01 Introduction
Scientific knowledge accumulates. A finding gets published, checked against other findings, and built on by people who never met the researchers who made it. Even so, science keeps only part of what it learns. Experiments that find nothing usually go unpublished, so other teams repeat them without knowing. A result that didn't work is still evidence.
Consumer companies keep far less. What a brand learns lives in its team and its slide decks, and it leaves when people leave or the company closes. What didn't work is rarely written down at all. Every new brand starts close to zero and has to relearn what thousands of brands before it already found out.
Markets make this harder than science does. An experiment can be run again, but a market never returns to the same state, so no result can be repeated exactly. That doesn't make a result less useful. It means the result has to be kept with the conditions that produced it, so a later decision can weigh it in context. A company that keeps its results this way knows more with every move. A company that loses them starts over.
02 Two ways learning fails
Most companies fail to learn in one of two ways. One is forgetting. Findings stay with the people who made them, teams don't share what they saw, and the knowledge disappears when the company closes.
The other is turning a single result into a rule. One winning ad becomes "people want this message," and that line starts steering decisions it was never checked against. The usual loop runs from result to lesson to rule, and it treats every result as a verdict. Most single results carry little signal and a lot of noise, so a company that learns this way ends up confidently wrong.
03 Evidence, not verdicts
Learning starts with beliefs, built by reviewing everything we already know, before any new result arrives. Each new result is weighed against those beliefs and against all the other evidence. Our beliefs change only as much as the evidence warrants, so a small result changes very little when a lot of other evidence points the other way.
Weighing results this way keeps the company from lurching after every result. It also means a belief has to earn trust before anyone relies on it. One way to check a belief is to ask it to predict results it hasn't seen yet, then compare the prediction with what actually happened. The practice is called backtesting, and it shows whether a belief holds up or only fits the results it came from.
Treat each result as evidence, not a verdict.
04 Infinite memory
Weighing results well is half of learning. The other half is keeping them. We call this infinite memory: nothing the portfolio learns is discarded, including results that showed nothing and results from brands that were pruned back. Every brand shares what it learns with the rest, and each new brand inherits what the brands before it learned. What gets passed on is a belief with its conditions attached: who it applies to, in what situation, and on what evidence.
A belief travels only as far as its evidence allows. Evidence that a formula supports calm and focus means one thing to a student the week before exams and something else to a person in their sixties who worries about brain fog. A finding about the student is a belief about that student in that situation. Whether it holds for the second person is a question to check with a small move before anyone relies on it there. If it holds, the belief widens. If it doesn't, the belief stays with the setting that produced it.
Evidence can matter more later than it does today. Minoxidil was developed to treat high blood pressure, and it was approved for that use in 1979. Doctors noticed that patients taking it were growing hair, a side effect with no bearing on blood pressure. In 1988, the same compound was approved as Rogaine, a treatment for hair loss. The side effect became the point because someone kept the observation. That is why no evidence gets discarded. As the terrain changes, its weight changes, and a finding that counts for little today can settle a question years later.
05 What it takes
Learning this way depends on three things that most companies don't have.
1. Evidence from far beyond our own results
The world produces far more information about demand than any one company's results. Consumer trends, market signals, performance data, scientific research, and visual styles all bear on what people want. Our own results are one source among many, and they get no special weight because they are ours. Gathering the rest at scale is hard, because it is scattered, noisy, and often expensive to reach.
2. A sense of how much each result should count
Every new result raises the same question: how much should this change what we believe? If a result counts for too little, the company misses real shifts in demand. If it counts for too much, the company chases noise. Setting that weight well for each kind of evidence is work that never finishes.
3. A record of why each decision was made
A portfolio has to inherit reasons as well as outcomes. That means knowing why each decision was made, what the evidence was at the time, and what happened next, including for the brands that were pruned back. A pruned brand still teaches the portfolio, but only if its reasons were kept.
06 A portfolio with infinite memory gets better with every brand
Treating every result as evidence is how we learn, and infinite memory is how that learning reaches the next brand. We call the capability we're building commercial intelligence. It runs research, product development, and brand design as one integrated process, guided at every step by how the market responds. It also connects every brand in the portfolio, so what one brand learns reaches all the others.
We believe a portfolio with infinite memory will make better decisions with every brand it launches. Findings travel further when each one carries the conditions that produced it, and each new brand begins with reasons as well as results. That depends on keeping the record honest, including for the brands that were pruned back.
Most consumer companies start every brand from zero. In a portfolio with infinite memory, each new brand starts from everything that came before it.