Initiating heuristic portfolios
Strategy research often infers that firms learn routines from their accumulated experience with organizational processes (Kale et al., 2002; Nelson and Winter, 1982; Zollo et al., 2002). In contrast, we find that firms learn portfolios of heuristics, so we confirm the conjectures of Eisenhardt and Sull (2001) with systematic, longitudinal data. Specifically, firms learn particular types of heuristics that focus on successfully capturing country entry opportunities. That is, consistent with the distinctions of Cohen et al. (1996), firms learn a common rule structure for a range of similar country entry problems (heuristics), but do not learn extensive details and precise steps to be applied consistently in every country entry (routines)9. Unexpectedly, firms initiate their heuristic portfolios with particular heuristics—selection and procedural.
We define selection heuristics as deliberate rules of thumb for guiding which sets of product or market opportunities to pursue (and which to ignore). Similar to Eisenhardt and Sull (2001), we coded heuristics as ‘selection’ if they specified particular countries or geographic regions to enter, products to sell, or customer types to address. We define procedural heuristics as deliberate rules of thumb for guiding the execution of a selected opportunity. We coded heuristics as ‘procedural’ if they specified country entry mode or an approach to functions like sales, design, staffing, or pricing policy. All six firms developed selection and procedural heuristics in their first country entry (Table 2).
A good example is U-Semi, a U.S.-based firm that creates semiconductor solutions for GPS-enabled mobile devices. Prior to entering their first country (China), executives planned to use Taiwan for the design layout of the firm's chips and China for manufacturing. As one executive stated, ‘Initially, we thought the product should be designed in Taiwan and manufactured in China because of the cheap labor.’ But after entering China, the executive team realized China was important for both manufacturing and design. The CEO noted, ‘We gradually discovered that even if they have R&D, most of the big Taiwanese companies also have their R&D centers in China. So actually what we found out is that our design activity should be in China and the manufacturing should also be in China.’ This learning prompted the executive team to create a procedural heuristic: use China for both design and manufacturing (Table 2). Later a team member confirmed, ‘Our design activity and manufacturing are both in China.’ Firm members used this heuristic to guide the execution of subsequent country entries.
Executives also learned other selection and procedural heuristics in China. When the firm first entered China, executives believed they should sell just their semiconductor products. But after several months of no sales, they realized that potential Chinese customers had weak engineering skills that prevented them from exploiting the cutting-edge technology of U-Semi's chips. The CEO said, ‘We found that in China, they (customers) didn't have much of a design process. You had to finish the whole design for them before they could use it.’ Therefore, the executive team decided that instead of selling only chips, they would sell turnkey solutions (selection heuristic). One member noted: 'Our experience clearly showed us that we needed to provide the total solution. We are not going into any country just having a chip and saying 'Here's our datasheet. Good luck!”
Another example is S-Security, a Singapore-based security software firm. During S-Security's pre-internationalization experience in Singapore, management was successful in targeting government agencies and banks and focusing on a technology ‘features’ sales approach directed at IT managers. The executive team decided to focus on the same customer (government agencies and banks) and use the same sales approach (features ‘sell’ to IT managers) when entering their first country—Hong Kong. After the firm's entry however, the CEO realized selling to IT managers was not working. Since new financial guidelines in several Asian countries required senior executives to understand information risks, many Hong Kong companies had shifted responsibility for information security from IT to audit. As the CEO said, ‘We learned from experience who is our buyer, who makes the decision—it's auditors instead of IT. In more and more of the organizations, IT security is out of IT. The budget is not in IT—it is part of risk management. So we changed according to the market.’ Thus, the executive team learned a selection heuristic: target the audit group within customer organizations to get sales.
Similarly, when trying to close sales, the CEO discovered the ‘features’ sales approach that had worked in Singapore yielded few sales in Hong Kong. Instead, he observed that Hong Kong firms preferred a ‘consultative’ approach with customized solutions to match their unique needs, noting, 'It's consultative selling, meaning that it's not 'Hey, this is a very good technology.” This mistake prompted the executive team to start using a consultative sales approach (procedural heuristic). The firm used these heuristics to guide subsequent country entries and, in doing so, avoided repeating the selection and procedural errors of Hong Kong. As the CEO remarked, ‘Malaysia is one year later than Hong Kong, so we didn't make the same mistakes.’
