Von Hippel is onto something with the idea of need-solution pairs.

This is how I understand the idea:

Most management literature is focused on problem definition. (See: James G March for one of the best models of how problems drives search). Often, searches are sparked by a problem. But what if it isn’t always sparked by a problem?

Von Hippel argues that search isn’t always motivated by formalized problems. Sometimes people are just discovering alternatives based on needs, and then, they compare a new alternative to the status quo. If the alternative is predicted to be better than the status quo, then the alternative will be substituted. If not, it’s rejected.

This is a kind of a neat way to think. Let’s run it through a common meme.

Henry Ford probably never said that if he listened to what his customers wanted, he’d just make a faster horse. But that doesn’t keep me from repeating that time honoured meme. Think about the need that a horse satisfied back in the day. It was transportation. You need to get from here to there. What were your substitutes at the time? Your feet. Your bicycle. A streetcar. A boat. A train. When you compare the different solutions to the need, the idea of a mechanical horse — an automobile — makes a lot of sense. It’s better than a horse on a number of quality attributes. It’s faster. It’s cleaner in terms of solids. It’s cheaper. It took a long time for the horse (and people!) to disappear from the roads. The substitute ultimately won.

That’s a pretty neat way of thinking, isn’t it?

An Interesting Tool

What if it’s easier to get problem definitions wrong than it is to get a need definition wrong?

The existence of so many frameworks and formalized processes around problem definition might be a clue to just how hard it is to define what a real problem is. How often are solutions really problems? Consider, for example, somebody who prefers the use of C# for everything. From one perspective, C# is a solution to every problem that comes along. From another, the use of C# for everything is the problem. How many meetings take on circular features because participants are confusing problems for solutions, not quite agreeing on the scope of a problem, or not agreeing on whether there’s a point to seeking the root of a problem?

The existence of so much complexity around defining a problem could be a clue that there could be a more elegant way of thinking about it.

Needs may be easier to describe.

Consider the need to quench thirst. I’m having a hard time imagining an argument that rejects the idea that quenching thirst is a need. It seems quite simple. The solution space to quench thirst is pretty expansive. But, at the very least, the need can be described as a simple matrix. Market segments along the top, and intensity of thirst along the side. There are several substitutes for clean, cold, water from a creek. Tap water and well water come to mind. Bottled water. Carbonated water. Water with additives like corn syrup, cane sugar, and caffeine. Different segments. Different solutions. Same need.

Need-Segment Surfaces

Assume a list of needs. This list seems kind of trivial to generate, and you might want to try enumerating a few yourself.

Security. Thirst. Hunger. Shelter. Rest. Clothing. Certainty. Reproduction. Love. Comfort. Compassion. Empathy. Belonging. Friendship. Entertainment. Excitement. Status. Support. Connection. Contribution. Achievement. Growth. Joy. Mourning. Creativity. Discovery. Fairness. Hope. Beauty. Privacy. Autonomy. Purpose. Self-actualization.

Assume a list of market segments. Let’s use the Geoffrey Moore (1991) definition of a market segment here: A group of people who refer to each other when making a purchasing decision.

This feels very trivial to generate, and you might want to try enumerating a few yourself:

Pumpkin Farmers, Canadian Fly Fisherman, Bungee Jump Operators, Suburban female teenagers, Rust Developers, Males with household incomes greater than $120,000 CAD and reside in postal codes beginning with V2.

The set of all market segments very likely exceeds the number of primary needs. I can be absolutely certain that the attributes of segments can be chained to absurdity. Consider the segment divorced males with six children, who own their own home, with gross assets between $300,000 and $400,000 CAD, aged 33 to 35, who hold Australian and Canadian passports, who are working towards their pilots license, who buy bologna or other meat slurry composites weekly. One can chain attributes into absurdity. The vast majority of market segments are the empty set. I don’t know if needs chain as neatly.

For instance, it could be argued that a Coach bag fulfills several needs: the need for beauty, the need for excitement, the need for clothing and the need for status. I’m careful here to separate out any Job To Be Done construct from the need. Holding things that you are not using at the moment, but that you might need later, is a job that you hire the bag to do. Holding things doesn’t seem to be a need. Somebody may have purchased a Coach bag in part because they needed to feel excitement by a purchase. We might call that a momentary hedonic good. One could argue that a Coach bag is beautiful, and people sate their need for beauty by surrounding themselves with beautiful things. And, sometimes, carrying a Coach bag lets other people who know what a Coach bag is that you could afford a Coach bag.

