The question of the rate at which alliances fail never goes away. Nor does the wide-spread article of faith that strategic alliances inherently have a high failure rate, usually well above 50%. To take just three recent quotes taken more or less randomly from the Internet:
“The failure rate of strategic alliances is significant. Depending upon which source is referenced, the failure rate can be as high as 82% or as low as 50%.”1
“Despite the advantages of alliances for both parties the failure rate is remarkably high, around 50%.”2
“Depending on which surveys you believe, the failure rate [of strategic partnerships and alliances] is somewhere between 60-72%.”3
What is missing, however, is any reliable data backing these figures or any attempt to analyse failure rates by industry segment, alliance type, date and so on. What data that is used to back up the estimates of failure rates is often seriously outdated, based on tiny sample populations or taken from inherently unreliable, theoretically-skewed, papers written by academics. So, for example, some of the data relied upon to arrive at these estimates is decades old and some comes from surveys (we use the word loosely) of little more than a handful of alliances.
The question as why any competent manager would bother investing in any activity with a likely failure rate of 82% is left unasked. By comparison, the failure rate on investments in the venture capital industry is around 66%.
To test the hypothesis that alliances have a failure rate of between 80% and 50% we decided to examine our own data gathered from the analysis of hundred of alliances. The question we asked was: what percentage of respondents in our surveys of current and past partners of clients agreed or disagreed with the following statements:
1. The alliance has been a success; or,
2. The alliance has met the expectations we had at the outset; or,
3. The alliance is on track to meet its objectives.
So that we could be sure that the reasonably is reasonably current we used data only for surveys conducted in the last three years.
The result? The percentage of respondents in our database disagreeing with any of those three statements was 18% (n=959), the percentage agreeing was 65%, the remaining 17% neither agreed nor disagreed with the statement in question.
A couple of footnotes to these figure are in order. First, our clients tend, by definition, to be large organisations with an alliance management function and with an interest in getting their alliance right so that would probably put a small positive bias in the data.
Second, a lot of the alliances we analyse for clients are intensive research-driven alliances with small biotechnology companies and academic institutions. These alliance probably have a higher failure rate than, for example, distribution and manufacturing alliances in other sectors so that may put a negative bias in the data.
These two biases probably pretty much cancel each other out.
So, in conclusion, we think that a failure rate of around 18-20%, for large organisations with a dedicated alliance management function and trained alliance managers, feels about right. If we had to use these figures to generalize across the market generally we would estimate a failure rate of between 20% and 25%. In short, far below the generally accepted estimates of the typical failure rate of alliances.
Pingback: Taking another look at alliance failure rates | Silico Research
Pingback: Taking another look at alliance failure rates | Silico Research
Research from ASAP would verify this. While the average success/failure rate is around 50%, when you look more closely what you find is that average tells you nothing. Companies that are ad hoc in managing their alliances have failure rates in the 80% range. Companies that have dedicated alliance managers and a deliberate process for managing alliances have an 80% success rate. What does that tell you.
Norma, thank you for the comment. We discuss possible differentiating factors between the two datasets in the reply to Peter’s comment.
Interesting post Emmett, one question though: In the summary you mention “a failure rate of around 18-20%, for large organisations with a dedicated alliance management function” and “to generalize across the market generally we would estimate a failure rate of between 20% and 25%”. The first figure I can understand, the second feels extremely low. As Norma states also a failure rate for companies with ad-hoc alliance management of around 80%, more a number I would expect and this makes quite a difference. Can you shed a light on this difference?
Peter, thank you for your interest in this thought piece and for your question. Our starting point is one of the largest and most comprehensive surveys in this area that was conducted by Jeffrey Dyer, Prashant Kale and Harbir Singh at the MIT and written up in How To Make Strategic Alliances Work (MIT Sloan Management Review Summer 2001). The survey addressed the difference in the success rate of alliances in companies with a dedicated alliance management function and those without. The authors found that organisations with a dedicated alliance management function achieved a 25% higher long-term success rate with their alliances than those without such a function (n=1,572). We have some issues with how the authors defined success and failure but that doesn’t invalidate the inter-set comparison. We think that the 25% difference in success rates between organisations with and those without an alliance management function is reasonable for the following reasons:
1) A lot of the causes of alliance failure have nothing to do with the management of the alliance (market failure, regulatory failure, failure in research and development, change of management, change of strategy, cultural clashes and so on); and,
2) Having an alliance management function does not, in and of itself, improve an alliance. Plenty of problems in alliances are introduced by alliance managers themselves; for example, personal friction between managers, communication failure or a lack of trust.
Anyway, using the failure rate in our dataset of 20% as a proxy for the better managed group in Dyer, Kale and Singh’s survey, unmanaged surveys generally would have a failure rate of around 30-35%. The mid-point for a generalized benchmark failure rate would be around 25-30%.
I haven’t seen the data that Norma is quoting but the difference may well be to do with issues such as the timing of the survey, the sample group, the questions asked or the definition of ‘failure’ used in that survey. Our test is very simple: does the partner consider the alliance a failure or not. Dyer Kale and Singh, on the other hand, use as a benchmark the average rating for survey population across four dimensions including the extent to which the alliance enhanced the competitive position of the parent company and the extent to which the alliance enabled each company to learn critical skills from the alliance partner. Using the average rating for the survey population as the test for success or failure, means, that whatever the level of partner satisfaction, 50% of alliances in the survey will be deemed to be a ‘failure’. This is methodologically acceptable when the research is comparing two sub-sets of the data but it would be inappropriate to use this figure as the basis for a generalized benchmark failure rate.
Pingback: A new view on Alliance failure rates — Simoons & Company