Overall, we find that firms explicitly learn heuristics as they begin their process experience. But since heuristics are often seen as dysfunctional (Busenitz and Barney, 1997; Carlson and Shu, 2007; Kahneman and Tversky, 1973; Holcomb et al., 2009), it is important to clarify why firms learn heuristics. One reason may be that heuristics are useful when time is short, information is limited, and the situation is novel (Newell and Simon, 1972). Heuristics speed action by requiring less information and simplifying cognitive processes. Consistent with this argument, we observed, for example, that when a third party in Australia contacted U-Analytics to encourage entry, executives used their selection heuristic of ‘restrict internationalization to English-speaking markets’ to respond swiftly even though they knew little about Australia. As a VP said, ‘We'd already decided that we're going to attack the English-speaking markets and so it (Australia) was just too good an opportunity to miss.’
A subtler reason firms may learn heuristics is that they are often surprisingly accurate. Research on ‘fast and frugal’ heuristics (Gigerenzer and Brighton, 2009; Goldstein and Gigerenzer, 2009) finds that simple heuristics can outperform analytically complicated and information-intensive approaches even when information and time are available. Heuristics are often accurate because they exploit information about context that individuals have, an attribute that laboratory-based research on ‘heuristics and biases’ usually lacks. Individuals seem to learn simple heuristics that fit with their understanding of the context and correlate with other information that also influences outcomes (Gigerenzer, 2008; Wilson and Schooler, 1991). The ‘English-speaking markets’ heuristic at U-Analytics, for example, takes advantage of a founder's U.K. rearing and familiarity with the British Commonwealth (most English-speaking countries) culture and so proxies for other useful information. This heuristic provided helpful guidance even though the founder could not anticipate what specific information would be useful. By contrast, analytically complex, information-intensive approaches may underperform because they ‘overfit’ experience, ineffectively weight diverse information, and do not exploit actors' knowledge of the situation (Goldstein and Gigerenzer, 2009). For example, investors using a single heuristic (invest equally in all asset classes) outperformed investment policies that relied on substantially more information, analysis, and computation (DeMiguel et al., 2009), while individuals using a single heuristic (take the midpoint between the two most distant crime scenes) solved serial crimes more quickly and accurately than a complex computational approach (Taylor et al., 2009).
Firms may also learn heuristics because they are easy to remember and improve. As cognitive science research indicates (Baddeley and Hitch, 1974; Cowan, 2001), knowledge retention is enhanced when lessons are simple because the significant capacity limits of short-term memory importantly restrict the amount of information that can be encoded in long-term memory. Without encoding, lessons are forgotten (Anderson, 2000; Craik and Lockhart, 1972). Simplicity also makes improving heuristics easier because it is easier to process feedback when actions are transparent and understood. Simplicity is particularly advantageous for organizational learning because individuals are better able to convey simple lessons and recipients are better able to remember them. As one cofounder described his firm's heuristics, ‘It's not really coded anywhere. It's been diffused in the company, so it gets into everybody's head.’
