the chalk mark, nine thousand nine hundred and ninety-nine for knowing where to put it. An hourly invoice for that intervention would have counted two days of walking around the plant. It would have billed the chalk.

The Hour Has Always Measured the Wrong Thing

Nothing sudden happened and I distrust explanations that blame the latest novelty.

The hour has always measured input, meaning how long I sit, never the result, meaning what changes for you. It worked for decades not because it was right, but because it was convenient and close enough to the truth: those who knew more took less time, but not incredibly less, and the math worked out.

The problem is that this approximation breaks down at the extremes. And my work lives at the extremes: the interventions where experience truly makes a difference are exactly those where the hour matters least. It's the case I started with. The more I know, the less the hour represents me.

Then there's a consequence that people prefer not to mention. An hourly model contains a structural incentive: the slower I am, the more I earn. I'm not saying that those who bill by the hour profit from slowness—that would be foolish and an insult to many serious professionals. I'm saying that the incentive is written into the structure, not the intentions, and it bothers me. I prefer a model where my interest and yours point in the same direction.

There's finally a practical reason, and perhaps the most decisive one. The tools of this trade change constantly. Every time one changes, the time needed to do a certain thing shifts one way or another. If the price is tied to time, every new tool forces me to reconsider the price list, while the value I deliver to you has remained identical. A price tied to value, however, doesn't move: it's linked to what truly matters to you. This isn't an original position. In the consulting market, the shift from the meter to a price defined in advance is mainly requested by those who buy, because a fixed-price quote transfers to the supplier a risk that was previously entirely borne by the client. It's the clients who want it, even before the suppliers.

First of All: Maintenance and Support Are Not the Same Thing

This distinction generates more confusion than any other, so I'll explain it before the rest.

Routine maintenance is continuous oversight, not an intervention. Core, theme, and plugin updates with compatibility verification, backups with actual restore testing, uptime monitoring, error reporting and known vulnerability alerts, anomaly resolution. It's not counted in terms of interventions because it's not made of interventions: it's made of constant attention. That's why it has an annual fee and, for this reason, is the mandatory foundation of any site I manage. Without that foundation I can't guarantee anything, and anyone who promises support on a site they don't oversee is selling smoke.

When a plugin breaks after an update, when a form stops sending, when a critical security patch arrives: I'm in there. I don't count them, I don't quote them, I don't invoice them. They are the service.

PowerUnit support is everything else. That is, everything that isn't routine maintenance: the new landing page, the product page, the newsletter, the SEO review, the accessibility intervention on material published after launch, the new analytics tracking. Work that adds something, not work that keeps something standing. The example I opened with is exactly the second case, and it's a boundary worth seeing in action. That plugin wasn't broken: it did exactly what it was designed to do. It was a limitation, and limitations aren't repaired: they're worked around. Routine maintenance deals with what has broken; support deals with what never worked as needed.

What a PowerUnit Is (and What It Isn't)

A PowerUnit is the unit with which I measure the value of an intervention: required competence, complexity, and tools used. Not duration.

The exchange rate is declared: 1 PowerUnit equals 30 euros in the prepaid package and 50 euros for a single intervention outside the package. Units are divisible into half points, have a validity of 48 months, and the updated balance is provided after each intervention. They're valid for everything I do, not just WordPress: also for SEO, digital marketing, and data analysis.

On price list transparency I want to be precise, because it's the point on which the credibility of the entire model is at stake and I have no intention of making it sound better than it is.

There is no publicly available price list. On the service page there are some indicative examples that give anyone evaluating the order of magnitude: how much it costs to insert a news item, how much for a landing page, how much for a complete accessibility audit. Those who activate the package receive a reference price list for the most common tasks.

On that price list the values can vary and the reason is simple: the same request on two different sites is not the same. A product page on a clean template and one on a template layered with three years of others' modifications are not the same work, and pretending they are means either overpricing the first case or losing money on the second. So each request is evaluated case by case, and the value in PowerUnits arrives before the intervention, with the scope written out. If something changes during the work, the additional part is quoted separately and is not executed without a yes.

Transparency isn't about having a price displayed in the window. It's about knowing it before saying yes.

It's also worth saying what a PowerUnit is not. It's not a subscription: the package is tied to services, not a fee, and if you don't ask for anything you don't consume anything. It's not a currency with a hidden exchange rate: the exchange rate against the euro is declared. And it's not a way to charge more for the same work: the convenience threshold is arithmetic, not rhetorical. The package is worthwhile from thirteen units up; below that threshold, the single intervention at full price costs less. I say this before the client asks, because it's a calculation anyone can redo in ten seconds.

Who Takes the Risk

This is the part that matters more than all the others combined and, usually, is the one that on commercial pages you read at the bottom, in small print.

By the hour, the risk of overrun is borne by the client. An estimate is agreed upon, work begins, and if the intervention proves more difficult than expected, the bill grows. The total is discovered at the end, when nothing can be done anymore.

