Chapter 3

Users Aren't Customers

The Seduction of Visible Metrics

At one point, I helped launch a small SaaS application aimed at technical professionals. We made what seemed like a generous and fairly rational decision: nearly everything would be free.

The theory was not unusual. Remove the friction. Let people experience the product before asking them to buy anything. Give them enough time to understand the value, incorporate it into their work, perhaps even become a little dependent on it. Once that happened, some reasonable percentage would upgrade when we introduced paid capabilities.

For a while, the numbers rewarded that theory.

Hundreds of people created accounts. A meaningful percentage returned. Some sent detailed feature requests. A few wrote messages about the product that were more enthusiastic than anything we would have been comfortable putting on our own website.

The dashboard looked good. That sentence has caused more trouble in software businesses than it deserves.

If I had shown you only the engagement data, I could have made a respectable argument that we had traction. People were signing up voluntarily. They were using the product. Some came back. They were engaged enough to suggest improvements. None of those numbers was fabricated or manipulated. We were not fooling ourselves with fake accounts or buying traffic from a bot farm in a distant country whose residents had developed an inexplicable passion for our niche SaaS application.

Real people were using it. Then we introduced a paid plan. Almost nobody bought it. My first instinct was exactly what you would expect from somebody with a builder’s brain: the paid version must not be good enough yet.

That explanation was wonderfully convenient because it came with a backlog. We could add more capabilities, move desirable features behind the paywall, improve the upgrade screen, make the distinction between plans clearer, and perhaps create one or two features impressive enough that a free user would finally surrender a credit card.

Before doing much of that, we started looking more closely at what the existing behavior actually meant.

We reduced the free offering. We made the professional use case more explicit. We moved the purchase decision earlier instead of allowing people to remain indefinitely in a comfortable free relationship with us. More importantly, we stopped asking users what additional features might persuade them to upgrade.

We started asking what they were actually doing with the product.

What was happening before they opened it? What happened if they could not use it? Was this part of paid work, or was it merely useful? Was the product saving time that had economic value, or time that was simply nice to get back?

Those conversations divided the user base in a way our analytics had not.

A smaller group was using the product as part of work they were paid to perform for other people. For them, saving an hour was not simply a convenience. It affected how much work they could complete and deliver. Their relationship with the same application was commercially different.

We rebuilt the paid proposition around that use case. We removed several features from the pitch rather than adding more.

We also raised the price.

The result looked strange if your favorite metric was the number of free users. Signups declined. Total active usage became less impressive.

Revenue went the other direction. Within the following months, the application had fewer active users than it had during the free-growth period, but the business was producing more than $10,000 in monthly recurring revenue and we had a much clearer idea of who the customer actually was. That experience changed the way I look at usage data. I stopped treating the word user as if it contained a business model.

The mistake was not that we had measured usage. We absolutely should have measured it. The mistake was allowing a set of true numbers to answer a question they had never been designed to answer.

This is the seduction of visible metrics.

Software gives us an extraordinary ability to observe behavior. We can know how many people created accounts yesterday, how many returned today, which screen they visited, how long they stayed, which feature they used, where they abandoned onboarding, how frequently they came back, and whether the little line on the graph is moving in a direction that allows us to sleep slightly better.

The numbers feel objective because they are objective. Twenty-seven signups are twenty-seven signups. Nine weekly active users are nine weekly active users. A forty-two percent increase in sessions is, assuming the analytics implementation is not doing something creative, a forty-two percent increase in sessions. The problem appears when we move from measurement to interpretation.

A founder sees a thousand accounts and thinks, A thousand people want this. Maybe. A founder sees increasing weekly usage and thinks, We’re getting traction. Possibly. A founder sees people returning and thinks, Monetization is the next step.

It might be.

Notice what happened in each sentence. The metric was precise. The interpretation quietly became larger than the metric.

Dashboards are especially good at encouraging this because they place different behaviors beside one another and make them look as though they belong to the same story. Visitors become signups. Signups become active users. Active users become retained users. Somewhere farther to the right, preferably after a tasteful arrow, the business becomes revenue.

Visually, the progression makes perfect sense. Commercially, every arrow contains an assumption.

The person who visits may never create an account. The person who creates an account may never use the product. The person who uses it may genuinely enjoy it and continue using it for years while having no reason whatsoever to pay.

That last possibility is the one builders often underestimate.

We tend to imagine that value naturally accumulates toward payment. If somebody uses the application repeatedly, the reasoning goes, they must value it. If they value it, eventually some appropriate premium feature, usage limit, price point, or upgrade message will convert that value into revenue.

Sometimes that is exactly what happens. Sometimes someone can honestly love your product and still have no commercial reason to buy it.

That is not a contradiction.