Unexpectedly, firms initially learn selection and procedural heuristics. This finding is important because prior literature does not anticipate that particular heuristics are learned first. It is also important because it emphasizes that heuristics relate to specific problem-solving contexts. So while prior research identifies universal heuristics such as anchoring (Tversky and Kahneman, 1974) and take-the-best (Gigerenzer and Goldstein, 1999) that individuals use to solve binary choice problems with correct answers (e.g., Does Cologne have a bigger population than Bonn?), we find selection and procedural heuristics that firms uniquely learn to help solve the common problem addressed by organizational processes—i.e., successful opportunity capture in an abundant flow of related yet heterogeneous opportunities. For example, Cisco's acquisition process attempts to make high-performing acquisitions from a large pool of heterogeneous, potential acquisitions. Similarly, our firms attempt to make effective country entries from a large pool of heterogeneous country entry opportunities. Selection heuristics help firms cope with this abundance by constraining the range of opportunities. For example, S-Security's selection heuristic of ‘restrict internationalization to Asia’ restricts the choice of countries. Similarly, procedural heuristics constrain how country entries should be made. While firms could flexibly improvise all facets of every entry, this would be slow and prone to mistakes. Thus, procedural heuristics speed entry, conserve attention, and improve reliability of opportunity capture by giving coherent guidance about entry without specifying precise details. In summary, we propose:
Proposition 1: When firms engage in repeated process experience, they initially learn selection and procedural heuristics for capturing opportunities.
Adding temporal and priority heuristics
Knowledge research identifies the importance of declarative and procedural knowledge categories (Grant, 1996; Moorman and Miner, 1998; Reagans et al., 2005). We also find that firms learn these knowledge types in their selection and procedural heuristics, respectively. But surprisingly, our firms also learn priority and temporal heuristics that focus on different knowledge. There are two unexpected findings. One is an expanded conception of temporal heuristics to include sequence and pace, not just rhythm. The other is developmental order—i.e., firms learn temporal and priority heuristics after they begin to learn selection and procedural heuristics.
We define temporal heuristics as deliberate rules of thumb for opportunity capture that relate to time. We coded heuristics as ‘temporal’ when they relate to time, such as sequence (e.g., order of approaching customer types), pace (e.g., complete one entry before beginning the next), and rhythm (e.g., number of entries per year). We define priority heuristics as deliberate rules of thumb that rank opportunities. We coded heuristics as ‘priority’ if they rank some acceptable opportunities as more important than others (e.g., preference ranking of some customers among all acceptable customers). All six firms began to learn temporal and priority heuristics after they started learning selection and procedural ones. Thus, heuristics are unexpectedly learned in a specific development order (Table 2).
To illustrate, U-Semi began to learn temporal and priority heuristics after selection and procedural ones. The executive team learned during their second country entry (Taiwan) that the firm should emphasize ‘tier-one’ countries (e.g., Japan, Germany, U.S.) over other countries with original device manufacturers (ODMs) and original equipment manufacturers (OEMs) (priority heuristic) because tier-one countries have the largest domestic markets for the mobile applications targeted by U-Semi's products. But, the team also realized entering tier-one countries immediately would be challenging because the firm lacked credibility. The sales vice president noted, ‘We started right off and tried to talk to Dell (U.S. tier-one). They wouldn't give us the time of day. They won't take you seriously until they see a validated platform.’ Based on this knowledge, executives decided the firm should (1) sell in ‘tier-three’ countries like Taiwan and then (2) use those reference accounts to gain customers in ‘tier-two’ countries like Korea. After gaining ‘tier-two’ customers, the firm should then (3) use those accounts to enter ‘tier-one’ countries like Japan (temporal heuristic: move from tier-three to tier-two to tier-one countries). The vice president summarized, ‘Our marketing strategy has been trying to get the credible players from the tier-three, then tier-two countries—the big fish in the small pond like Hyundai in Korea and BenQ in Taiwan. If they adopt your platform and you ship in mass production, then you leverage that to get into the tier-ones in Japan, Germany, and North America.’ The firm used this heuristic to guide subsequent country entries.
U-Semi also learned other temporal heuristics. The executive team learned in their fourth entry (Japan) to sequence new product introductions by country (i.e., first Taiwan, then Korea, then China), and to sequence their selling efforts by industry sector within a country (e.g., first auto, then PDA, then mobile handset). The former sequence was based on decreasing engineering skill. The latter was based on increasing market rivalry. As one executive explained, ‘We learned that it is much tougher to get the handset providers in the short term. So we're starting with auto.’