With PowerUnits, the risk is mine. The agreed value refers to the scope defined together: if it takes me twice as long as expected because I underestimated the problem, that extra time doesn't enter the bill.

This works for structural reasons, not generosity. The person who gives you the price is the same person who then executes the work. If the value were set by someone who doesn't bear the consequence of the overrun, value-based pricing would be a bet made with someone else's money, and the person making it covers themselves in the obvious way: inflating the quote. Taking the risk is credible only when the person quoting and the person working are the same. There's also an effect I only noticed after a few months. With the stopwatch running, no one calls you for a trivial matter. The message that begins with "sorry, maybe it's silly, but I noticed that…" doesn't get sent, because the client wonders if they'll receive an invoice for two minutes of attention. Those unreported "trivial matters" are at the origin of a huge portion of the problems I see: an ignored warning, a plugin behaving strangely for weeks, a form that occasionally doesn't send.

On a Client's Site the First Priority Isn't Speed

Working on your own content and working on someone else's site are two different professions, and the difference isn't one of scale. It's one of responsibility.

On my site I can afford to experiment. If something breaks, I broke it. On yours, no. On yours there's a form from which requests arrive, there are pages that rank, there's people's data and a policy that must remain compliant. Every intervention, even a small one, carries a question that doesn't concern the intervention itself: what does it touch, what can it break, what happens if it goes wrong, and how do you go back.

The dependencies between parts of a site are clearly visible on an e-commerce. Changing how shipping costs are calculated might not break anything visible and, meanwhile, silently cause some orders to fail, only in certain combinations. It's the worst kind of failure that exists, because it leaves no trace anywhere: those orders don't even come into existence, and there's nothing to look at in the admin.

Then there's a fourth question: it's the one that determines where the work stops on its own. Some actions are reversible and others aren't: sending something externally, touching people's data, modifying definitively, modifying a price list. Before those, the work stops and asks you, even when the answer already seems clear. Thresholds are set beforehand, because afterward they're no longer useful.

That part is never visible. It doesn't appear in the results, there's nothing to show, and when it's done well, it seems like it wasn't there. But it's often longer than the intervention itself and is the foundation of my work.

And there's a reason why you can't skip it by looking at the final result. Work done badly in an obvious way is recognized immediately, and in fact that's not what causes damage. The dangerous one is the other: the result that seems complete, orderly, plausible and that, precisely for this reason, no one rereads. Things that break in production almost never look wrong.

And it's also the part that doesn't get shorter. The tools with which you produce something change and can make it faster to put together: what doesn't change is the verification, because it doesn't depend on how it was produced, but on what can break when you put it online. If anything, the opposite is true: the easier it becomes to produce quickly, the more important is the person who checks before publishing.

That's why I don't sell speed, and I'm a bit wary of those who offer it. I don't promise I'll take less time. I promise that when I'm done the site will work.

On where exactly this threshold lies, incidentally, the European legislator has expressed itself more clearly than many quotes. In the artificial intelligence regulation, in Article 50, it was established when content can be considered attributable to someone, and two conditions are indicated that must apply together: substantial review by someone with professional competence on the matter and an identifiable subject who assumes responsibility for it. With a clarification worth rereading: substantive verification is the minimum, correction of form is not enough. It's the most concise definition I've read of what distinguishes signed work from simply delivered work. And here lies the most concrete reason why the hour, in this profession, doesn't work. Control is invisible and, by the hour, everything that is invisible must be justified after the fact to a client who has only seen the result. A value agreed upon in advance doesn't need to be justified afterward.

When a PowerUnit Isn't Enough: The Fixed-Scope Project

A unit of value works well for recognizable interventions. It loses meaning when the work is complex, articulated, and composed of parts that influence each other: a complete redesign, a platform migration, building a members area.

There, counting units would be like weighing a house in bricks. The correct model is the fixed-scope project: an agreed price for a described result, broken down into phases with verifiable deliverables. Here too the risk of overrun remains mine, but defining the scope becomes a job in itself and, in fact, is done before signing anything.

It's not a novelty and doesn't need to be defended: it's the way complex work has always been commissioned.

When Even the Scope Can't Be Defined: Fixed-Budget Research

Then there's the most interesting and most awkward case to price.

A client asks whether a process that currently takes a person two days a month can be automated. Or whether two management systems that don't talk to each other can be integrated without rewriting half the system. Or whether a platform migration can preserve ranking history instead of resetting it.

These are feasibility questions and they have a characteristic that breaks both previous models: neither party knows, at the outset, what will be found. PowerUnits don't work because there's no recognizable deliverable to put on the price list. Hours don't work because no one can estimate them in advance.

In these cases, the model I use is the fixed-budget exploratory phase: we agree on how much is spent and how much is explored, not on what is found. Fixed budget, defined duration, open outcome.

And here the thing must be said that makes the model acceptable instead of embarrassing: the deliverable is not the solution, but the documented answer. A "it's not feasible with this platform, here's why, here are the three alternatives and what each would cost" is a delivered result in every respect. It's worth exactly as much as a yes and much more than the money the client would have spent discovering it themselves after commissioning the implementation. The value of research is not success: it's the reduction of uncertainty before a major expense.

