The Drone Blog
In 2020 I got the chance to join a new very interesting initiative. One of the factors which made it interesting was that it did not start when I was asked, so I got the chance to initiate it from scratch instead of joining something which was running already like on most of the previous assignments.
The intention of the initiative was to try if it was feasable to implement a drone based camera system to check the placement of pallets in the warehouse. Several Vendors were invited and interviewed. The focus was to find a good match between offered capabilities of the drone solution and the requirements and wishes of the customer.
But next to the pure elaboration of the current solution the wish was as well to find a strategic development partner which is willing to actively support the process of identifying and analysing new business cases where a process could benefit from a smart usage of indoor drones.
The below is of course not to be understood as an accurate sequence. Many activities were done in parallel and overlapped. But the below gives a good impression of the steps taken und how the customer achieved the wished goal in a comparably short timeline.
High flying visons
Drones are a wonderful toy and people tend to become very creative when new technology is on its way.
The ideas which are created are very wide and open and those brainstorm sessions are always fun and productive. Anyway, sooner or later reality kicks in and one needs to agree upon the ideas which have the biggest potential.
So after two broader workshops to collect Ideas from the customer and do reality checks with the vendor we had to drill down to one reasonable business case to be used for a prove of concept. For the beginning the focus was put on an increase of the stock accuracy in the warehouse.
A second area whihc was put on the "high potential" list was an option to perform safety checks in the WH.
07 '20
Two remaining Vendors
Two vendors have been identified as potential candidates for a long-term relationship.
One offers a fully autonomous system which allows the drones to perform checks on the different storage locations in the warehouse during the non-operational hours in the night and the weekend.
The other offers a semi-autonomous system which requires an operator on site observing the drone activities and which focuses on complete checks of specified racks.
Next to the two different approaches the two vendors came as well with significantly different price tags and different company backgrounds. One big player in the supply chain since many years and one newbie in this area.
Hopefully we can benefit from the experience of a global player as well as from the out of the box thinking a start-up can do.
08 '20
Different sites - different processes
The customer operates several retail sites with embedded warehouses but also central distribution units (DC) and customer order distribution centers (CDC).
The stores perform stock checks in most cases based on the range. Meaning all locations which hold a specific article shall be checked. This leads to a chaotic distribution of check tasks across the complete warehouse, where a limited number of locations per rack must be checked.
The different distribution units perform the stock checks per area. Meaning the counting tasks are not spread across the warehouse but instead they are bundled in racks or sections of a rack.
This setup led to the decision to still test both systems in both unit types, but the focus for stores was put on fully autonomous solution and for the CDCs on the semi-autonomous solution.
DCs are not in scope currently.
09 '20
Ongoing preparation
Drones seem to be a very appealing topic. The search for ambassadors turned out to be simple in this case. Main stakeholders have been identified as well and the budget is secured.
In parallel the risk assessment is performed on both solutions. Different aspects are considered, but the three most important topics are:
Safety: What needs to be in place to ensure that the drones don't harm co-worker or customers?
Legal compliance: Which documentation and follow ups need to be in place to result in a tax compliant stock check?
Data security: Are GDPR-rules followed and will the solution fit as well other legislations like "China cyber law", "California consumer privacy act" or Russian cross border rules?
Some of these questions are difficult to answer and not all of them must be answered in the first step but can be taken when implementation in the according regions is planned, but it is good to have them in mind already at this stage.
10 '20
Pilots identified
The first test sites have been identified. The fully autonomous solution will be tested in a store in Switzerland and the semi-autonomous in a store in Italy.
In the coming month another fully autonomous pilot will be performed in a customer order distribution center. This is not reflecting the original plans, but the customer is the king.
The stores and the distribution units use different WMS systems. Both need to support in the future the standard interface which is supposed to be developed in close co-operation between the drone vendors and the customer.
11 '20
Proof of concept
For the first store a proof of concept has been agreed and set up.
Without any substantial IT-development a simple SQL query was created which fetches information from the WMS system. That data is forwarded as a straight .csv file via mail to the drone vendor.
On the vendors side the data is parsed and fed into the drone system. Based on that data the drone flights are planned and scheduled. A response back to the WMS system is not planned for the prove of concept since the drone vendor offers front end which is good enough for this state.
