08 - Side project launch: real Italian case studies

Why you will not find stories with revenue figures here

Stories of people who built on their own a product earning thousands of euros a month are the most-read material on the subject and the least useful for deciding. The stories come from those who made it, and nothing is told about those who tried and earned nothing: the sample is rigged from the start. On top of that, a revenue figure attributed to a person without a verifiable public source is not data, it is hearsay. That is why the article contains no cases with names and revenue: I found none I could document with an open source, and inventing them would be worse than staying silent.

What you find instead is a tool: how to read a case when you meet one, and a program that shows you where the real uncertainty lies in a launch plan. The numbers in the example are all declared invented.

Five questions to ask of every success story

Faced with a story with a figure, before letting yourself get excited, run the text through these questions, in the order in which it pays to ask them.

• Who says so? The author, an interviewer or a third party reporting? A second-hand figure is hearsay.
• What does it measure? Revenue, turnover, takings or profit are not the same thing, and neither are "recurring" and "one-off".
• Over what period? An exceptional month is not a year.
• From what starting point? An audience already built, a network of contacts, an already known product change everything.
• At what cost? Hours of work, expenses, contributions and taxes take away a part of what you read.

The order is not random: the first question decides whether it is worth asking the others. If the answer to the first is "it is not known", stop there and treat the story as a tale, not as a precedent. The last two are the ones that separate what can be imitated from what depended on conditions you do not have: a success story with an audience of thousands of people already acquired tells you nothing about your first month. For the numbers on the progress of a small project it is more reliable to build your own, with explicit assumptions, as in the next section.

Where the uncertainty really lies: a simulation

Take a simple launch plan: a certain number of visitors each month, a percentage who buy, a price. Whoever makes the plan often thinks luck is the main risk: "maybe fewer will come". The program below tries to measure how much chance weighs and how much the assumption weighs. It simulates a thousand launches for each of three conversion percentages (0.5%, 1% and 2%) with 3,000 visitors a month for six months; all the parameters are invented and serve only to show the mechanism.

The function random.binomialvariate (available from Python 3.12) generates how many visitors out of 18,000 buy, given the rate. The generator random.Random(42) has a fixed seed, so two runs give the same result; statistics.quantiles with n=10 returns the deciles, from which we read the 10th, 50th and 90th percentiles. The monthly revenue at the end assumes nobody cancels, a very optimistic assumption that the program declares in the printed text.

# How many paying customers after 6 months? A simulation with INVENTED numbers,
# to show where the uncertainty lies. Requires Python 3.12 or later
# (random.binomialvariate). The fixed seed makes the output repeatable.
import random
import statistics

VISITATORI_MESE = 3000      # assumption
MESI = 6
SIMULAZIONI = 1000
PREZZO = 9                  # euros per month, assumption

rng = random.Random(42)
for conversione in (0.005, 0.01, 0.02):       # 0.5%, 1%, 2% of visitors pay
    clienti = [rng.binomialvariate(VISITATORI_MESE * MESI, conversione)
               for _ in range(SIMULAZIONI)]
    d = statistics.quantiles(clienti, n=10)     # deciles: d[0] = 10%, d[4] = 50%, d[8] = 90%
    print(f"conversion {conversione:.1%}: customers at 6 months, 10% {d[0]:.0f} | median {d[4]:.0f} | 90% {d[8]:.0f}"
          f" -> monthly revenue at the median {d[4] * PREZZO:.0f} EUR (if nobody cancels)")

# Output:
# conversion 0.5%: customers at 6 months, 10% 78 | median 90 | 90% 102 -> monthly revenue at the median 810 EUR (if nobody cancels)
# conversion 1.0%: customers at 6 months, 10% 162 | median 180 | 90% 197 -> monthly revenue at the median 1620 EUR (if nobody cancels)
# conversion 2.0%: customers at 6 months, 10% 334 | median 360 | 90% 382 -> monthly revenue at the median 3240 EUR (if nobody cancels)

Look at the width of the intervals: with 1% conversion the 10th percentile is 162 customers and the 90th is 197, a band of about thirty people around 180. Chance, at the same rate, moves little. What moves a great deal is the rate: from 0.5% to 2% the median goes from 90 to 360 customers, four times as many. That is why a story that says "I converted 2%" is no use to you if your real rate is 0.5%: the source of uncertainty is the assumption, not fate. The program does not know what your market will do: it tells you which numbers are worth measuring first, namely visits and purchase percentage, before spending months building.

How to get your own rate, before building

You discover the rate with a small experiment, not with a story. Publish the page that describes the product, state a price and a way to book or pay in advance, bring fifty or a hundred real people there (not friends) and count how many act. Fifty or a hundred is a rule of thumb, not a standard: it serves to see whether the percentage is closer to 0.5% or to 5%. The pre-mortem helps you choose which risk to verify first, and the article on organic marketing shows how to bring visitors with no budget.

Once you have a measured rate, the other numbers can be calculated. How much is left of a 9-euro subscription you see in the article on Stripe (the price-list fees) and in the first of the series on the break-even point; how much is left after tax under the regime forfettario, with a simplified example.

Frequently asked questions

Are there documented Italian cases of indie products?

Probably yes, but in this article I cite none: I found none with an open, readable public source. If you want to look for some yourself, apply the five questions above and distrust every figure without a primary source, such as a post signed by the author or a public document.

How many customers do you need to live off a digital product?

It depends on the price, the costs and the taxes, not on a universal number. The program in the first article calculates how many subscribers are needed to reach a net target: change your parameters and read the result. Then add taxes and contributions with your accountant.

Does the simulation predict my result?

No. It uses invented numbers and only shows how uncertainty is distributed between chance and assumption. If you redo it with your measured data, it tells you how wide a band you should expect, under the same assumptions.

What you can do now

You can read a success story with five questions, tell the primary source from the second-hand tale, simulate a launch plan to see whether chance or the assumption weighs more, and decide which number to measure first. It is the end point of the series: from the stack to support, from hosting to payments, each article tried to give you a number you can redo.