Sebastian Demleitner
No matter if the object is ignition switches, floor mats, Barbie dolls or peanut butter: product quality and safety matter. For this reason, product recalls have become a regular occurrence among many industries; playing the largest role within the automotive, food, toy, and pharmaceutical industries.
The Importance of Product Recalls
Product recalls may be voluntary (initiated by the company without external influence) or mandatory (initiated by regulatory entities e.g. the NHTSA).
Therefore, product recalls are highly relevant in not only maintaining customer satisfaction but also product safety and ensuring organizational compliance. Especially in industries like pharmaceutical, food or automotive, organizations must consider the importance of product safety regarding both personal harm and legal implications. Issues with products may lead to injury, illness, or even death and consequentially to lawsuits and loss of confidence in the company by different stakeholders such as customers or investors.
Landmark cases such as the Takata airbag disaster or more recently Boeing’s airplane door incidents shed a strong light on the importance of product recalls. The aerospace company’s stock price dropped almost 10% within the first week of the 737 incident and has since lost more than 20% at the time of submission of my paper.
Machine Learning for Recall Analysis
For my bachelor thesis, I analyzed existing academic literature regarding product recalls in the automotive sector. My task was to find and distinguish electric vehicle recall analysis using machine learning methodologies.
Existing literature may firstly be divided into research regarding electric vehicles using machine learning and general vehicle recall analysis. For electric vehicles, the focus within the literature laid on safety concerns and battery health.
General vehicle recall analysis may further be divided into research ex-post and ex-ante. While much smaller, the ex-ante focused research provided insights on an internet forum for Toyota owners. Through text mining and dictionary-based analysis methods, the researchers were able to correctly predict half of all recalls before their occurrence. The ex-ante focused field proves very promising especially for practitioners, as avoidance may be cheaper than the handling of product recalls themselves. More on the costs of recalls later in this text.
Proving to be the largest and most variable area of academia when it comes to research methodology, the ex-post focused research made up for the main part of my work. Here, both qualitative and quantitative types of methods can be distinguished and further categorized by point of interest. Namely and in order of magnitude: general recall trend analysis, reliability, accidents, quality and safety, brand preference, firm value and advertising, recall timing, lobbying and compliance, voluntary versus involuntary recalls, plant operations.
Within these categories and through different research methods, valuable insights for managers involved in automotive product recalls could be gained.
Trends in Recall Frequency and Costs
In general recall trend analysis and through mostly statistical analysis, researchers found that recalls tend to be two to three times more expensive than a regular forward distribution setting (i.e. selling the car as planned, regular aftersales timeline). Additionally, a 2019 study showed that both frequency and severity of automotive recalls are on an upwards trajectory. Researchers identified two main factors causing this: increased technological complexity of modern vehicles and an increase in customer complaints. Within those increasing numbers of recalls, voluntary ones are on the rise while involuntary recalls are on the decline. This leads to the assumption that automotive manufacturers are acutely concerned with customer satisfaction.
The literature on reliability, accidents, quality and safety has found product recalls to be a large force of good overall. They reduce the number and severity of accidents as well as the number of future recalls and injuries. Interestingly, the impact of an impending product recall could be split up into two effects: drivers’ behavioural response, i.e. being more careful when aware of potential threats, and the elimination of hazardous defects through the recall process itself.
Brand Preference and Advertising Strategies During Recalls
On a more marketing-related note, brand preference, firm value and advertising are heavily influenced by firms’ decision-making surrounding product recalls. Mainly, product recalls have been found to negatively impact both brand preference and firm value. This effect increases with brand quality, media coverage and recall severity.
To mitigate these effects, firms may opt to adjust their advertising strategies. Here, researchers found that during the crisis, it may be wiser to focus on promotion advertisement over brand advertisement while reversing that focus after the recall is concluded. If an automaker advertises the affected name plate (e.g. Camry) or a promotion (e.g. 0% finance) during the product harm crisis, the effects on brand preference and firm value are positive in the short-term but negative in the long-term. On the contrary, brand advertising (e.g. Toyota) has a positive effect at least on firm value in the long run.