Think about how many useful things you consume without ever seriously considering what you would pay for them. Free utilities, open-source software, browser extensions, newsletters, templates, mobile apps, online calculators, AI tools, community resources, and countless other products can become genuinely useful parts of someone’s life without crossing the threshold into a purchase.

The user is not being dishonest. The product is not necessarily failing. The founder may simply be observing a different behavior from the one the business ultimately requires.

This becomes uncomfortable because usage is visible much earlier than payment for many products. A new SaaS founder may wait weeks for the first serious customer while accumulating dozens or hundreds of smaller actions that look increasingly encouraging. Those actions are irresistible raw material for a story about progress.

Five signups become twenty-five. Twenty-five become one hundred.

People start coming back.

The founder is not inventing optimism. The optimism has data.

That is precisely why evidence classification matters.

Weak evidence is rarely false evidence. It is usually real evidence being asked to support a stronger conclusion.

A user is a person who uses the product. A customer is a person or organization involved in the economic exchange that supports the business.

Sometimes those are the same person. Sometimes the user and payer are different people inside the same company. Sometimes millions of users support a perfectly legitimate business through advertising, transaction fees, network effects, sponsorship, or another model in which the user never pays directly.

So the point is not that only credit-card transactions matter. The point is that you have to know which behavior supports your business model.

If your business requires people to pay $79 a month, then 10,000 people enthusiastically using the free version is not the same commercial result as 1,000 people paying $79 a month. You may prefer the first group for product learning, growth potential, word of mouth, or future conversion. It may become extraordinarily valuable.

But may become is doing important work in that sentence. Until the bridge between usage and economics is demonstrated, it remains a bridge you believe exists. That is the discipline I wish I had applied earlier. When I see a dashboard now, I try not to ask whether the numbers look good. I ask what each number actually knows.

Free Changes Behavior

Free is not simply a lower price. It changes the decision.

That distinction matters because founders often treat a free user as if they are a paying customer who has not yet encountered the invoice. Sometimes that is directionally useful. Frequently it is not.

When the price is zero, a person can say yes for reasons that disappear the moment money enters the conversation. They can be curious. They can be bored. They can want to compare your product with something else. They can think it might be useful someday.

They can enjoy trying new software. They can sign up because a friend sent them the link and they have three minutes before the next meeting.

None of those motives are illegitimate. They are simply compatible with free in a way they may not be compatible with paid.

The moment you introduce a price, the question changes from:

Is this worth trying?

to something closer to:

Is this worth giving up something else for?

That is a much harder question.

Money creates opportunity cost. Even a small purchase competes with other purchases, other tools, other priorities, and sometimes the general human preference for continuing to pay nothing when paying nothing has been working remarkably well.

In business software, price can introduce additional consequences. The person may need approval. They may need to justify the expense. A credit card may not be enough because the company wants an invoice, a vendor record, a security review, or a procurement process. A $50 monthly tool can create paperwork whose emotional cost exceeds the $50 by a suspiciously large margin.

Free avoids much of that. That makes free extremely useful when you want to reduce friction around evaluation. It also makes free dangerous when you mistake evaluation behavior for purchase behavior. Imagine I offer you two versions of the same tool.

The first says:

Try it now. No account required.

The second says:

$49 per month. Start subscription.

The product can be identical behind both buttons. The person clicking the first button is deciding whether the expected benefit exceeds a few seconds of attention. The person clicking the second is deciding whether the expected benefit exceeds $49, whatever alternatives that money could purchase, the possibility of forgetting to cancel, and the general annoyance of accumulating another recurring charge. Those are different experiments. This is why removing price can dramatically increase activity without necessarily telling you much about willingness to pay.

Suppose 2,000 people try a free tool. That is useful evidence that 2,000 people were willing to try the tool under free conditions. Now suppose you add a $10 charge and fifty people buy. You have not necessarily discovered that 1,950 people suddenly stopped valuing the product. You discovered that fifty people valued it enough to cross a different threshold.

That threshold is commercially important.

The remaining users may still find the product helpful. Some may even use it constantly. They may recommend it, request features, participate in a community, write glowing comments, and tell you they would be devastated if it disappeared.

Then you ask them to pay. Some will disappear with astonishing efficiency. Again, this does not make them hypocrites.

Their previous behavior was real. So is the new behavior.

The problem arises when we assume the first must eventually produce the second. This assumption appears in many forms. “We’ll get the users first and monetize later.” “Once they see how valuable it is, upgrading will be obvious.” “We don’t want to scare people away with pricing yet.”

“Let’s make it free until we have enough traction.” All of those strategies can be rational.

Some of the largest software businesses in the world have used free products, free tiers, trials, network effects, advertising, or subsidized adoption extremely effectively. Free can lower acquisition friction, create distribution, improve the product through usage, support virality, help establish a category, and allow buyers to evaluate something before committing.