Two different stock check processes performed in two different WMS systems with two different drone vendors will require a quite flexible interface. But like always, tools and processes are different and change from time to time but the underlaying information is very similar and stable.
In the end we all talk about warehouse locations, one or several SSCC-numbers stored per location. Those SSCC's hold one or several SKUs with a defined quantity. On top we have some time stamps and statuses which can be used to identify the current state.
One of the main tasks during this proof of concept will be to identify those information objects which are relevant and map them in a way that one standard interface can be used for all.
12 '20
Many findings on Working Methods
The proof of concept proved to be a very good idea.
Many cases were identified where the stores did not follow the global working method and created work arounds to ensure that they still stick to the legal requirements.
This was mainly found in the handling of multi deep storage, non-selling goods and SSCC - label usage.
After several meetings and workshops with the Business stakeholders we decided to accept some of the ways of working from the POC store and consider them in the future interfaces.
To avoid those which were not accepted in the future a training program called "raise the base" was agreed. We try to get a permanent follow up on that raise the base initiative by adding relevant questions to the generic store review process.
01 '21
Buisiness Case approved
After pulling out a lot of data during the last months, a lot of meetings and calculations, the business case got approved and the actual development work can start.
This is an important milestone for the initiative.
It helped for sure that the idea won a company internal award for the most innovative initiative.
Like mentioned above: Drones seem to be very appealing.
02 '21
IT development started
All that glitter is not gold.
The award made this initiative very public in the company and the original idea to test and try and conclude changed to: "Make it happen – now!". So, there is a lot of pressure created now to deliver a running solution.
The pro on that is, that the initiative was staffed with an additional developer and that the development of a standard API has started.
The Con is that we are forced to take short tracks which will leave technical dept. And with an internal innovation award the expectation management is significantly more difficult to handle.
03 '21
Roadmap with strategic partner agreed
Development is not only needed in the customers WMS system and the existing integration gateway for all kinds of automations, but as well on the vendor’s side.
Next to the actual interfaces there is one big change requested from the vendor. One part of the application is today not available in any acceptable enterprise standard on a supportable and manageable secure platform but only on a standalone PC.
It took some time but now there is roadmap, and the questioned application will be moved from this PC into the cloud. Since this is a massive change for the vendor the change is estimated by the vendor to be completed by April 2022.
As of now we will continue with the current application and architecture and got an ok to deploy this setup in up to max 5 stores.
04 '21
First Tests
Like often the actual IT development is the fastest part of an IT initiative.
The first parts of the API are on their way and the first tests are performed to ensure connectivity and communication between the locally hosted WMS and the AWS hosted drone solution.
For the customer as well as for me it is the first time to work with the amazon cloud, so some findings are to be expected.
All initial tests are performed with mock-up data, so I am very curious about the first "real" transactions.
05 '21
Roll out material
In parallel to the development, the roll out material is created.
Doing such a change to a customer with so many different sites require a lot of preparation.
Awareness videos, technical instructions, education material, hyper care support flow documents, etc. are created to get those stakeholders on board which have not been involved that much up till now.
This is an important phase since this material and the preparation of people working at organisations like the customer internal first level support, or the project implementation teams will drive to a huge extend the perception and the acceptance of the solution.
06 '21
Go Live Step 1 & 2
The first pilot is in place and up and running.
First the transactions from the WMS system to the drone system were implemented. The new API worked smooth and without any major problems from day one.
The drone vendor decided to receive the data via the new API but keep it in the shadow for at least one week to compare it to the data the receive via the report based POC set up.
After 2 weeks without any problems the second step was taken and the API which handles the communication of the drone results back into the WMS system was activated.
For the time being we consider in the WMS system only the "OK" results. Locations where the drone "said" they are correct are considered as correct in different WMS procedures. Those which were flagged with any kind of issue by the drones are not updated in any way in the WMS.
So currently the drone will not fix any problems, but "only" find the problems.
We decided on this approach since it saves the team from creating complex business logic but still saves the user 95% of the workload.
Auto correction of data in the WMS will be something which can be developed in small steps case by case in the line of business and the future product owner without a globally supported initiative.