If managers are concerned with a positive brand image, customer satisfaction and avoidance of legal and other costs, they should try to avoid any unnecessary recalls but at the same time handle those necessary ones as well as they can, i.e. being accommodating with customers and cooperative with regulatory entities, no matter if they are voluntary or involuntary. This may include a short time to recall, efficient communication both internally and directly with customers, and remedial marketing activities.
The High Cost of Delayed Recalls and Factors Influencing Timing
Regarding recall timing, research suggests that, considering estimated average costs $12 million for each recall, avoiding such should be managers’ priority. The literature found factors that tend to increase firms’ time to recall. Mainly, when a) recalls originate externally (e.g. complaints, policy), b) recalls stem from issues with supplied parts, c) recalls arise from design rather than manufacturing mistakes, d) multiple models are involved in the recall. Recalls also take longer for more severe issues and within more diversified brands.
Research on automotive product recalls in the context of lobbying and compliance brought to the forefront that simply considering the numbers, automotive companies have a strong incentive to increase lobbying expenditure. The literature shows that for an additional lobbying spend of $404,367, one fewer voluntary recall is observed. Similarly, for $1.66 million invested into lobbying, one fewer involuntary recall is observed. Comparing this with the aforementioned average costs of $12 million for each recall, the implication becomes clear. Additionally, as electric vehicles are twice as likely to experience a recall than internal combustion engine cars, managers may opt to increase lobbying expenditure in line with an increase in production and sales of electric vehicles.
Voluntary Recalls and Organizational Learning
This should be taken with a grain of salt however, as automotive manufacturers firstly may not be able to reduce their product recalls to zero simply by spending as much money as possible on lobbying. Secondly, product recalls are an important contributor to customer welfare and public safety and should therefore be carried out when necessary.
Comparing voluntary and involuntary automotive recalls, researchers found that voluntary recalls result in more learning than mandated recalls. Learning here meaning the reduction of future recalls. This implies that firms which focus on recalling their products when necessary – without waiting for regulatory influence or only considering recalls with regulatory pressure – tend to suffer less recalls in the future. Managers should consider these positive effects of product recalls, namely reducing the number of future recalls, and not only focus on the negative short-term consequences, e.g. costs.
Impact of Plant Operations on Recall Frequency
As the automotive industry is still heavily manufacturing based, insights into plant operations and their impact on product recalls were provided by different studies. They found that both increases in product variety (i.e. how many different models are built in the same plant) and plant utilization (i.e. percentage of capacity) lead to additional manufacturing-related recalls and therefore costs. As does an increase in plant variety variety (i.e. in how many different plants the same model is produced) combined with increased plant utilization. Direct figures are also provided: firstly, if four options for a certain model are added, this increase in product variety statistically leads to two additional recalls and $46.2 million in associated costs. Secondly, products produced in a plant exceeding 100% capacity experience eight additional recalls estimated at $167 million. Managers should tread lightly when considering an increase in any of these areas and always weigh up potential costs and benefits.
Future Directions in Recall Prediction
Finally, a small part of the ex-ante focused research has already made use of machine learning methodologies. Here, different algorithms were compared regarding their performance in vehicle quality analysis and further methodologies were tested regarding their performance in vehicle recall analysis. One main advantage of using machine learning for these tasks proved to be the much larger amount of data that machines can handle compared to traditional statistical analysis. Additionally, once fully trained, machine learning algorithms are able to perform different tasks without human intervention and can thus provide further insights.
In conclusion, I was able to provide a wide overview of the existing literature regarding automotive product recalls and their respective methodologies. It became apparent that while statistical analysis forms the fundamental basis for existing literature, many future opportunities lie within the application of machine learning methodologies. Additionally, I was able to work out many objective insights from academia that may prove valuable to both managers and policymakers going forward.