GLS: Creating more efficient delivery tours with Honeycomb

Introduction

In today’s fast-paced logistics industry, efficiency is everything. People expect deliveries the same day or at most the next day, avoiding companies with poor delivery experiences altogether. For delivery services, even minor improvements and optimisations can translate to substantial savings in time, costs, and fuel — improving profitability and reducing their environmental impact.

The collaboration between GLS Studio and Honeycomb came about to improve the efficiency of delivery routes and specifically to improve the UI of the GLS ParcelPlanner app: used by fleet managers and subcontractors to optimise delivery routes. GLS Studio is an interdisciplinary high impact unit driving innovation within the GLS Group, one of the leading parcel service providers in Europe and part of the Royal Mail Group: delivering over a million parcels a day in peak season.

We spoke to the co-founders of GLS Studio, Stefano Bourscheid and Sinan Theuvsen, about their experience using Honeycomb and how it helped optimise their deliveries.

Visualising driver areas

Geospatial data is a core part of GLS’ business. Location is at the basis of how they divide work between drivers and plan to allocate resources. Efficiently managing who's going to pick up which parcel and when a specific delivery is optimally positioned and sequenced requires constant optimisation. To help subcontractors with this process, GLS Studio was developing the core of its ParcelPlanner app, but they struggled to find a good solution to visualise different driver areas. The goal was to create a system that automatically allocates parcels to a specific driver based on geo-fenced driver areas that could easily be drawn, edited, and implemented.

GLS hadn’t yet developed a standard internal tool for this purpose and other external tools which offered map views involved continuous back and forth between developers and data analysts, downloading and re-uploading data sets. This process sucked up a lot of resources before a simple data set could be visualised, analysed, and improved.

“If you use Tableau, you have to have the data in Tableau and then need to find a way to transfer it to another tool like Metabase. This also means you cannot easily pass on learnings from one team to another, because each team is working with their own code so you have to send along the query. We just want to look at some data, which shouldn’t be rocket science: show me a map with data and let me interact with it.” – Sinan

The process had previously been very informal with drivers claiming certain streets or addresses and sorting through parcels at the depot to deliver to those addresses. It was crucial to get this information on a map in an aggregated form through H3, because ParcelPlanner wasn’t allowed to display individual driver stops for legal reasons. It was at this point that GLS Studio first started working with Honeycomb.

“The anonymized, out-of-the-box data visualization allowed us to get started really fast with defining the different areas for drivers” – Stefano

Honeycomb immediately proved a great fit. Without requiring complicated pipelines to be built and connected, the tool visualises data from the H3 grid – already used by GLS Studio to perform area calculations. The fact that this functionality is built in, that driver areas can be geofenced based on custom shapes, and that desired KPIs are automatically calculated for those specific areas was exactly what GLS Studio was looking for.

Speed and interactivity

Starting to work with Honeycomb was a very quick and frictionless process. It took their team two days to implement from the very first meeting to signing a contract and implementing Honeycomb on their frontend web app. From a technical perspective, GLS Studio found integrating Honeycomb just as easy and quick as adding a few lines of iframe code to their custom app. There’s no need for database credentials – provide Honeycomb with the url of a secure S3 bucket and it automatically pulls data from there.

“Honeycomb makes it super easy to share and visualise data. Whether you work through uploading a CSV or through the integration with Snowflake, Honeycomb works at a really fast speed, which I love. GLS Studio is not a corporate that has a fixed pipeline that we follow for the next 3 months. If I have an idea, I want to see it in two hours – not in months. Honeycomb is perfect for this.” – Sinan

With other tools, investigating and getting the visualisation required would take GLS Studio at least two or three days. If they didn’t get the results they expected the first time, they’d run another round of tests which would take another few days. With Honeycomb, they found that data exploration tasks often produced desired results within 90 minutes – from defining a hypothesis, getting data, uploading the CSV, examining the data, drawing conclusions, and sharing the results – and were able to move on to the next investigation. The collaboration reduced these tasks from at least a few days to a few hours at most.

A key draw of using Honeycomb was how it works entirely in the frontend (running in a user’s browser) and how users can interact with the tool in real-time. Another was that it hardly requires any development work from partners. GLS Studio didn't have to build all of the backend infrastructure and logic to crunch numbers and display their KPIs. At that point ParcelPlanner only had a login page, so Honeycomb was its very first feature. Throughout the collaboration, new features were added to improve the interface but the core technology and functionalities stayed the same. Honeycomb quickly became GLS Studio’s go-to tool for instantly and interactively testing, viewing and sharing results.

