«Abstract We conducted a qualitative study of Nokia to understand its rapid downfall over the 2005–2010 period from its position as a world-dominant ...»
We started with open coding (Strauss and Corbin, 1998), which included reading the interview transcripts and marking codes to describe the content of the interviews. We coded nearly 2,000 segments of data. The initial codes covered various topics, such as ‘‘MMs criticizing TMs for being ‘lost’,’’ but we also included emotion-related themes such as ‘‘perceived lack of emotionally appealing communication.’’ We further categorized such first-order codes into more-abstract theoretical dimensions. We also wrote numerous memos during the process to develop theoretical insights.
As we iterated between coding and data, our theorizing evolved in four main steps: (1) Nokia’s failure to innovate caused by TMs’ inaccurate understanding Vuori and Huy 11 of capabilities; (2) Nokia’s failure to adjust during the innovation process caused by inauthentic TM–MM interactions; (3) the interim outcome, temporal myopia—excessive focus on short-term innovation even though long-term activities might lead to more beneficial outcomes (Levinthal and March, 1993)—caused by TMs’ externally focused fear and MMs’ internally focused fear and low external fear; and (4) Nokia’s attention structures that in part elicited such fears, which created decoupling interactions between TMs and MMs, causing temporal myopia.
We refined our coding procedures according to our evolving understanding (Strauss and Corbin, 1998). To code emotions more systematically, we used the appraisal theories of emotions (Lazarus, 1991; Ellsworth and Scherer, 2003;
see also Huy, 2011; Huy, Corley, and Kraatz, 2014), which indicate that emotions result from specific appraisal processes and each basic emotion is associated with a prototypical appraisal pattern. For example, fear is elicited when people perceive potential harm in their relationship with their environment and consider their coping potential as low or uncertain. We thus inferred fear when people described their concern that undesirable things might happen to them— that is, uncertain future harm—and did not feel they had full control over the threat or the ability to eliminate it. In addition, the informants sometimes explicitly described various emotions they had felt, by using phrases such as ‘‘I was afraid that...,’’ or they recalled physiological reactions such as shaking or a racing heartbeat and their context, which allowed us to infer specific emotions.
When multiple informants described having experienced a similar emotion and/ or reported that colleagues experienced the same emotional reaction associated with a similar appraisal, we inferred that shared emotions were present.
Labeling emotions felt jointly but individually as ‘‘shared’’ is consistent with the way ‘‘shared’’ is used in research on shared mental models that uses a compositional view (see Kozlowski and Klein, 2000; Mathieu et al., 2000). Initially, we recognized a variety of emotions from the informants’ accounts, but we ultimately focused on fear because it was the emotion most commonly described, and it explained most plausibly the key dynamics that we describe in our inductive model.
Beyond coding, we also used a practice similar to constant comparisons in grounded theory (Strauss and Corbin, 1998) to identify differences and similarities among various data segments. For example, by comparing an MM’s fear accounts with those of other MMs, we could see that they had experienced similar emotions associated with similar appraisals, thus suggesting shared emotions. In addition, when MMs’ expressed targets of fear differed from those of TMs, we ultimately inferred the difference between internally focused and externally focused fear and started to analyze their antecedents and consequences. This iterative process resulted in the data structure presented in figure 1, which shows how the theoretical concepts we developed are grounded in the empirical data.
We used a factual timeline of events, theoretical logic, and the informants’ descriptions, which contained various events at different points in time and relationships between the events, to identify the temporal dynamic of our process model, i.e., reflecting how some factors influence subsequent factors over time. For example, we created several causal-loop diagrams to map the links between concepts that were abstracted from the data. Likewise, we 12 Administrative Science Quarterly XX (2015) Figure 1. Code-aggregation diagram.
matched the informants’ fear-related statements with their descriptions of TM– MM interactions through axial coding (Strauss and Corbin, 1998).
Checking the Model’s Validity We conducted a three-step member-check process to increase the validity of our inductive model (Shah and Corley, 2006). First, we presented the key elements of the model individually to 23 informants who were spread across different groups and functions. As shown in table 2, 21 of these informants, including TMs and software VPs and directors, concurred that our model accurately Vuori and Huy 13 described what had occurred at Nokia.2 Second, we sent a four-page summary of our key findings to 331 TMs and MMs who worked for Nokia during our study period, with an invitation to send us feedback, ‘‘especially if you disagree with the findings.’’ We received 37 written responses, all but three confirming our key findings. Several MMs also shared additional examples of how they had felt fear or how fear had prevented them from reporting negative information to TMs, even when they knew it would have been the right thing to do. Third, we conducted 27 follow-up phone calls and face-to-face conversations during the revision process to double-check that our informants had indeed meant that they had felt fear as an emotion (as opposed to using ‘‘fear’’ as shorthand for a cognitive evaluation of a situation) when describing their own experiences and their interpretations of others’ experiences. Twenty-two (81 percent) of them confirmed that they had indeed meant the emotion fear and that this emotion had influenced their behavior. Hence Nokia informants overwhelmingly agreed with the key elements of the inductive model that we present below, in particular the role of externally focused and internally focused fear.
