Is it a bad thing that 80% of new products fail?

Posted by Ray Poynter, March 2013

---

In a recent LinkedIn discussion, one contributor suggested that 80% of new product launches fail. This sort of statistic occurs in marketing discussions on a regular basis, with varying definitions of failing and various values being quoted, sometimes as high as 95% sometimes as low as 75%. But I feel that these discussions are often addressing the wrong issue. When they discuss ‘Why is market research concept testing so bad?’
Whenever I hear stats like ‘80% of product launches fail’, I ask myself, what percentage of new launches should fail?

I think that we need to move the question away from percentages and ask ‘How many product launches, in absolute terms, could be successful?’ As people like Mark Earls (author of Herd) have shown, we are creatures of habit and we mostly copy behaviour. As an individual most of us are only going to fully adopt a handful of new products each year, and we are more likely to adopt a product if others do.

So, I would contend that within any specific market, there is a limit to the absolute number of new products that can be successful. In my 35 years in research the number, of new product launches has gone up and up, and I feel it has gone up faster than the number of new successful products has gone up. I suspect that the ratio of 5:1 could well be close to the current ratio between the number of products that organisations feel they have to launch each year and the number that could be successful.

If I am a product manager at a company, I may be obliged (by the C-suite and the organisation’s strategy) to launch a range of products every year. If a fast moving consumer brand felt it needed to launch ten new lines or variants every year to keep getting space on the shelf, to get press coverage, to motivate sales, and to keep the brand looking fresh, it is not going to suddenly launch no products if it feels they might not succeed – not launching would, of course, guarantee failure.

If I am a brand manager, I might have 100 concepts to evaluate, I might be expected to launch 10, and I know that parity is 2 (i.e. 20% success) – as long as market research helps me pick the right 10, i.e. the 10 that increases the chance of getting at least 2 successes, I will feel research has helped.

One of the key things we are seeing from modelling, and again from authors like Mark Earls, is that the role of luck is enormous. If 5 equally good and attractive products are launched into the same market sector, only one will succeed (normally), i.e. if 5 equally good products are launched into the same market, one of them will do much better than the rest. The mechanic of success will be social copying, supported by marketing and luck.

So, I am all in favour of better concept testing, and I think that market research should keep pushing for better and better options. But, I think that even if we had perfect testing, the majority of products would fail – where I am defining perfect testing as something that would identify every product that had a chance to be massively successful. However, if every product we predicted to be successful, was successful, nobody would ever launch a product with a 90% chance of success – which would elevate the elimination of Type I errors (false positives) at the cost of a massive number of Type II errors (false negatives).

When real innovators and successful marketers criticize market research, it is not usually because we have allowed too many new ideas to reach the market. It is usually that we have stopped too many products that would have been successful from being launched. Market research should balance Type I (false positives) with Type II (false negatives). The history of market research is that we have probably been too keen on eliminating Type I errors, which means we have helped prevent millions of potentially great products.

In my, humble, opinion the aim of any new concept testing, market predicting technique should be to be more predictive than other tools, not to reach some arbitrary level of accuracy.

However, I will make one forecast. In ten years the failure rate of new product launches will be higher than it is now! This is because I think the number of new products launched will grow faster than the number of new products that can be successful – which will increase the failure rate. The number that will be launched is a function of falling barriers to market, increased retail complexity, and the nature of modern marketing. The number that can be successful is limited to how many things people, like us, will change in any one year – are you going to change your coffee, AND your favourite jam, AND your favourite beer, AND your toothpaste. AND so on – and then stick with your new choices long enough to make them successful?