A second illustration is S-Enterprise. The executive team began to learn temporal and priority heuristics after they began to learn selection and procedural heuristics in Taiwan (first country). After their entry into Taiwan, they decided that Japan should be their highest priority Asian country (priority heuristic) because they saw in their Taiwanese experience that many customer trends begin in Japan. But as they probed for early customers in Japan, executive team members realized Japanese customers were not interested in buying unless the firm had U.S. references. A founder stated:
‘Japanese customers did not respect us when we said that we were a Singaporean company. I think it is a prejudice on their part. Japan looks to the U.S. as being at the forefront of technology, but not to the rest of Asia. They see themselves foremost in Asia. It can't possibly be that this small company from Southeast Asia has technology that we don't have.’
To address what executives termed the ‘pecking order of nations,’ they decided to sequence their country entries: (1) first enter the U.S., (2) use U.S. customers as references to enter Japan, and then (3) use Japanese customers as references for entry into the rest of Asia (temporal heuristic). A founder noted the counterintuitive sequence of expanding outside Asia to expand further in Asia: ‘If you go to Japan, you first have to have success in the U.S. If you have success in Singapore (headquarters country) and you go to Japan, you may not be able to sell. So it means that you have to have success in the U.S., then you go to Japan, and then from Japan you can go elsewhere in Asia.’ The value of this temporal heuristic was later confirmed as an executive noted: ‘Their (Japanese customers) faces changed when we said we were a U.S. company and started giving out the Inc. business card. They were much more receptive then.’
S-Enterprise executives also added priority heuristics. The founders noticed business activity around a new standard, Rosetta Net, in their home country of Singapore and their first country (Taiwan). But, they did not understand the implications of this standard when they entered their second and third countries (U.S. and Japan) where Rosetta Net was less actively promoted. They then realized that, within Asian countries (selection heuristic), it was advantageous to ‘ride on the coattails of this new international standard’ because selling in countries with Rosetta Net activity took advantage of S-Enterprise's understanding of the standard from Singapore and Taiwan and signaled the likely existence of a sophisticated customer base in the country that would buy S-Enterprise's leading-edge products. So, they added a priority heuristic such that they continued to enter Asian countries (selection heuristic) but with a preference for Asian countries that actively promoted Rosetta Net (priority heuristic). One team member stated, ‘Rosetta Net is the entry point for us. We learned that when you embrace a common standard, your processes are similar, country to country.’ Thus, when deciding their fourth country entry, executives quickly converged on Malaysia because Rosetta Net had recently opened an office there. The CEO explained, ‘We saw the opportunity with this new standard being taught to Malaysia. That's why we started moving in.’ Another executive member concurred, ‘Malaysia happened to have a strong enough consumer base that they embraced the (Rosetta Net) standard. So it was a natural next step for us.’
Overall, we find that firms explicitly learn temporal and priority heuristics. This further emphasizes the key point that heuristics relate to specific problem-solving contexts. Similar to the prior heuristics, these heuristics relate to the common problem of organizational processes—i.e., successful opportunity capture in an abundant flow of heterogeneous opportunities. Heterogeneity suggests that some opportunities may be more attractive (e.g., higher growth). Priority heuristics guide executives to avoid lower-value (albeit acceptable) opportunities when higher-value ones exist. Heterogeneity also suggests opportunities may have features (e.g., customer engineering before sales call) that make temporal heuristics helpful for sequencing activities. Also, since opportunity capture often requires internal coordination of limited resources, heuristics that set a rhythm or pace can be especially advantageous (Brown and Eisenhardt, 1997; Vermeulen and Barkema, 2002). For example, as F-Supplysoft's CEO explained, ‘We have a timetable. We take one continent at a time. So, if the U.S. (current North American entry) goes as planned, then we start in China. If not, we delay China.’
More intriguing, our data also indicate a developmental order—firms begin to learn temporal and priority heuristics after they begin to learn procedural and selection heuristics. This is significant because it indicates an unexpected phased development of heuristics. One reason for this development is that temporal and priority heuristics involve relationships among opportunities and so require more experience to learn. Firm members often need to learn about single opportunities before they can relate those opportunities to one another by ranking or sequencing them. In contrast, selection and procedural heuristics relate to single opportunities and so require less experience to learn (Table 3).
