Semi-autonomous vehicles: What the UMTRI and GM study reveals about driver assistance systems

What the UMTRI-GM study teaches about Automation in e-shops

The study presented by DesignNews on the collaboration between the University of Michigan Transportation Research Institute, UMTRI, and General Motors around ADAS, or advanced driver assistance systems, is not directly relevant to e-commerce. It concerns systems such as automatic emergency braking, forward collision alert, lane keeping, and reverse automatic braking, technologies designed to reduce specific types of accidents before they happen. However, for an e-shop owner, the real interest lies in the logic behind the findings: when a system continuously monitors critical data, recognizes patterns of risk, and triggers the right intervention at the right time, the result can be a measurable reduction in errors, costs, and losses.

This is Automation at its most mature. It’s not just sending an automated email or connecting two tools together. It’s creating an operational “assistance system,” similar in philosophy to ADAS, that helps the e-shop prevent problems before they impact the customer experience or revenue. In the automotive industry, the problem might be a rear-end collision or a mistake while reversing. In e-commerce, the corresponding problem might be an abandoned cart, an out-of-stock product at peak demand, a suspicious transaction, a late shipment, a bad review, or a campaign that keeps spending budget without bringing in profitable sales.

The value of the UMTRI-GM study lies in the fact that it examines real-world usage data and crash outcomes, not just theoretical technology promises. This is also critical for e-commerce owners: Automation should be evaluated based on results, not on how impressive a tool sounds. If an automation does not reduce time, cost, risk, or revenue loss, then it is simply technical complexity. On the contrary, when linked to clear KPIs, it can function as a digital business accident prevention system.

According to the findings reported in the UMTRI-GM study and presented by DesignNews, different ADAS technologies are associated with different levels of reduction in specific crash categories. The graph below illustrates how important the “right system for the right risk” match is.

From ADAS to e-commerce: the same prevention logic

The most useful transfer from automotive to e-commerce is not technical, but strategic. Advanced driver assistance systems do not completely replace the driver; they support him in moments when delay, fatigue, blind spot or misjudgment can be costly. In the same way, e-commerce automation should not be treated as a replacement for business judgment. It should act as a mechanism that identifies the “blind spots” of the online store: customers who are ready to leave, inventory that is running out, campaigns that are experiencing a drop in ROAS, high-value customers who have not returned for repurchases, products with increased returns or suspicious orders that need additional scrutiny.

For example, automatic emergency braking is activated when the system recognizes an increased likelihood of a collision. In e-shopping, the equivalent is a rule or predictive model that triggers an alert when a bestseller drops below a safe stock level, when the conversion rate of a landing page drops unusually, or when a series of failed payments indicates a problem at checkout. Forward collision alert warns the driver before an accident. In e-commerce, a well-designed alerting system warns the team before sales are lost. Lane keeping helps the vehicle stay on the right track. In digital commerce, this corresponds to dashboards and triggers that keep marketing, operations, and customer support teams aligned on the same KPIs.

Technology becomes truly useful when it is connected to specific scenarios. A fashion e-shop can use inventory management automation to predict which sizes will sell out first. A cosmetics e-shop can apply personalization to customers who buy repeat products, sending refill reminders based on the actual consumption cycle. An electronics online store can trigger fraud detection rules on high-value orders with unusual behavior. A B2B e-commerce can automate repurchase offers based on order history. In all these cases, Automation is not a “nice to have”, but a loss control mechanism.

The data behind the decision: why intuition is not enough

One of the most practical messages of the GM UMTRI study is that technologies are judged by use case. It makes no sense to talk about “safety” in general if we don’t know what type of accident is being reduced, in what environment, and with what technology. The same goes for e-commerce automation. It’s not enough for an e-shop owner to say “I want AI automation” or “I want marketing automation.” They must first define what loss they want to reduce: cart abandonment, low repurchase, high service costs, late shipments, wrong stock, poor ad targeting, or payment fraud.

Independent data from the IIHS and HLDI confirm the same principle: systems that actively intervene can have a greater impact than systems that simply warn. In the field of road safety, front crash prevention with automatic emergency braking has been associated with a significant reduction in front-to-rear crashes and injury crashes. For an e-commerce owner, this difference between “warning” and “intervention” is critical. A dashboard that simply shows that cart abandonment is high is useful, but it doesn’t solve the problem. An abandoned cart automation flow that triggers email, SMS or remarketing audiences based on cart value, product availability and customer history is an intervention.

As shown in the graph below, IIHS-HLDI data shows a higher impact when technology is not limited to information, but takes action to reduce risk.

In e-commerce, this philosophy translates into data-driven decisions. If the data shows that users are abandoning at the checkout step, the problem is not solved by more social media posts. If the highest percentage of returns comes from a specific product category, the solution may lie in better size guides, clearer photos or automated post-purchase updates. If customer support is constantly receiving the same questions, a self-service help center and a carefully trained chatbot can reduce the load without reducing the quality of the experience. The point is to choose automation based on the point of friction, not based on market trends.