“In terms of that level of interactivity and allowing us to be hands off, there isn’t another tool that could provide such a tailored service for us, even compared to the most expensive off-the-shelf data visualisation solutions.” – Stefano

Delivery tour efficiency

GLS Studio’s goals are not only to reduce driving time through ParcelPlanner, but also to equalise it across drivers. Reducing driving time and preparation time means drivers can deliver more packages, helping drivers earn more money in the same working day.

Before ParcelPlanner, drivers would need to drive to the depot, pick out their parcels, organise them and put them in the van before being able to start driving their tours. Internal research found that within a 9.6 hour day, drivers only spent about 5.8 hours delivering on average. Every minute is valuable in this process and improving the productivity within these timeframes is key. Now, drivers arrive at the depot with parcels already organized in an optimal sequence for them by a separate team so they can quickly load the van and leave immediately: saving 10,000 drivers in Germany alone up to 3 hours per day.

“Honeycomb helped kickstart this work, providing ParcelPlanner with key functionalities (with basically no code) which not only makes it possible to view the areas, but also to edit them, save them, and see those changes immediately reflected in terms of the labels and owners of our driver areas. It’s really important to be able to do this in real time and see this reflected quickly. This allowed the team to focus on user testing, testing assumptions on vital tasks like relabelling parcels, and arranging pre-picking.” – Stefano

There are naturally some imbalances in terms of driving time, because some drivers are more experienced or motivated than others. However, qualitative research showed that this leads to complaints from drivers who perceive that they’re doing a lot more work than others. That’s why distributing work between drivers more evenly – and thereby improving work satisfaction – is another goal.

To avoid one driver delivering 300 parcels and another delivering 500, load balancing comes into play: either assigning additional stops to a driver in a quieter area on a specific day or through redrawing driver areas on a more long-term basis. Other factors also need to be taken into account like the number of packages per address and the weight of the packages: there’s a big difference between delivering 200 parcels at 2 stops and delivering 200 parcels at 200 stops, while the weight of the packages affects how many a van can carry due to legal weight limits. Honeycomb helped the GLS Studio team develop the entire tour management system that they now have in place, without spending six months building custom maps with their own tools.

The results are eye-catching. GLS Studio launched a small-scale pilot with a new user over the course of three weeks to test their assumption that introducing their map tool, powered by Honeycomb, would help users stabilise tours across delivery drivers and consequently improve performance. Stable tours were immediately obvious on the map, while bad tours stood out because they look very random with drivers going all over the place: providing visual evidence of inefficiencies in the delivery process. While only based on limited data, they immediately saw major improvements as the stability of tours steadily increased up to around 90%.

“We proved that we can improve things just by making them transparent and tangible through visualising the data on a map. Honeycomb was very helpful in creating transparency and improving communication around our data.” – Sinan

End user experience

Using the new system required some adjustment from the fleet managers and delivery drivers that use ParcelPlanner on a daily basis to manage the day ahead. They were used to viewing and discussing deliveries based on postal codes, but the downside of this approach is that it’s hard to determine the distance between different places (without encyclopedic knowledge of the area). When GLS Studio initially floated the idea of using maps instead, users had been reluctant as they weren’t used to using digital maps for deliveries. However, they were pleasantly surprised when they were introduced to Honeycomb's solution and immediately saw the benefits.

“Our users were very impressed with the look and feel, and how data was visualised through Honeycomb’s hexagonal cells (H3). They thought it was a very innovative way to anonymise and obfuscate data without losing too much geospatial info.” – Stefano

Users quickly started adopting the solution and got used to working with it on a daily basis. They particularly like the real time interactivity when they change driver areas and immediately see the numbers reflected in the UI. Past solutions at GLS weren’t able to update in real-time: users would need to click a button, wait minutes for a task to run, come back to the page and reload it before getting the updated numbers. This involved a lot of feedback and was a very tedious process compared to using Honeycomb.

“Honeycomb does everything in the frontend and seeing it perform so fast in terms of changing areas and updating numbers really impressed end users.” – Stefano

Moving forward

Following the successful collaboration around ParcelPlanner, GLS Studio is looking into other areas where Honeycomb’s visualisation capabilities could benefit the business, including data exploration. Now Honeycomb is fully integrated in Snowflake, users can take any table with geospatial information in Snowflake and easily visualise that.

“With the integrated Honeycomb solution, they have the magic button that everybody always wants. One button, done.” – Sinan