I am not interested in declaring free good or bad. I am interested in what free prevents you from learning if you never create another decision. A free product can teach you whether people use it. It can teach you where they struggle. It can show you which workflows matter.

It can reveal retention patterns, usage frequency, feature adoption, referrals, and whether people voluntarily return when nobody is reminding them. Those are valuable observations. But if your eventual business model requires the user to pay, free cannot by itself tell you what happens when the price stops being zero. That sounds painfully obvious, yet founders routinely build entire commercial theories on the assumption that usage will convert once the correct paid features arrive. The phrase I pay attention to is once we add.

Once we add team features. Once we add exports. Once we add the integration. Once we add better reports. Once we add AI.

Once we add enough limits to the free tier that upgrading becomes necessary. Sometimes the sentence is correct. Sometimes it is a way of postponing an uncomfortable discovery. The discovery may be that the current users chose the product partly because it was free. That is not an insult to the product.

Price is part of the product’s commercial environment. Change the environment and behavior changes.

You can see this in ordinary life without any software involved.

If someone offers free samples outside a bakery, plenty of people will accept one. That does not mean the bakery has discovered the correct market size for a twelve-dollar loaf of bread.

The sample tested willingness to eat bread. The purchase tests something else. Software founders occasionally perform the digital equivalent of handing out bread for six months and then becoming offended when the crowd behaves differently at the cash register.

I say that with sympathy because free feels like evidence accumulation. Every new user gives you more data. More feedback. More usage. More feature requests. The product feels increasingly real because more people are interacting with it.

And some free users will absolutely become excellent customers. The point is not to distrust them. It is to avoid assuming which ones will make that transition before they actually do. That is why the context of free usage matters so much. A person using your free product casually once a month is one kind of signal.

A person using it every day inside paid client work is another. A person who has invited a team, imported real data, and built a workflow around it is another. A person who says, “I love this,” while using it only when bored on Sunday afternoon is another. The interface may classify all four as active users. Commercially, they may be living in completely different worlds.

This is what we eventually discovered in the SaaS application I described earlier. The most important distinction was not simply between active and inactive users.

It was between people for whom the product was convenient and people for whom the product affected economically meaningful work. Free had allowed both groups to look similar for a while. Payment forced the difference into view. That is one of the useful things price does.

It creates information.

Not perfect information. A person can refuse to pay because the price is wrong, the offer is weak, the timing is bad, the buyer is wrong, or a dozen other reasons we will deal with later.

But price introduces a constraint that free removes. You get to observe what happens when value has to compete with cost. If your plan depends on converting free users eventually, I would want to know exactly what transition you believe will occur. What will become valuable enough to justify payment? What kind of user is most likely to reach that point?

What behavior today suggests they are approaching it? What happens when you actually ask? Those questions are much more useful than saying, “We have 5,000 users, so we only need one percent to convert.” Mathematically, that sentence is delightful. Commercially, the one percent has opinions.

Free can help you build an audience. It can help you build a product. It can help you learn. It can even be the correct permanent price for one side of a business model. Just do not let zero dollars quietly become proof of what people will do when the number changes.

Engagement May Still Be Valuable

At this point, it would be easy to conclude that I am suspicious of every metric except payment. I am not. That would be just as lazy as treating every signup as demand.

Usage can be extraordinarily valuable evidence. In some businesses, it is one of the most important things you can measure. The mistake is not caring about engagement. The mistake is assuming that engagement means the same thing in every business model.

Consider three applications. The first is a paid SaaS tool where each user pays a monthly subscription. The second is a free consumer application supported by advertising. The third is a collaboration product where one person’s usage makes the product more useful to other people and widespread adoption helps create the conditions for later monetization.

All three can report the same metric:

100,000 monthly active users.

Commercially, that number belongs to three different stories.

For the subscription product, active users may matter because continued use can support renewal, expansion, and perceived value. But if none of those users is paying and the business depends on subscriptions, the founder still has an unresolved bridge between usage and revenue.

For the advertising business, usage may be much closer to the economic engine. More qualified attention can create more advertising inventory. Session frequency, time spent, audience composition, and retention can have direct commercial relevance even though the user never pays the company.

For the collaboration product, adoption may create network value. A team is less likely to pay for software nobody else on the team uses. In that case, free or low-friction usage may be a deliberate part of how the product becomes valuable enough for someone else to purchase.

Same number. Different business meaning. This is why I hesitate whenever somebody labels a metric “vanity” without first asking what business model sits behind it. Impressions can be vanity. Impressions can also be the inventory being sold.