Where should an e-shop owner start with Automation?

The right starting point is not buying a tool. It is mapping risks and opportunities. In an e-shop, there are four areas where Automation usually brings fast and measurable results: acquisition, conversion, operations and retention. In acquisition, automation helps in campaign control, audience creation and budget optimization. In conversion, it affects checkout, product recommendations, abandoned cart flows and conversion rate optimization. In operations, it is connected to inventory management, ERP, warehouse, shipments and availability notifications. In retention, it concerns loyalty, personalization, win-back campaigns and customer experience after purchase.

A common mistake is that businesses start with the most impressive use case, not the most painful one. For a small e-shop, perhaps the biggest gain lies not in complex predictive analytics, but in reliable abandoned cart automation that works properly, with clean segmentation and thoughtful copy. For an e-shop with a large catalog, the biggest gain may lie in stock alerts and feeds that update without errors. For a brand with high repeat purchase potential, the priority may be personalization. McKinsey has reported that companies that excel at personalization generate 40% more revenue from these activities than average players. This does not mean that every e-shop will automatically see the same increase, but it shows that personalization, when based on data and not generic templates, has real business value.

The chart below captures one of the most useful benchmarks for e-commerce owners considering personalization as part of their automation strategy.

Step-by-Step application guide

Step 1: Record the loss points. Start with numbers, not assumptions. Check conversion rate by device, checkout abandonment, return rate, out-of-stock sessions, support response time, repurchase by category, and ROAS by campaign. The logic is the same as ADAS: first we identify where the “accidents” are happening.

Step 2: Prioritize by financial impact. If an issue affects a few customers but has a high order value, it may be more important than a frequent but small issue. Create a table with three columns: potential revenue loss, difficulty of implementation, and speed of delivery. The first automations should be where the value is high and the complexity is manageable.

Step 3: Choose the type of intervention. For abandoned cart automation, define different flows by cart value and customer type. For inventory management, create triggers when demand increases faster than available stock. For fraud detection, combine risk rules with manual review for high-value orders. For customer experience, automate shipping updates, user instructions, and post-purchase follow-ups.

Step 4: Measure before and after. Every automation should have a baseline. If you implement abandoned cart flow, record recovery rate, revenue recovered, unsubscribe rate, and time to purchase. If you implement personalization, measure average order value, click-through rate, conversion rate, and repeat purchase. If you implement support automation, measure resolution time and customer satisfaction. Without a baseline, there is no proof of value.

Step 5: Put human control at critical points. Just as ADAS does not remove driver responsibility, e-commerce automation should not operate unchecked in high-risk areas. High-value discounts, refunds, suspicious orders, price changes, and crisis communications need approval rules. The best system is one that automates the repetitive and keeps humans where the critical is needed.

The biggest mistakes you should avoid

The first mistake is over-automation. Many e-shops try to automate everything at once and end up with complicated flows that no one controls. The result is duplicate emails, wrong segments, inappropriate offers or customers receiving messages that are inconsistent with their actual experience. The second mistake is poor data quality. If products do not have correct attributes, if CRM has duplicate records or if analytics setup events are not recorded correctly, then Automation will accelerate the error instead of correcting it.

The third mistake is the lack of ownership. Every automation must have a responsible person. Who checks if it works? Who updates the templates? Who views the reports? Who decides when a flow should be changed? Automation without ownership gets old quickly. The fourth mistake is the one-dimensional evaluation based only on revenue. A flow can bring sales but increase returns, complaints or unsubscribes. That is why the evaluation must also include customer experience metrics.

The fifth mistake is the perception that AI automation will solve organizational problems. If the company does not have a clear inventory policy, a clear commercial strategy and consistent communication, artificial intelligence will not fix them on its own. It will require the right data architecture, clear rules and continuous improvement. The lesson from the UMTRI-GM study is exactly that: technology performs when it is placed on a clearly defined problem and measured by real outcomes.

How to build a more “safe” and profitable e-shop

If we see the e-shop as a vehicle moving in an environment full of unpredictable conditions, then Automation is the assistance system that helps the business react faster and more consistently. It will not replace commercial strategy, brand or customer relationship. But it can reduce errors, protect profit margin and create a better customer experience at scale. The UMTRI-GM study on ADAS shows that technology has the greatest value when it is linked to a specific risk and when its effectiveness is measured with real data.

For an e-commerce owner, the practical application is clear: start with the most expensive problems, implement targeted automations, measure before and after, and improve gradually. The goal is not to have more tools than the competition. It is to have fewer blind spots, faster reactions, and more reliable decisions. In a market where advertising costs are rising and consumer patience is decreasing, properly designed Automation can become one of the most essential competitive advantages of an e-shop.

DesignNews: UMTRI-GM ADAS Study

University of Michigan Transportation Research Institute, UMTRI

IIHS: Front crash prevention and rear-end crash reduction

IIHS: Advanced driver assistance systems overview

McKinsey: The value of getting personalization right