A Coach bag may have several quality attributes that are intended to address several different needs, but I am skeptical that a quality attribute, unto itself, constitutes a need.

It’s quite possible to identify discrete, atomic, needs, but I’m not convinced that it’s possible to connect needs in an infinitely deep tree the same way that market segment can be defined as such. I remain open to be convinced.

As a result, the need-segment surface would likely be a rectangle rather than a square. I’d put the needs along the top, the columns, along the X-axis, and I’d enumerate the infinity of segments along the rows, along the Y-axis. I’d argue that needs, in the absence of a framework that enough people agree with, are nominal variables. There may be a few communities that would try to put those needs in some kind of order. And, I’d also argue that the segments could be nominal variables. Though, it’s often very useful to impose some kind of ordinal order on them in some way. (Careful! Trees in a random forest can ignite and explode in a clusterf of complexity!)

What does the surface look like?

Consider adding a third dimension to this rectangular sheet. Let’s call that z-axis the number of solutions. If there is an arbitrary order to the nominal variables, you’d expect to see a pretty arbitrary distribution on the surface.

If you were to re-order the variables and cut the fabric, you might see some pretty interesting patterns. For instance, I’d anticipate that some market segments would have more available solutions to quenching their thirst than others. Some market segments just have dirty water. For others, it’s either the communal well down the street, a coca-cola, or a local beer. And, for other still, they have over 120 SKU’s to choose from at their supermarket.

Surfaces and Learning

What is particularly useful about Von Hippel’s contribution is how much simpler and bubblier, it makes search. There’s plenty of evidence that suggests people aren’t that exhaustive in their search for alternatives. They’re satisficers. Bellman has the receipts. So it creates a very nice binary comparison to alternatives.

As Y, I have a I need, X, that is getting satisfied by A, given an alternative, B, is it so much better than A that the switching costs are worth it? If Yes, then substitute, If No, then continue with life.

It may also make estimating that much easier. What I really like is that resolves a tension I’ve had with Thiel for awhile.

I love startups. And often I’ll encounter ideas for which I simply do not have the imagination for which market segment it could possibly appeal to. In part, the lack of imagination has to do with an unawareness of real needs that may, or may not, exist. As I sit here in February 2022, I am aware of but a tiniest subset of segment-need pairs. I know what I know. I don’t know what I don’t know. I have a vague awareness of just how huge the need for Belonging among Japanese teenagers aged 9 to 16 is, but I really don’t have any conception of the need for privacy amongst landless rural Kansans.

Sometimes the answer to the observation “There’s nothing on the market for landless rural Kansans that helps them keep their privacy” is, “do they have a need?” and follow up, “why landless rural Kansans?”.

Thiel’s observation was that the hardest bit was getting from 0 to 1 (haha, get it, bit?). If, as I’ve been arguing, that most cells on the surface are empty, then it follows that the second most common set is 1 person. The hardest bit about gradient ascent is likely all the common local maximas in the fabric — the persistent threat of the harbinger customer. Expressed in a different way – maybe some people are not connected to one another along certain need pathways in such a way that information about an alternative can be transmitted from a niche segment to a mass segment.

What if it’s far better to execute a search based on the X-axis first, and then generate a substitute for a segment along the Y-axis, and then do an adjacency check that there is another rational cell. That there is another market to grow to.


Von Hippel’s onto something with need-solution pairs. It’s a neat tool. I used it to construct a model of all need-solution-segment arrays, just so I could get a higher view of it all. I suggested a way of thinking about structuring and searching that surface for a cell for which to compare a substitute to the status quo. It might be a faster way of learning, and a way of de-risking getting trapped in a local maxima.

It might be an edge.


March, J. G. (1994). Primer on decision making: How decisions happen. Simon and Schuster.

Moore, G. A., & McKenna, R. (1991). Crossing the chasm.

Thiel, P. A., & Masters, B. (2014). Zero to one: Notes on startups, or how to build the future. Currency.

von Hippel, E., & Von Krogh, G. (2016). Crossroads—Identifying viable “need–solution pairs”: Problem solving without problem formulation. Organization Science27(1), 207-221.

von Hippel, E., & Kaulartz, S. (2021). Next-generation consumer innovation search: Identifying early-stage need-solution pairs on the web. Research Policy50(8), 104056.