Table 3. Characteristics of opportunity-capture heuristics
A less obvious reason may be heuristics that involve relationships among opportunities require not only more experience, but also more cognitive sophistication to learn. Individuals must simultaneously keep in mind information about several experiences while making cognitive links among them. For example, S-Enterprise executives solidified their heuristic to focus on Asian countries (selection heuristic) early on. But they did not learn their preference for selling in Rosetta Net countries (priority heuristic) until after they had experienced selling in both Rosetta Net and non-Rosetta Net countries and had made cognitive links among different entry experiences to determine benefits of the Rosetta Net standard for country entry.
A related reason for this development order is that temporal heuristics, in particular, are likely to be learned later because explicit knowledge about time is often abstracted from experiences that happen first (Boltz, Kupperman, and Dunne, 1998). That is, temporal knowledge often builds on nontemporal knowledge (Zakay and Block, 1998) and links together implications of experiences that occur in different time frames. For example, U-Semi executives had to first learn through experiences in several countries that they faced many more commercial rivals in some customer industries (e.g., mobile handsets) than others (e.g., autos). It was only after they had learned these lessons and developed sufficient cognitive understanding of their implications that they were able to learn a temporal heuristic for sequencing customer types (i.e., first auto, then PDA, then mobile handset). In general, our findings are consistent with cognitive science research that shows temporal concepts are learned after nontemporal ones (Fraisse, 1982; Hambrick and Engle, 2002).
Overall, this developmental order is consistent with cognitive science research on experts. This work finds that novices become experts through experience (Chase and Simon, 1973; Feltovich, Prietula, and Ericsson, 2006). Moreover, experts such as chess masters are able to hold in mind multiple actions simultaneously while novices cannot (de Groot, 1978; Ericsson and Kintsch, 1995). Experts also think in terms of relationships among features like priorities and sequences and integrate past and future time frames while novices focus on isolated features in the present (Ericsson, Patel, and Kintsch, 2000; Friedman, 2000; North et al, 2009). For example, expert firefighters interpret a fire scene by what preceded and what events are likely to follow while novices focus on immediate features like color and intensity (Klein, 1998). Thus, later learning of temporal and priority heuristics is consistent with transition from ‘novice’ to ‘expert.’ In sum, we propose:
Proposition 2: Temporal and priority heuristics are learned after selection and procedural heuristics.
Engaging in simplification cycling
Much research argues that routines improve task efficiency by enhancing speed and reliability (Davis et al., 2009; Helfat and Peteraf, 2003; Nelson and Winter, 1982). As firms gain experience with routines, they elaborate them to accommodate added lessons and so develop an increasingly reliable and complete set of action steps (Eisenhardt and Tabrizi, 1995; Kale and Singh, 2007; Szulanski and Jensen, 2006). In contrast, we observe both elaboration and simplification. Specifically, our data show that as firms manage the content of what they explicitly learn by developing more and higher-order heuristics, they also manage the complexity of what they explicitly learn with simplification cycling.
Simplification cycling exhibits two patterns that emerge from the data: the first pattern, elaboration, was expected. Consistent with prior research (Kale et al., 2002; Sapienza et al., 2006), executive teams elaborate the number and detail of country entry heuristics as they gain experience. This creates more current, comprehensive heuristic portfolios. The second pattern, simplification, was not expected. Executive teams purposefully simplify their heuristic portfolios by pruning heuristics as they gain experience. We assessed simplification cycling by tracking the addition and deletion of heuristics over time. All six firms engaged in simplification cycling—i.e., they began with a few heuristics, added more, replaced some, and subtracted others (Table 4).
