Maersk: Probabilistic Attribution for 2026 Supply Chains

Trying to pin down marketing attribution in the global supply chain is a nightmare of logistics and data. For a company like Maersk, with ships and trucks all over the planet, they can’t just look at the last click and call it a day, not when their B2B sales cycles are long and involve huge contracts. Those simple first-click or last-click models totally miss the complex journey a customer takes. So what’s the alternative? Probabilistic attribution, which gives a much clearer, more realistic picture of what’s actually working for a logistics giant.

Beyond Last-Click: The Imperative for Probabilistic Attribution in Supply Chain Marketing

For years, marketers got by with lazy attribution, mostly last-click. It was easy: just give 100% of the credit to whatever the customer did right before buying. But that’s a completely broken way to look at B2B sales, where deals take months and touch dozens of channels. Think about a typical Maersk client’s path: they see a LinkedIn ad, poke around the website, join a webinar, grab a whitepaper, and talk to a sales rep before they finally sign. If you give all the credit to that last sales call, you’re pretending all the marketing that got them there didn’t matter.

Probabilistic attribution is the smarter way. It uses statistical models and machine learning to calculate the odds that any given touchpoint actually influenced the final sale, then assigns partial credit accordingly. It’s all about using your historical data to find the real patterns and causal links. The model looks at everything, the sequence of interactions, the time between them, the specific content someone looked at, and learns what combinations tend to lead to a conversion. This gives you a real understanding of how your campaigns are performing, all the way from a first glance at an ad to the final signed contract.

Maersk’s complex global supply chain only makes this kind of advanced attribution more necessary. Their customers are all over the world, in different industries, with totally different ways of making decisions. A shipping manager in Rotterdam isn’t going to follow the same path as a logistics director in Shanghai, and a good probabilistic model can handle that mess, giving you a truer picture of your marketing ROI by region and segment. The focus shifts from “which channel got the final click?” to “how much did each channel contribute to this person converting?” That change in perspective has a direct line to how you set your budgets and plan your next move.

Data Foundation: The Foundation of Effective Probabilistic Modeling

A probabilistic attribution model is only as good as the data you feed it. Period. For a company the size of Maersk, that means pulling together a ton of different information streams. You need data from web analytics platforms like Google Analytics 4, from your Salesforce CRM, from your email marketing software, from all your ad accounts like Google Ads and LinkedIn, and you even need to get the offline stuff your sales team is logging. The real work is the painful process of centralizing, cleaning, and stitching it all together so you can actually follow a single customer’s journey across all those systems.

The dream, of course, is a unified customer ID. That’s a single, persistent ID you can use to track a customer as they hop between their laptop, their phone, and different channels. You can get there with deterministic matching (which is pretty reliable, using an email or login ID) or you can try probabilistic matching (which is more like an educated guess, using IP addresses and device fingerprints to connect the dots). The better you are at building this unified customer view, the more you can trust your attribution model. If you don’t have that data infrastructure sorted out, your fancy algorithms are just going to spit out junk.

Once you have the customer journey data locked down, the next level is pulling in external factors to make the model even sharper. Think about things like major economic shifts, seasonal shipping rushes, new global trade policies, or what your competitors are up to. All of that affects customer behavior and how well your marketing works. It’s a huge project to try and integrate all of those variables, but the point is to build a model that actually mirrors the messy reality Maersk’s customers live in. To even attempt this, you have to get marketing, data science, and IT into a room and working together, which is a political challenge a lot of companies just can’t seem to solve.

Building the Model: Algorithms and Variables for Supply Chain Context

With your data pipeline finally built, you can get to the fun part: picking and training the algorithms for your probabilistic attribution. You’ve got a few common options, like Markov chains, Shapley values, or standard machine learning models like logistic regression and neural networks. There’s no single “best” one. The right choice depends on your data and what you’re trying to figure out. For instance, Markov chains are great for this kind of work because they’re designed to model sequences, like the probability of a user going from ‘viewed an ad’ to ‘requested a quote’.