07 '21
Handover and closure
The hand over is completed and the initiative will be closed this month.
The solution is by now up and running in three stores and the receivers are very pleased with the results.
Next to those three Pilots 35 additional stores have been identified globally which will be prioritised to receive the solution. The Vendor plans to scale up the implementation teams on their side so that they should be able to deliver appr. 50 implementations per year in the future.
An implementation plan on store level was created which guides the according store implementation team step by step through the process per store. For each of the steps the needed material was provided.
The development teams on the customer- and the vendor-side have handed over to the implementation teams and the according contacts between my customer and the vendor have been established as well on a personal level.
And finally, the report solution from the original POC has been closed.
A long list of future potentials has been handed over to the line of business and now it is up to them to fine tune the new product and to adjust it to fit new upcoming business needs.
08 '21
Reflection
The initiative took longer than expected. The proof of concept and by that as well the actual IT-development started about 3 months later than originally planned. This was mainly caused by many discussions about the need for this at all and other conflicting priorities. Once those discussions came to an end the complete initiative suddenly got more attention than wished, which created a lot of pressure.
This showed that a clear plan with a clear timeline should be in place always upfront. Unfortunately, most plans do not survive the first contact with reality and become null and void to a certain extent.
So, planning on the one side but staying flexible to adjust to new circumstances was – like very often - key success criteria.
In this case the actual delivery suffered from changes in priority and direction mainly in two areas:
1. There is no tested or even used interface for the second vendor.
2. There is no tested or even used interface for the WMS system in the distribution units.
Still the most important delivery was achieved. A solution which is running and ready to be implemented in the 400+ retail sites of the customer.
Next to that there are two new IT developers which learned a lot about the structure of the customers IT organisation and the used applications. Both will be kept in the IT department of the customer to support as well future initiatives.
Even though there were some struggles in the beginning and the political dimension became too powerful, I'd say it was a successful initiative which resulted in a beneficial solution which will improve the customers stock accuracy and decrease the manual workload.
09 '21
Update
Over the years I got contacted by the customer or the vendor regarding this Solution every now and then, so I never lost contact completely. Now 4 years later I'd like to take the chance to give a short update.
The Stream with the second vendor was not followed but the first vendor showed to be a very good choice. The Solution has been implemented by now in more than 150 sites in many countries across Europe and North America. The Asia Pacific Area is still not on board, but the implementation continues so that region might come as well.
The Solution has been further developed with many very positive effects. The flight time of the drones has increased and the quality of the cameras has been improved significantly which allowed the vendor to open a lot more doors in the area of machine learning in picture recognition.
They have by now a really impressive module which identifies safety issues in the rack like pallets which are not placed correctly, loose hanging plastic straps or damages on the racking or the sprinkler system as such.
Next to that they started as well a module which is capable to measure the fill rate of locations which is very interesting especially for those locations which are accessed by customers and where the goods are due to that not always placed in a very structured manner.
The follow up and issue identification user interfaces have been improved a lot and reached actually a level where the customer decided to not implement any automatic stock update in the own ERP system. So users in the stores nowadays work actually in the drone system and it is absolutely good enough to serve the many.
The technical foundation has been lifted to a modern enterprise level, where the entire application is hosted in the cloud and supports single sign on.
The integration has been moved to the new customers integration platform which is today used in all different business units and supports the still existing landscape of different warehouse management systems.
Overall a very positive development.
One could splitting the entire solution into hardware (camera and drone) and software (processing the pictures). Especially on the drones as such I think there could be better options. Permanently mounted cameras or cameras mounted on autonomously driving vehicles could be alternatives, both would come without the limitations caused by the high energy consumtion and limited weight capacity of a flying device. On the other hand especially in this initiative the "drone" was the fancy part which gave the needed attention from stakeholders and decision makers.
On the software part the vendor has proven in an impressive way that modern visual systems with machine learning based picture recognition have today reached a level where they can replace bar code scanning or RFIDs in many use cases.
Plus the benefit that pictures - once taken - can be used for a lot of different use cases (e.g. stock and safety can be checked on the same picture.
With more initiatives like this where customers and Suppliers build strong relationships there are many great opportunities and a lot of up till now undiscovered business cases can be realized.
11 '25