Downloads can be vanity. Downloads can also be a necessary first step in a consumer product where monetization depends on scale. Free users can be commercially irrelevant. Free users can also be the supply side of a marketplace, the audience supporting advertisers, the source of referrals, the pool from which teams spread, or the top of a very deliberate conversion system. The metric does not know whether it is vain.

The business model decides.

That matters because simple rules are attractive when the commercial side feels complicated. “Ignore users, focus on revenue” sounds wonderfully disciplined until you apply it to a product where user behavior is part of the revenue mechanism.

The better rule is more annoying:

Know why the metric matters. Suppose you run a freemium developer tool.

Ten thousand people use the free version each month. Five hundred use it daily. Two hundred have invited someone else. Eighty have created projects that appear to be connected to real work. Twenty-five companies are paying.

You could look at that data and say only twenty-five customers matter. That would throw away useful information.

The free users can tell you which workflows become habitual. They can show you which features cause people to return. They can reveal which users spread the product to coworkers. They can help you identify the point at which individual usage begins turning into team adoption.

The commercially useful question is not whether the free users matter. It is how they matter. Maybe the strongest signal is repeated use. Maybe it is collaboration. Maybe it is reaching a usage limit.

Maybe the important transition occurs when someone connects company data. Maybe the product becomes economically meaningful only when a second person joins. Those patterns can help you understand where usage begins moving toward a business outcome. But notice the discipline required here.

You are still not saying:

“People use it, therefore they will pay.”

You are saying:

“This behavior appears connected to a later commercial event, and we should test whether that relationship is real.” That is a much stronger claim because it can eventually be checked. If users who invite teammates convert at five times the rate of users who do not, that is interesting. If the pattern repeats consistently, more interesting. If you redesign onboarding to encourage the relevant collaborative behavior and paid conversion improves while other important variables remain reasonably stable, now you are learning something about the commercial system.

Engagement becomes useful when you can connect it to a decision that matters. The same principle applies after payment. Suppose a customer subscribes but never uses the product.

I care about that.

Payment is powerful commercial evidence, but it does not magically make usage irrelevant. If the product depends on recurring value, a customer who pays and never activates may be a customer who cancels as soon as the credit-card statement reminds them that the subscription exists.

We will deal with retention much later because it deserves its own chapter. For now, the important point is that behavior after payment can matter every bit as much as behavior before it.

Usage can tell you whether the product is actually becoming part of the customer’s work. Frequency can tell you whether the problem recurs often enough to sustain a relationship. Depth of use can show whether customers are receiving only superficial value or building meaningful workflows. Invitations can reveal organizational spread. Repeated use of one narrow feature can show that your sprawling application may, in practice, be a business built around one thing.

Those are not vanity observations. They can change what you build, how you package, whom you target, how you onboard, and eventually how you price. The discipline is to keep asking what the behavior predicts or enables. If you cannot answer that yet, say so. “We have strong engagement, but we do not yet know whether it predicts payment.” That is a perfectly respectable statement.

In fact, I trust it more than:

“Engagement proves people love the product.” The first sentence tells me what was measured and what remains uncertain. The second tells me someone has become emotionally attached to a dashboard. There is also an important difference between leading indicators and substitutes. A leading indicator is a behavior that appears earlier than the commercial result and may help you anticipate it.

A substitute is a metric you use instead of the commercial result because the commercial result is disappointing or absent. Those can look identical on a slide. Active usage might genuinely be a leading indicator of renewal. Or active usage might be the number you highlight because nobody is paying. Invitations might predict team conversion.

Or invitations might simply be the nicest-looking graph available this month. The distinction is not aesthetic. It is empirical. What happened afterward? That question eventually determines whether the metric deserves the interpretation you gave it.

I like engagement data most when it helps explain differences between people. Who came once and vanished? Who came back? Who built something substantial? Who invited somebody else?

Who hit a limit? Who contacted support? Who asked about pricing? Who upgraded? You are no longer treating “users” as one undifferentiated mass. You are looking for behavior that separates casual utility from stronger commercial relevance.

That is what our free SaaS users eventually taught us. The important discovery was not that free usage had been meaningless.

It had not.

The usage gave us enough surface area to notice that one group behaved differently. They were using the product in paid professional work. The same time savings meant more to them because the underlying activity had economic consequences.

Without the free users, we might have learned that more slowly. Without eventually asking for payment, we might never have learned it at all. Both forms of evidence mattered. They simply answered different questions. So keep the engagement dashboard.

Measure retention. Watch usage frequency. Study activation. Look at invitations, sessions, workflows completed, exports, projects created, or whatever behavior plausibly matters inside your product. Just give every metric a job description.

What question is this number allowed to answer? What commercial outcome do we believe it relates to? What evidence would show that belief is wrong? When you can answer those questions, engagement stops being decoration.

It becomes instrumentation.