The model is only as smart as the variables you feed it. For a business like Maersk’s, you’d absolutely need to be tracking:

• Touchpoint type: Was it a display ad, a search ad, an email, a webinar, a sales call, or just a visit to the website? They all have different weights.
• Engagement metrics: You can’t just count a visit. How long were they on the page? Did they click any CTAs? How far did they scroll? Deeper engagement usually means more influence.
• Time decay: A touchpoint from yesterday probably matters more than one from six months ago. Your model needs to account for that with a time-decay factor.
• Position in journey: Top-of-funnel brand awareness ads do a different job than a late-stage pricing page visit. The model should be able to tell the difference.
• Customer segment: Different types of customers will react to your marketing in different ways. You can build separate models for each segment or just include customer characteristics as variables.

The model gets smarter over time by learning from your historical data, figuring out which sequences of touchpoints are most likely to result in a sale and assigning credit based on those patterns. It’s a living system. As more data flows in, the model constantly refines its understanding of how customers behave. In a market like global shipping that can turn on a dime, that kind of adaptability is exactly what you need. You’re not aiming for a crystal ball that’s 100% perfect, you’re aiming for an estimate of channel effectiveness that gets better and better every quarter.

Interpreting Results and Driving Actionable Insights

So you’ve built this complex probabilistic attribution model. Now what? Building the model is one challenge, but actually using its output to make smart marketing decisions is the entire point. The model will give you a detailed report card on every channel and campaign, showing you how much it actually influenced conversions. For instance, you might find out that your flashy display ad campaign is getting tons of impressions but has almost zero impact on actual sales, while a small, targeted email campaign is quietly driving a huge percentage of your deals.

With that kind of insight, the marketing team can reallocate the budget with some actual data to back them up. If the model shows that a certain content marketing initiative is a beast for generating early-stage leads, Maersk might invest more in that content type. If a pricey ad channel is shown to be a waste of money, you can finally kill it without a political fight. This approach replaces “this is how we’ve always done it” with hard evidence and optimizes the whole funnel.

These models are also great at finding hidden gems in your marketing. You might discover that some forgotten resource buried in Maersk’s whitepaper library shows up again and again in the journeys of your best customers. That’s a huge signal. You can start promoting that resource more heavily or make sure your sales team is using it in their conversations. The detailed insights from these models let you make strategic adjustments that have a real impact on the bottom line, instead of just making big, blind bets.

Challenges and Continuous Refinement in a Dynamic Ecosystem

Let’s be real, implementing probabilistic attribution at an organization like Maersk has its challenges. The initial setup requires a big investment in data infrastructure and data science talent, not to mention getting departments to work together. You also have to navigate the nightmare of privacy laws like GDPR and CCPA, which means careful data anonymization and consent tracking, especially with global customer data. Because Maersk’s operations are global, you’re also wrestling with different data standards and legal rules in every country, which makes just getting all the data in one place a huge headache.

The work doesn’t stop once the model is live, either. Digital marketing and customer habits are always in flux. New platforms pop up, existing ones change their algorithms, and customer preferences drift. A model that’s dead-on accurate today will slowly become useless if you don’t keep it updated. This requires a feedback loop for regularly checking the model’s predictions against real-world sales numbers. Running A/B tests based on what the model tells you is a great way to both validate its insights and keep refining it. This constant tuning is the only way the attribution system stays relevant.

Look, probabilistic attribution isn’t some magic fix for all your marketing problems. It is, however, a powerful analytical tool that gives you a much truer read on marketing effectiveness than you’ll ever get from last-click. For a global leader like Maersk, it provides the clarity they need to make sense of their complex B2B marketing efforts and make sure every dollar is actually working toward their goals. It’s about finally understanding influence instead of just counting clicks.

When a company like Maersk adopts probabilistic attribution, they get a clear map of what’s actually driving people through their long and complicated customer journeys. That clarity leads to smarter budgets and better campaigns. The result is a much healthier marketing ROI, and as a side benefit, it’s great for spotting and fixing funnel leaks you never knew you had.