Table 4. Simplification cycling
An example is F-Supplysoft, a Finnish firm selling point of sale software to help retailers manage inventory. The executive team created several heuristics during their first country entry (Sweden). One was a procedural heuristic specifying entry mode - ‘enter new countries through acquisition’. Based on their Swedish success, executives realized that relying on acquisitions would offer benefits like gaining quick access to employees who would know how to conduct business in the local market. As the CEO stated, ‘Our idea was first of all to buy (a firm to enter a country).’ During their second entry (Norway), the executive team created another procedural heuristic related to acquisition. Upon reflection, team members realized that another key to their Swedish success was the enthusiastic support of acquired senior managers. So, in addition to their first heuristic of ‘enter new countries through acquisitions,’ executives added a heuristic to ensure gaining such support in later entries: ‘ensure pre-acquisition integration of target executives.’ An executive team member described, ‘Based on our experience (in Sweden), we're talking a lot with them (company managers) before acquisition, trying to find out if they will back us up 100 percent. If they will, then we make the acquisition.’ Executives then added detail to this second heuristic after entering their third and fourth countries (France and Germany). F-Supplysoft's executives observed problems with their French acquisition that signaled the need to motivate acquired managers after the sale. This prompted executives to revise the ‘ensure pre-acquisition integration’ heuristic in their next entry (Germany) to include ‘high investment in post-acquisition integration.’ The VP overseeing Germany noted that F-Supplysoft executives ‘spent a lot of time in Germany after we made our acquisitions to make sure that we continued to integrate them and explain our values to them and how they would address new markets.’ This elaboration added emphasis on cultural and business integration post-acquisition to the heuristic.
Although these acquisition-related heuristics facilitated entry into F-Supplysoft's first four countries, the executive team later realized that acquisitions as an entry mode could also be expensive and slow. But, rather than adding further heuristics about when (and when not) to use acquisitions or when (and when not) to use other entry modes, they cut heuristics. The executive team eliminated their procedural heuristics regarding entry mode (e.g., ‘enter new countries through acquisition’ and ‘ensure pre- and post-acquisition integration’) and did not replace them with more elaborated heuristics about using acquisitions or new procedural heuristics about when to use other entry modes. This pruning of heuristics enabled firm members to improvise the entry mode based on country-specific conditions at the time, not heuristics. For example, the need to service big potential customers in the U.K. (fifth entry) led leaders to use a greenfield entry mode to establish a presence quickly. The executive in charge recalled, ‘We decided to go in, establish an office, and get a presence. So we rented a small office in a serviced office with furniture, telephone lines, everything you need. I think that has saved time. It is the right way to do it in the U.K.’ Alternatively, the country-specific conditions in the U.S. (sixth entry) led leaders to choose an alliance entry mode. The CEO said, ‘We had to make a decision how to establish operations in the U.S. We found very good acquisition targets, but prices were sky high and it was taking too much time. So we decided to go with partners.’
Another illustration is U-Analytics, a U.S.-based enterprise software firm. The two founders created customer relationship management (CRM) software to help firms ‘mine’ their data. During the firm's first and second country entries (Australia, U.K.), the executive team relied on a few selection heuristics (e.g., ‘restrict internationalization to English-speaking markets,’ ‘sell real-time analytics’) and a few procedural heuristics (e.g., ‘use implementation partners,’ ‘create a strong HQ liaison for each country’). During the third to fifth country entries, however, U-Analytics' executive team jettisoned some heuristics (Table 4). They dropped the selection heuristic of ‘restrict internationalization to English-speaking markets.’ This heuristic had been valuable because it exploited the British background of a founder and facilitated entry by focusing on linguistically and culturally similar countries like Australia and the U.K. But now, executives saw that it prevented them from addressing attractive opportunities in non-English-speaking markets like France, Germany, and Korea (entries three to five). Rather than substituting this selection heuristic with a priority heuristic (e.g., give preference to English-speaking markets) or elaborating the selection heuristic (e.g., enter English-speaking and/or large markets) to guide country selection, they simplified their portfolio by eliminating the heuristic. U-Analytics also eliminated its procedural heuristic ‘create a strong HQ liaison for each country’ because executives observed that being a liaison consumed too much time. As a HQ leader noted, ‘My expectation was that (the country) would be more autonomous than it ended up being. It required a lot of attention. It wasn't really until we were going that I realized what a huge job I had to do in headquarters to keep the U.K. and Australia in the mind-set.’ Although the executive team could have replaced their existing heuristic of creating a HQ liaison with an updated one or added more elaborate heuristics for directing the activities a liaison should do and avoid, they did not. Rather, they simplified their heuristic portfolio by eliminating heuristics.
