How much research does a go-to-market strategy need?

We once followed twelve people in Madrid who were about to buy something and hadn’t yet decided what. It could be a television or a pair of running shoes. For three weeks they kept a diary on their phones, we went into the shops they chose with each one of them, and we interviewed them once they’d paid. The study was for Amazon, and the pattern that kept coming back was easy to describe and expensive to ignore. Many people needed to see the product in a shop first, to compare, touch it or ask someone. Then many bought it online at the best price they could find, while others paid in a shop they trusted.

If you’re planning to sell in Spain only through a website, that pattern can change your plan more than an estimate of market size would. That’s what go-to-market research is for: finding the assumptions in a launch plan that are wrong while there’s still time to change them.

So how much of it do you need? It depends on how expensive each decision is to get wrong and how easily you can undo it. And a study only earns its cost if you’re unsure enough for the answer to change what you do.

Why this is a challenge

Launch teams tend to fall into one of two habits. Some research everything, because a launch feels like a bet with the whole company watching. Others research almost nothing, because the plan worked at home and the date is already in the diary. Both camps tend to quote the same statistic, the claim that 80% or even 95% of new products fail.

That figure is hard to pin down. The 95% is often credited to Clayton Christensen, but we couldn’t find a study behind it, and a 2013 paper in the Journal of Product Innovation Management called the “80% or higher” failure rate a persistent myth: the studies it reviewed, going back to 1977, put the failure rate of new products at 40% or less.¹ A 2012 meta-analysis in the same journal, by contrast, opened by saying that success rates are generally below 25%.² Part of the gap is what gets counted (products that reached the market, or every idea that entered development) and how each study defines failure.

The careful numbers are still sobering. Researchers who tracked 83,719 new products (SKUs) launched in 31 US consumer goods categories between 2002 and 2009 found that a quarter of them were no longer being bought a year later, and about 40% after two years.³ Failure is common, but it isn’t destiny. The more useful question is which of your decisions are worth paying to be more certain about.

Five decisions, five kinds of evidence

A go-to-market strategy is usually five decisions dressed up as one plan. Each can fail in its own way, and each needs a different kind of evidence.

• Who you serve first. You need to know which groups exist, what drives each one and how big they are. That usually means interviews to find the groups, a survey to size them and a segmentation the sales team can actually use (we wrote about why so many segmentations die in a slide deck).
• Where you go. Market size is the part most plans already have. Walmart left Germany in 2006 after eight years there. Matthias Kaelberer puts the failure down to a strategy too rigid for local conditions, more than to the rigidity of the German market itself, and Susan Christopherson shows how a retail model shaped by US market rules ran into German institutions.⁴⁵ Culture, rules, habits and the relationships local players already have rarely show up in a market sizing.
• What you charge. Pricing research gets closer to real behaviour when people’s answers have consequences, as they do in a shop. When Klaus Miller and colleagues compared four ways of measuring willingness to pay against real purchases, the two methods with real stakes passed their tests, while hypothetical answers showed bias, even if they sometimes still pointed to the right price.⁶
• How people buy. Online, in shops, through partners, or the mix the Madrid shoppers showed us. Here, watching people buy, or asking them while it’s happening, tends to beat asking them weeks later, because people tidy up their memory of a purchase once it’s over.
• What you say. Messages are cheap to test and easy to change, but you still need the words your customers use. When a smartphone brand doubled its online conversation in Spain in a year, a year of social listening (published as a composite case, with names removed) showed that the biggest peaks came from campaigns with famous faces on X, and that the conversation about the product picked up when launches repeated a single, concrete camera claim.

How much is enough: one-way and two-way doors

Once the decisions are on the table, sort them by how hard they are to undo. Jeff Bezos put this neatly in his 2015 letter to Amazon’s shareholders: some decisions are one-way doors, consequential and almost impossible to reverse, and they deserve slow, careful work.⁷ Most decisions are two-way doors you can walk back through.

In a launch, the two-way doors include the headline, the offer, a price test on a small share of online traffic and most digital touchpoints. For these, testing in the market is usually cheaper than researching in advance, as long as you have the traffic. Ron Kohavi and Stefan Thomke put the bar at a few thousand daily active users, which a brand-new launch may not have yet. Testing is humbling, too: in their account in Harvard Business Review, only about 10% to 20% of experiments at Google and Bing produced positive results.⁸ If most ideas at companies with that much data don’t move the numbers they were meant to move, it’s worth letting real behaviour decide.

The one-way doors include entering a country, signing a distribution deal, choosing the positioning you’ll build the brand on and committing a factory to a product. These are the decisions where research before the decision pays, because the market won’t give you a cheap second try.

There’s a simple test for whether a study is worth buying. Before you commission it, write down what you’d do if the answer came back one way, and what you’d do if it came back the other. If both lists are the same, don’t buy the study. Decision analysts call this the value of information, and Ronald Howard’s 1966 paper on information value theory is a classic statement of it: information is worth what it can change in a decision, weighed by what’s at stake.⁹

Howard also showed that resolving several uncertainties together can be worth more or less than resolving each one separately.⁹ Launch decisions interact in a similar way: the segment you choose changes the price you can charge, and the channel changes what you need to say. Five studies commissioned separately can leave nobody looking at where those decisions meet.

“Won’t research slow the launch down?”

It’s a fair worry, and the best-known studies suggest it’s often misplaced. Robert Cooper and Elko Kleinschmidt spent years studying how companies develop new products. In a study of 203 new products, they found that the early, predevelopment activities were among the most critical steps in the whole process.¹⁰ In a later study of 103 projects in the chemical industry, solid up-front homework was the second-strongest driver of projects that finished on time.¹¹ A 2001 meta-analysis in the Journal of Marketing Research reached a similar conclusion, with proficiency in predevelopment tasks among the factors with the biggest effect on new product success.¹²

These studies compare how managers described past projects, so they show an association, not proof that the homework caused the success. The 2012 meta-analysis mentioned earlier updated the 2001 one with studies from 1999 to 2011, and found weaker effects that shrank over time.² Still, in the one study that measured speed, the projects that did their homework weren’t the slow ones. Skipping it often doesn’t save the time it promises. It can move the questions to the moment when answering them means a relaunch.

This is what we do

If you’re asking how much research your launch needs, start with the five decisions and ask two questions about each: what happens if you get it wrong, and could you undo it next month? The expensive, irreversible decisions deserve research before anyone commits. The cheap, reversible ones need a test plan. Some decisions need neither, because no answer would change what you’d do. When you bring us a launch, that’s where our first conversation starts too: with the decision in front of you, before any method.

Sometimes the right plan is a pilot. The care app we researched and co-designed with a Spanish startup was built on 100 interviews and three co-creation workshops, and it’s now in pilot with two councils, two care homes and five families. Sometimes it’s a year of listening, as in the smartphone case. And when the decision is a one-way door into a new market, it often pays to use all three lenses at once (interviews, surveys and social listening), which is what our H3™ method was built for. If your launch doesn’t need a study, you’ll hear that from us first.

Spend most of your research budget on the doors that only open one way.