Ten products can still be one bet

Four icons, perhaps one bet

Imagine a small team whose first app fails to gain traction. It builds three more. The apps serve different purposes, but all depend on the same store-keyword channel, similar subscription pricing and the assumption that people will search for the solution and pay. The team now sees four chances. Yet if acquisition is the untested premise, the fourth app may simply reproduce the first uncertainty.

This is a hypothetical case, not a diagnosis of a particular company. The question is more specific than whether teams should focus or diversify: before a shared acquisition condition has been validated, adding products that depend on it may add maintenance obligations without adding comparable independent chances of success.

That proposition can be wrong. A new product that attracts users when the original channel fails, or produces distinct learning at little extra cost, can strengthen the portfolio. The relevant count is not the number of icons. It is the number of ways those icons can survive.

Identical individual odds, different portfolios

Consider a deliberately simplified calculation. Suppose each product has a known 20% chance of reaching the same sustainable-business threshold within a fixed period. If four outcomes are independent, the probability that at least one succeeds is 1−0.8⁴, or 59.04%.

Now construct a different world. All four must first pass one shared gate, with a 25% chance of passage. Conditional on passage, each has an independent 80% chance of success. Each individual product still has a 25%×80%=20% chance. But the probability that at least one of four succeeds is now 25%×(1−0.2⁴), or 24.96%. Adding further products cannot lift that probability above the gate’s 25%.

These are invented inputs, not industry statistics or estimates of startup success. The models hold individual odds constant and change only the dependency structure. They demonstrate why individual odds alone cannot describe portfolio survival. With a fixed staffing budget, even that assumption is generous: spreading effort may change each product’s odds.

A shared gate might be a channel, a customer segment’s willingness to pay, or an ability to earn trust. Reality rarely makes it a binary switch. The switch is a device for exposing the common dependency.

The best reusable capabilities also create dependence

Shared components, release processes and support tools can reduce the cost of another product. But reusing code and reusing a business assumption are different acts. Build time, failure rates and maintenance hours can evaluate the former. The latter still requires asking why this person arrives, stays and pays.

Suppose a team has demonstrated that it can sell a particular kind of language-learning app to a defined audience. Adjacent products for that audience may be more sensible than a sudden move into racing games. The shared channel is then an accumulated capability, not merely an untested bet. Correlated risks can be acceptable when cost advantages justify them.

The objective is therefore not minimum correlation. Unrelated products may require unrelated expertise, content and support, exhausting the organization while diversifying demand. A better distinction is between shared conditions supported by evidence and shared wishes copied into every business plan. Reuse the demonstrated capability; make the next experiment address a remaining uncertainty.

Cheaper development does not guarantee a cheaper portfolio

A new product has at least two costs: creating it and remaining responsible for it while its value is unproven. Responsibility can include compatibility updates, payment problems, privacy explanations, support messages and decisions about its future. Maintenance need not be expensive for this distinction to matter. Its cost is simply not determined by the speed of initial code generation.

Attention switching is easier to miss. A prototype with no users may require little upkeep. Ten prototypes with a few users each can generate repeated decisions about small issues. Individually cheap tasks may fragment the uninterrupted time needed to test the central assumption.

Price another product against a defined period of minimum service, not just its launch date. Estimate the maintenance hours you will commit over the next three months and compare them with spending those hours on an existing product. That is a budgeting window, not permission to abandon users when it ends.

Ten failures without users may provide one piece of information

If ten apps receive few qualified visitors, they have not necessarily delivered ten tests of demand. They may repeatedly reveal one fact: the current channel does not bring the intended audience to the product. Ranking their revenue at that point can confuse not being seen with not being worth buying.

Selecting the app with the most downloads can make the opposite mistake: choosing a burst of exposure rather than a repeatable advantage. More candidates create more opportunities for an unusually good-looking result. Replicability needs a fresh observation, rather than declaring validation from the same data used to select the winner.

A portfolio can therefore be evaluated by the distinctions it makes possible. Which judgment becomes clearer with each addition? If success or failure both leave us saying only that we need more traffic, the experiment is weak. A small test separating failure to attract users from failure to help arriving users complete a task may teach more than four complete launches.

Test whether you are diversifying or copying uncertainty

Choose two or three candidates. For each, specify how qualified users arrive, which completed task demonstrates value and which behavior would support the business assumption. Label shared dependencies consistently. Calling channels store discovery, SEO and word of mouth does not make them independent if they all rely on the same keywords and need.

Test the common gate first. If acquisition is uncertain, test two clear task entry points over comparable periods with comparable audiences and budgets where appropriate. Record qualified visits, actual task completion, payment or meaningful purchase intent, and reasons for stopping. Intent is not revenue; a store click is not an installation. Products serving different audiences should be checked against their own prewritten business conditions rather than forced into one conversion-rate ranking.

Give a promising candidate a fresh test with a new group, a later window or another viable acquisition source. One replication guarantees nothing, but helps distinguish exposure from capability. If every candidate stalls at the same gate, investigate that gate. If arrivals cannot complete the task, investigate the product. If completion produces no willingness to pay, investigate the relationship between delivered value and pricing.

With tiny samples, a handful of clicks or zero purchases cannot establish precise success probabilities. Record the conclusion as unresolved, retain behavioral evidence and interviews, and set the next affordable budget and stopping condition in advance. The aim is not to calculate destiny. It is to stop answering the same failure by making another product.

A portfolio earns its value through different ways to survive

Multiple products can offer two distinct benefits: shared capabilities make each cheaper, and different survival conditions reduce the chance of losing every revenue stream together. The first can be strong while the second is weak. The second can be strong but too expensive. Separate them to understand what the portfolio buys.

A worthwhile portfolio need not look maximally varied or contain the most download buttons. It should answer a concrete question: if our most important shared condition stopped holding tomorrow, which products would still be needed, and how could we reach the people who need them?

If the answer is none, the company need not immediately enter an unfamiliar field. It has a concentrated business. Recognizing that concentration makes it possible to assign appropriate cash, service commitments and a genuinely informative next experiment.