Why do firms engage in simplification cycling? An obvious reason is that some heuristics become obsolete. But a subtler reason is that firms often replace initial, naive heuristics with strategic ones. For example, F-Supplysoft executives began with a selection heuristic that emphasized entering Scandinavian countries, beginning with Sweden, because these Finns were very familiar with Sweden. But they later substituted a more strategic heuristic around market size that was much more related to success. Firms also chunked granular heuristics into abstract ones. For example, S-Security changed a selection heuristic from ‘sell to governments, insurance companies, and banks’ to ‘sell to organizations with extensive proprietary data and ability to pay.’ This heuristic led the firm to target oil companies in Saudi Arabia, insurance firms in Malaysia, and manufacturing firms in China. Firms also substitute heuristics with greater precision, such as when S-Enterprise added their priority heuristic for the Rosetta Net standard.
Overall, replacing superficial heuristics with higher-quality ones (i.e., more strategic, abstract, and precise) again resembles the transition from novice to expert. Cognitive science research finds that experts in diverse domains like bridge, physics, baseball, and electronics use heuristics based on strategic aspects of their situation (e.g., threats and opportunities) (Feltovich et al., 2006; Chi, Feltovich, and Glaser, 1981). For example, bridge experts pay attention to the number of cards in each suit, which is more closely related to winning than number of aces (which novices track). In contrast to novices, chess experts pay particular attention to the location of the king (Charness et al., 2001), while venture experts rivet attention on customer problems (Baron and Ensley, 2006). Experts also use abstract heuristics that chunk information and generalize across situations (Charness, 1979). For example, physics experts rely on general laws like conservation of momentum to solve problems while novices attend to concrete problem features like whether the problem involves a spring or inclined plane (Chi et al., 1981). Finally, consistent with simplification cycling, experts are reflective about what and how they know and so frequently restructure and refine their representation of knowledge to access new and existing information more efficiently (Feltovich et al., 2006).
But replacing and reorganizing heuristics does not explain why firms keep their heuristics portfolios small. Psychology research suggests an intriguing reason—i.e., a fundamental trade-off between adding new heuristics to efficiently fit every situation (Goldstein and Gigerenzer, 2009) versus using a few heuristics and flexibly engaging in real-time problem solving (Switzer and Sniezek, 1991). On the one hand, adding heuristics may ‘overfit’ heuristics to experiences, create confusion, and even offer conflicting guidance. On the other hand, deleting heuristics may underexploit past experience and create mistakes. Simplification cycling may help balance this tension. Strikingly, a similar trade-off occurs in the strategy literature where extensive structures like large heuristics portfolios conserve attention and reduce mistakes by providing efficient guidance, while minimal structures flexibly open up the range of action but also introduce errors (Davis et al., 2009).
Finally, firms may keep their number of heuristics small to maintain neural plasticity, which is the degree to which cognitive systems can change (Anderson, 2000; Shepherd, 1991). Neural plasticity is highly dependent on long-term knowledge organization at the biophysical level (Hawkins, Kandel, and Siegelbaum, 1993; Koch, 1999). When that organization is streamlined into simple cognitive structures, adding new information is easy and searching existing information is quick (Cowan, 2001). So when the organization of process experience is streamlined into a few heuristics, it is easier to add or reorganize heuristics in long-term memory. This is particularly important because it enables firms to improvise action within a simple structure of rules that keeps behavior at least partially coherent (Eisenhardt, Furr, and Bingham, 2010; Miner, Bassoff, and Moorman, 2001). Overall, simplification cycling produces an increasingly able, yet small, set of heuristics that are better remembered among firm members. In sum, we propose:
Proposition 3: As experience increases, firms are likely to elaborate and then simplify their heuristics.