Each phase closes with a decision, not a delivery: move forward or stop. If moving forward, the next phase returns to being a fixed-scope project, because now the scope is known. I declare the limit of this model myself, because it's what maintains the honesty of the rest: research is the model that shifts the most risk onto the client; therefore, its budget must be small relative to the project it serves to decide. If an exploratory phase costs as much as the implementation it should enable, it's not research. It's a disguised quote and should be refused.

Why in Italy This Fourth Model Is the Most Difficult

Of these four ways of measuring, the last is the one that struggles most to establish itself here. And the reason isn't economic, but cultural: it concerns which part of the work we're accustomed to considering worthy of being paid.

Thirty years ago no entrepreneur would have dreamed of having a qualified technician work without paying them. The expert who came to calibrate the machine was compensated. "You have to pay them!" was a statement with an exclamation point. For certain categories of work, and digital is among these, over the years it has become a question: do you have to pay them?

The reason, I believe, is not stinginess. It's that digital work often doesn't leave an object.

The mason leaves and a wall remains. Instead, someone who handles site maintenance leaves and the site remains exactly as it was before. The paradox of successful maintenance is that it looks very much like having done nothing, and the only times it becomes visible are those when it was missing. It's a profession that's only seen from the hole it leaves, exactly like the control I was talking about above.

With research the same thing happens, but worse. Because there the missing object is not only the visible result: it's also the only part we're accustomed to evaluating, namely the thing built. In few other fields is diagnosis taken for granted as free: you pay for the visit even when the doctor says there's nothing, you pay for the surveyor's inspection even when they conclude that wall can't be knocked down. In digital, the phase with the highest density of competence has been normalized as complimentary; and in fact asking to pay for an exploratory phase means asking to pay for exactly the part that everyone has learned to receive for free.

A fact, to show it's not just a feeling. Law 49/2023 on fair compensation also applies to professionals not registered with professional bodies, but only in relationships with public administration, banks, insurance companies, and companies with over 50 employees and over 10 million in revenue: about 17,000 companies out of 6 million. SMEs, meaning the type of client I work with every day, are outside the scope. Meanwhile, according to the Cribis study on the second quarter of 2026, in Italy only 42% of companies pay within the agreed deadline, down from 44% a year earlier, and serious delays beyond ninety days are increasing. I'm not asking for that law to be extended, and I don't believe a pricing model solves a cultural problem. The reason this observation fits into an article about prices is more concrete: a model that declares in advance what you're buying and how much it costs does exactly one thing, and it's the only thing it can do. It makes discussable what was previously implicit. The client who receives a quote for an exploratory phase can say no, and no is a legitimate outcome. What they can no longer do is receive that work without knowing they've received it.

Four Measures, Not One

The point of the article is not that PowerUnits are the answer to everything. It's that there is no answer to everything, and insisting on using a single measure was the real mistake, even before the hourly unit of measure was.

Then there remain cases where time is still the right measure, and I distrust those who write that the hourly rate is dead. Supporting an internal team, where the scope changes every day and defining it in advance would be fiction. Live training, where the time spent together is literally the product. Open emergencies, those where no one can know if it will take two hours or two days: there a value estimate would be a bet disguised as a quote, and whoever proposes it is either reckless or loading it with a safety margin that you pay. In these cases, the sensible variant is hours with an agreed maximum cap: the actual time is billed, but the client knows the cap from the first minute.

The reverse also applies. Value-based pricing works under three conditions: that the scope is specialized, that the value is communicated and approved before the intervention always and without exceptions, and that whoever sets the value is whoever executes the work. Without the second, it becomes a price decided by feel, case by case, based on how rich the client seems. Without the third it becomes an estimate made by someone who won't bear its consequences, and in that case hours are more honest. And it's exactly why, rightly, many view value-based pricing with suspicion.

In summary, the four models and the question that weighs more than all: who pays if the work proves longer than expected.

How It Works, in Practice

Maintenance is the foundation and remains an annual fee. Interventions outside maintenance are quoted in PowerUnits, with the value communicated and approved in advance. Large jobs become fixed-scope projects. Questions without certain answers become fixed-budget exploratory phases, with a final decision rather than a delivery. The two service levels, value examples, and conditions are available on the ongoing WordPress support page, along with what is included in the fee and what is not.

Closing

I return to those unreadable emails.

The value of that intervention wasn't in the few hours it took me. It was the years that allowed me to understand, by reading the description of a symptom, where it originated and where it didn't. Time was the residue, and the least interesting thing I could put on the invoice. I haven't stopped counting. I've just stopped counting the wrong thing.

Sources

• Law of April 21, 2023, No. 49, provisions on fair compensation for professional services (Normattiva)
• Cribis, Payment Study, second quarter 2026
• Regulation (EU) 2024/1689, art. 50, and EU Commission guidelines, July 2026 (EUR-Lex)

I also discussed these four models in a seven-minute conversation, generated with AI from this article.