FINDINGS Various types of fear experienced by Nokia’s TMs and MMs caused a decoupling interaction pattern between the two groups that harmed Nokia’s innovation process. Two factors influenced TMs’ and MMs’ fears: Nokia’s structural distribution of attention produced differences in appraisals and shared fears among TM and MM groups, and TMs’ past aggressive behaviors—and widely shared stories about them—triggered intuitive fear reactions in MMs because they had a lower position in the organizational hierarchy than TMs.
Structural, Attention-related Antecedents of Shared Fears Structural distribution of attention. Nokia’s organizational structure included the division of labor between TMs and MMs such that the former had a focus on the external environment and the latter on the implementation of TMs’ directives (March and Simon, 1958; Joseph and Ocasio, 2012). TMs focused on the external environment through direct interaction with customers and competitors and also through staff channels (corporate market intelligence). A TM told us that he had plenty of interactions with customers, especially large network operators such as AT&T and France Telecom, and the CEO publicly admitted that he was ‘‘paranoid about all the competition’’ (MarketWatch, 2007).3 Nokia had clearly defined processes for managing We showed a late version of the figures in this paper and explained them, after which the informants were asked whether they felt the model was a reasonable description of what had happened at Nokia, in light of their own experiences. Twenty-one participants agreed with the model, and two expressed doubts: one who had a significant role in radio-technology development before 2007 and did not witness decoupling interactions between TMs and MMs in his or her area of activity, and one who claimed that Nokia did not have problems in software quality.
When we refer to the data from our private interviews, we have anonymized the individuals such that top managers (CEO and three executive vice presidents) are marked as TM, senior vice presidents and vice presidents as UMM#1.#N (upper middle managers), and directors and managers as MM#1.#N (middle managers). We also indicate the primary focus area of the individuals, such as ‘‘software,’’ with the quotes when this has theoretical relevance. When we use data from public sources, we refer to the actual titles of the people.
14 Administrative Science Quarterly XX (2015) emerging strategic issues (e.g., Kajanto et al., 2004) and strategic agility (Doz and Kosonen, 2008), according to the CEO, even using ‘‘cultural anthropologists... to find out what’s important in people’s lives because everything changes so quickly’’ (Korn-Ferry Institute, 2010). Nokia’s active sensing of the market was manifest in its response to the iPhone threat about one year before its launch, which it knew about despite Apple’s famed secrecy. In other words, cognitive inertia (Fransman, 1994; Tripsas and Gavetti, 2000) cannot be seen as a plausible or sufficient explanation for Nokia’s failure. As members in senior
strategic positions at Nokia revealed:
We knew [the iPhone] was coming out about a year in advance. We had pretty good specifications for it.... The first thing [a MM] flagged up [in a market review] was that we didn’t have touch screens or touch-screen development.... That message went right up to the top of the organization, and was well received too. [The CEO] forwarded [the] email to his subordinates. Our market analysis showed that our biggest competitive weakness was the lack of touch-screen products; he agreed, and wrote ‘‘Please take action on this.’’ (MM#1, business) Confidential prospective documents that were shown to us confirmed that the TMs did indeed recognize this issue, gave it a high priority, and included it in their requests to MMs. UMM#1, who was working close to the CEO, likewise
described how the CEO put clear pressure on MMs so that they would implement what he thought would be essential:
One of the first things [the CEO] brought up was the touch screen.... He felt it was the next big thing... He brought it up with the executive group every way he could.
And he spoke directly with technical middle management.... In every single executive-group meeting, they went over our outlook with the touch screen. And this was right after he was made CEO [a year before iPhone launch]... I recall many such cases vividly, where he brought up the right concern [what Nokia should do], he discussed it directly with the technical middle and upper management, he went straight to the topic, put pressure on people, put it up in all possible goals [for MMs], followed up on it in every single meeting.
A TM further elaborated in a private interview with us that he saw it as TMs’ task to translate market expectations into clear goals for the organization: ‘‘The tight schedule and challenges come from the markets. They put time pressure on you, and you need to pass it on to the organization.’’ Given that Nokia failed to develop an innovative offering to match the iPhone, despite TMs’ intentions and directives, we need to better understand how the implementation process unfolded within Nokia’s structural and historical context.
MMs’ focus of attention was predominantly on the implementation of TMs’ orders rather than the external environment and was split between two key groups: product units (PUs) in charge of defining specific products and software units in charge of developing Nokia’s two OSs, Symbian and MeeGo, and modifying them to meet the needs of specific phone models based on requests from the PUs. Although the primary communication channel flowed from TMs to PUs to software units, the software units also had direct interactions with TMs, both formally and informally. This suggests that a purely structural explanation—lack of opportunities for interaction (cf. Henderson and Clark, 1990; Ocasio, 1997)—does not explain the assessment gap between TMs and Vuori and Huy 15 software MMs that we describe below. The PUs’ attention was focused on TMs’ requests and narrow market segments corresponding to the specific products they were developing. Nokia had five PUs, each responsible for a specific market segment such as music phones. Within each unit, there were several programs, such that Nokia had ‘‘40 to 50 product programs per year’’ (UMM#2), and program leaders were tasked with implementing their respective programs. Each PU focused on delivering phones as directed by TMs. The key aspects that TMs emphasized for PUs were product segmentation, cost, and schedule. The software units’ attention was focused on satisfying both
TMs’ and PUs’ requests. As our informants related: