Monday, 6:30 a.m., at a third-party logistics warehouse near Rotterdam. The WMS shows 412 orders waiting to be picked, and the sorter has been idle for twenty minutes because an overnight update changed a message format coming from the control layer. Across the yard, the transport team is re-typing orders from a major retail client by hand, because the client’s new ERP stopped sending files in the agreed format. A subcontracted carrier in Poland is on the phone asking why a load of apparel has no SENT reference number yet. The chief operating officer listens to three different explanations and asks the question that resurfaces in every logistics business every few years: shouldn’t we finally have our own system?

That question has two bad answers. The first is “yes, let’s build everything from scratch, because no packaged system understands how we work.” The second is “no, let’s buy one suite from one vendor and let them worry about it.” Both treat the TMS and the WMS as one block that gets one decision.

This article is for COOs and operations directors at carriers, freight forwarders, 3PLs and in-house logistics teams at manufacturers and retailers. It argues a position not everyone will accept: in logistics, the line between build and buy does not run between the TMS and the WMS. It runs between physics and commerce. Whatever controls the physical movement of goods (warehouse automation, scanners, the WMS execution core, driver hours recording) is worth buying. Whatever makes the company money and sets it apart (pricing, carrier allocation, the customer portal, planning rules, billing) is worth building or extending. And the integration layer that connects you to carriers, customers, EDI, telematics and regulators should be yours in every scenario, whichever system you buy.

Why is “custom TMS/WMS or off-the-shelf” the wrong question in logistics?

Because it assumes a logistics system is one thing. In practice, a logistics business relies on a dozen or more functions of very different kinds: driving a conveyor, printing a label, planning a load, pricing a shipment, exchanging status updates with a customer, invoicing freight. Each of them answers differently to three questions: does it create an advantage, what does a failure cost, and how fast do the requirements change?

We have already covered the general build-versus-buy logic for systems with a similar dilemma. In the article on custom CRM development, we showed that the most expensive option is often a packaged CRM customised beyond reason. In the piece on custom ERP, we broke the decision down by process and explained why “extend” is really four different decisions. Those principles apply here and we will not repeat them. Logistics, however, has three traits that change the arithmetic.

First, the system touches the physical world. A CRM bug means a miscalculated commission. A bug in a WMS that drives automation can stop a line, and a bug in the driver app leaves a truck at the dock without paperwork. The cost of failure is counted in hours of downtime, contractual penalties and missed delivery windows, not in user frustration.

Second, much of the system’s value lives outside it. A TMS without connections to carriers, load boards, customers, telematics and authorities is a notebook. The number of those connections grows with every new customer and carrier, and each one has its own format, authentication method and data quality.

Third, regulation moves faster than vendor release cycles. In the EU, authorities are preparing to accept electronic freight information under eFTI, smart tachograph rules are extending to vans in international haulage, and several member states run their own electronic monitoring or e-invoicing regimes. In the US, the ELD rule reshaped driver hours recording years ago, and customer EDI requirements keep evolving. If every such change means waiting for the next product release, the company pays for it in manual work.

So rather than asking whether to build a TMS or a WMS, ask about each function separately: does it control the physical movement of goods, does it determine how much the company earns, or does it connect the company to the outside world? Those three categories have three different answers.

How do a TMS and a WMS differ, and why not make one decision for both?

A transportation management system covers what happens between locations: order intake, pricing, choosing a carrier or an own vehicle, route and load planning, status tracking, transport documents and freight settlement. A warehouse management system covers what happens inside the building: receiving, putaway, replenishment, picking, packing, shipping and cycle counts.

On paper, they are two similar transactional systems. In practice, they differ in four ways that drive the decision.

How standardised the process is. Warehouse processes are largely repeatable across companies. Receiving, storage, picking and shipping look much the same at an auto-parts distributor and at a cosmetics e-commerce business. The details differ, the logic does not. Transport processes vary more because they depend on the business model. A freight broker without its own fleet earns the spread between the customer price and the carrier rate, so pricing and carrier selection are everything. An asset-based carrier earns on fleet utilisation, so planning and driver hours dominate. A shipper’s logistics department mostly wants low cost and on-time delivery.

Proximity to hardware. A WMS in an automated warehouse talks to the equipment control layer, sometimes through intermediate layers known as warehouse control systems (WCS) or warehouse execution systems (WES). A TMS mainly talks to other systems and to people: dispatchers, drivers, customers. The closer you are to hardware, the higher the cost of failure and the weaker the case for custom code.

Source of advantage. In a warehouse, advantage usually comes from building layout, automation, labour organisation and process discipline, not from unique software logic. In transport, advantage more often lies in data and rules: how the company prices, which carriers it trusts, how it consolidates loads, how quickly it reacts to exceptions.

Number of external connections. A WMS typically has a handful: the ERP, client systems for a 3PL, parcel carriers for labels. A TMS can have dozens or hundreds, because every customer and every carrier is potentially a separate interface.

The first conclusion follows: in a typical company, the WMS decision leans towards buying, and the TMS decision more often justifies building some functions. That is only a starting point. A layered view, described below, gives a better map.

Why should a WMS that drives automation almost always be bought?

A warehouse with conveyors, sorters, automated storage and retrieval or a fleet of autonomous mobile robots runs on several software layers. At the bottom are the device controllers. Above them, a control layer turns commands into motion: move this tote from location A to station B. Higher up, an execution layer decides in what order and with which resources work gets done. At the top sits the WMS, which knows orders, inventory and locations.

The case for buying this part rests on three arguments.

Accountability for safety and throughput. The automation vendor is accountable for the line reaching its designed throughput and operating safely. If the customer’s own code sits between the vendor’s control layer and the WMS, every dispute about why the sorter underperforms starts with whose code is at fault. When negotiating an automation contract, check how the vendor treats integration with a system it does not know and what that means for warranties. That is not a technical detail. It is risk allocation.

Communication standards that are still maturing. For mobile robots, an open standard called VDA 5050 is developed jointly by the German associations VDA and VDMA, together with the Karlsruhe Institute of Technology, to describe communication between a robot fleet and central fleet control. That is a good direction, because it reduces dependence on a single manufacturer. But a standard does not mean you should write your own fleet controller. It means you should require compliance when you buy, so that you can add robots from another manufacturer later.

Commissioning and testing cost. Automation control software is tested on live equipment, often in maintenance windows, at night or on weekends. An in-house team would have to build a simulator, go through commissioning with the automation vendor and maintain skills the company needs once every few years. A packaged system has been through this many times.

Does that mean there is nothing to build in the warehouse? Quite the opposite. Above the execution core there is room for the functions that differentiate a 3PL: order prioritisation rules based on customer contracts, client reporting on on-time shipping and picking errors, a portal where each client sees its stock and inbound appointments, billing for warehouse services under non-standard tariffs, integration with temperature sensors in cold storage. Those are the places where custom code creates an advantage and a failure does not stop the line.

There is one exception. A non-automated warehouse with simple scanner-based operations, in a company whose process is genuinely unusual, for example goods that require individual identification and heavy documentation. There, a custom WMS can be justified, because it touches no hardware beyond scanners and printers. Even then, check whether the unusual process is real or just habit that a packaged system would help clean up.

When does a custom TMS make sense, and when is it an expensive hobby?

In transport, the arithmetic is different, because much of the system’s value is in rules rather than in hardware control. A custom TMS, or a substantial part of one, makes sense when at least two of the following are true.

The company earns its margin on pricing. If a broker’s margin depends on how quickly and accurately it prices quote requests, using rate history on the lane, seasonality, fuel surcharges and carrier risk, then the pricing engine is the heart of the business. Packaged systems have rating modules, but they are usually based on tariffs and tables. If your edge is a model your competitors do not have, it is hard to fit it into someone else’s table.

Carrier allocation follows its own logic. A company that works with a stable carrier base, scores carriers on on-time performance, claims, availability and price, and then proposes allocations automatically, holds knowledge in that logic that it should not hand over to a software vendor. Especially since the vendor may sell the same module to competitors.

The freight is unusual. Oversize loads with permits and escorts, dangerous goods with heavy documentation, cold chain with continuous temperature records, intermodal with rail and short-sea legs. Packaged systems cover these cases to varying degrees, and the gap between “supports” and “supports it the way we need” can be large.

Customers expect an outstanding digital experience. If the customer portal, shipment tracking and reporting are part of the commercial offer rather than an add-on, it is worth owning them.

A custom TMS is an expensive hobby when the company moves standard full-truckload or less-than-truckload freight, by standard rules, for customers who mostly want price and reliability. Then the advantage comes from operations, not software, and every euro or dollar spent building a system that does what a packaged one does is taken away from the fleet, the people or the margin.

There is also a middle path, and in practice it is the most common one: a packaged TMS as the transactional core (orders, documents, settlement), plus custom modules where the advantage lies (pricing, allocation, customer portal), plus an integration layer the company owns. This setup demands architectural discipline, because it is easy to slide into modifying the core, a cost we described for ERP. With well-designed interfaces, though, it gives the company what matters most: control over how it makes money, without maintaining what is a commodity.

Why do carrier integrations and EDI drive cost more than features?

When choosing a logistics system, boards tend to compare feature lists. Yet in logistics, implementation and running costs are driven more by the number and quality of external connections than by feature differences.

Look at who a typical 3PL’s TMS talks to. Customers send orders: some through a portal, some as files, some as EDI messages. In Europe that usually means UN/EDIFACT, maintained by UN/CEFACT under the UN Economic Commission for Europe. In North America it usually means ASC X12 transaction sets such as 204 (Motor Carrier Load Tender), 990 (Response to a Load Tender), 214 (Transportation Carrier Shipment Status Message), 210 (Motor Carrier Freight Details and Invoice) and 856 (Ship Notice/Manifest). Parcel carriers expose APIs for booking, labels and tracking, each with its own data model. Full-truckload carriers use load boards and their own systems. Telematics providers deliver vehicle positions and data. Then there are regulators and authorities: national e-invoicing platforms, national transport monitoring schemes, and soon eFTI platforms.

Each of those connections is paid for three times. Once when it is built: data mapping, error handling, partner testing. Again with every change on the partner’s side: a new API version, a changed format, a deprecated endpoint. And again with every new customer or carrier that must be connected before the company can start earning on them.

That is the second part of the argument: the integration layer should belong to the company. This does not mean writing every connector from scratch. It means having one logistics data model that the company controls (order, shipment, handling unit, status, document, event) and one layer that translates partner formats into that model. Then replacing the TMS or WMS does not mean re-wiring a hundred connections, and a new customer does not wait three months for the system vendor to find time for their integration.

Such a layer usually has a few standard parts: a partner catalogue with formats and versions, mappings, a message queue that survives a partner’s temporary outage (an event broker like Apache Kafka often fits here), monitoring that shows which connection has gone silent, and tools to reprocess a message after a fix. We described the general patterns in the article on software integration patterns, and our glossary explains systems integration in more depth.

A practical test when evaluating a packaged system: ask the vendor to show how a new parcel carrier and a new customer with its own order format are added, how long it takes and who does it. If the answer is “that’s a job for our integration team, priced individually,” count how many new partners you onboard per year. That number multiplied by the cost of one onboarding is often a bigger line than the licence.

How do e-CMR, eFTI, tachograph rules and national regimes shape the architecture?

Freight regulation in Europe and North America is moving from paper to data, and that shift touches the dispatcher’s daily work more than any feature list.

e-CMR. The Additional Protocol to the CMR Convention concerning the electronic consignment note entered into force on 5 June 2011, according to the UN treaty registry. Poland, a major road haulage country, acceded on 13 June 2019. In international carriage between states that are parties to the protocol, an electronic consignment note is legally possible, provided the parties to the carriage agree to it. For the system, that means electronic signatures at loading and unloading (usually in the driver app), document storage and sharing with the parties.

eFTI. Regulation (EU) 2020/1056 on electronic freight transport information aims to replace paper freight documents with data stored on certified platforms and shared with authorities on request, for example through a QR code. According to the European Commission, member-state authorities may start accepting data from certified eFTI platforms from January 2026, and must accept it from 9 July 2027. Some implementing specifications are still being finalised. For a company choosing a system today, the conclusion is simple: the data model for transport documents should be structured enough to be exposed to an eFTI platform without re-keying.

Tachographs and vans. Regulation (EU) No 165/2014 introduced smart tachographs that also record vehicle location from satellite positioning. According to the European Commission, from 1 July 2026 light commercial vehicles between 2.5 and 3.5 tonnes engaged in international transport for hire or reward must also be fitted with a tachograph. Companies that never had to handle tachograph data for vans now have to download and analyse it. In the US, the FMCSA rule on electronic logging devices has required most drivers who keep records of duty status to use an ELD since 18 December 2017, with carriers using older automatic on-board recording devices allowed to continue only until 16 December 2019.

National regimes. Some countries add their own layer. Poland is a good example for anyone moving freight into, out of or through it: its SENT system, run by the National Revenue Administration on the PUESC platform, monitors road and rail transport of selected goods. Senders, recipients and carriers must register shipments and keep the registration up to date, and the carrier must ensure current geolocation data for the vehicle is transmitted throughout the journey, through an app or a registered external location system or on-board unit. The list of goods changes: according to PUESC, apparel and footwear (CN chapters 61, 62 and 64) were added from 17 March 2026. The same geolocation infrastructure serves the e-TOLL road charging system. Poland has also made its national e-invoicing system, KSeF, mandatory during 2026, so freight invoices issued from a TMS have to reach it, usually through the finance system.

The common thread is that logistics regulation is mostly about data and its flow, not about screens. That is why it matters so much who controls the data model and the integration layer. A packaged system that supports a regime “its own way” may be enough. But if every rule change means waiting for a new release while dispatchers file registrations by hand, count the cost of those weeks.

What must a driver app do, and how should it connect to tachographs and telematics?

The driver app is where the logistics system meets reality: a dock with no signal, gloves, rain, and a driver with no time for five screens. The feature set is usually similar: today’s job list, navigation to the stop, loading and delivery confirmation, damage photos, recipient signature, exception reporting (delay, refused delivery, missing goods), transport documents including e-CMR, and any regulatory reference numbers the load requires.

Three requirements decide whether the app works.

Working without connectivity. The app has to work offline and sync when the connection returns, without losing confirmations or duplicating statuses. That sounds obvious, yet it is one of the hardest technical requirements, because it needs deliberate conflict resolution: what happens if the dispatcher re-sequenced the stops while the driver was out of coverage?

Hardware. Drivers’ own phones, company smartphones, rugged devices with scanners, tablets mounted in the cab. Each has different costs and limitations. The hardware decision has to be made before the app is built, not after.

Integration with what is already in the vehicle. This is where the physics argument returns. Tachographs and ELDs are regulated devices, and the rules around driver hours change and differ by jurisdiction. Writing your own software to read and analyse tachograph or ELD data rarely makes sense. It is the domain of specialised telematics providers who track regulatory changes and device specifications. It is more sensible to buy that service and pull what planning needs through an API: remaining driving time, planned breaks, vehicle position. Custom code belongs where that data meets business logic: in planning that will not assign a trip to a driver who lacks the hours, and in an app that shows tasks in an order consistent with the driver’s remaining time.

A driver app is a good candidate for custom development if drivers handle unusual processes (multiple consignees, returnable packaging, cash on delivery, installation at the customer site) or if the company wants one app for its own drivers and subcontractors. Apps bundled with a TMS are usually enough when the process is simple and the drivers are mostly employees. If you build your own, remember that a mobile app has its own release cycle, testing across many devices and a maintenance cost that lasts for years.

Should route optimization be built in-house or bought as an engine?

Route optimization is the area where it is easiest to overrate the value of your own algorithm and underrate the value of data.

Mathematically, the vehicle routing problem is a generalisation of the travelling salesperson problem and is NP-hard. In practice, that means for a realistic number of stops nobody looks for the optimal solution, only a good one, using heuristics. Mature engines exist, including open-source ones such as Google OR-Tools, which support capacity constraints and delivery time windows. Commercial engines add maps, time-of-day travel times and a dispatcher interface.

Writing your own solver from scratch rarely pays off. But buying an engine does not solve the problem either. Routes produced by the best algorithm are useless if the input data is wrong, and in logistics input data is wrong more often than anyone likes to admit.

What decides route quality? First, address geocoding: the point on the map has to match where a truck can actually pull in, not the middle of the parcel or the customer’s head office. Second, service times: how long unloading really takes at a specific consignee, not an average across the whole customer base. Third, vehicle and site restrictions: access bans, bridge weight limits, low clearances, dock opening hours, tail-lift requirements. Fourth, business rules: a customer who must receive goods before 9 a.m., a driver who knows the procedures at a particular distribution centre, loads that must not be combined.

So the sensible split is: buy the engine or use an open-source one, and build the layer around it. That means a data model with company-specific constraints, a process to correct address data and service times using telematics history, an interface where the dispatcher can accept or adjust the proposal, and measurement of plan versus actual. The last one matters most: without comparing the plan with what actually happened, nobody knows whether optimization delivers anything.

When the board hears a promise of savings from route optimization, ask: compared with what? If the baseline is manual planning by an experienced dispatcher who knows the territory, the gain may be smaller than the sales material suggests. If the baseline is spreadsheet planning with a growing number of stops, it may be large. Only a pilot on your own data, measuring kilometres driven and on-time performance before and after, gives an honest answer.

How should typical modules be split between buy, build and integrate?

The table below brings together the conclusions so far. It is a starting point for discussion, not a verdict: a given company may have a reason to decide differently in any row. What matters is that the reason is named.

Module or functionCategoryDefault recommendationWhen to change the decision
Warehouse automation control (WCS/WES, robot fleet)PhysicsBuy, ideally from or certified by the automation vendorPractically never; instead, require open standards such as VDA 5050
WMS core (receiving, locations, picking, shipping)PhysicsBuyNon-automated warehouse with a genuinely unusual identification and documentation process
Priority rules, reporting and client portal for the warehouseCommerceBuild, or extend through interfacesWhen clients expect nothing beyond the standard
TMS core (orders, documents, status, settlement)CommodityBuyUnusual freight that packaged systems do not model
Load pricing and carrier allocationCommerceBuild, if it is the source of marginWhen the company competes mainly on price and scale, not on its pricing model
Route optimizationMixedBuy or use an open-source engine; build the data and rules layerVery large fleet with unique constraints and its own analytics team
Driver appMixedBuy for a simple process; build for an unusual oneOne app for own drivers and subcontractors
Tachographs, ELDs, driver hours, telematicsPhysicsBuy the service, consume data through an APIPractically never
Integration with carriers, customers, EDIIntegrationOwn the layer and the data model; buy or build connectorsWhen partners are few and stable
e-CMR, eFTI, national monitoring and e-invoicingIntegrationInside the integration layer, built for quick changeWhen the packaged system updates faster than the rules change

Notice the pattern: the category column largely determines the recommendation. Physics and commodity: buy. Commerce: build where it is the source of advantage. Integration: always your own layer. The pattern is easy to remember and easy to apply to functions the table does not list.

Before filling in such a table for your own company, describe the processes before anyone talks about systems. In the article on BPMN 2.0 in practice, we showed how to model a process precisely enough for a vendor to price it, and how to capture the exceptions that, in logistics, generate most of the dispatchers’ work.

What does each option cost, and what drives cost up in logistics?

The honest answer is that cost depends on a few factors worth quantifying before you talk to vendors. Rather than quoting market ranges without a source, here is what pushes cost up and how to estimate it parametrically.

Factors that drive cost up in logistics more than elsewhere:

  1. The number of partners to integrate and how fast new ones arrive. Every new order format, parcel carrier and customer system is separate work.
  2. Availability requirements. A warehouse running three shifts and transport moving around the clock mean no daytime maintenance windows and support that works nights and weekends.
  3. Seasonal peaks. Pre-holiday retail, harvest season in agricultural haulage, sales events in e-commerce. The system must handle a multiple of average load, and go-lives must be planned to avoid the peak.
  4. Mobile devices and their variety. Scanners, label printers, drivers’ phones, in-cab tablets, each on a different operating system version.
  5. Regulation. e-CMR, eFTI, tachograph and ELD rules, national monitoring and e-invoicing regimes, and for international transport, the rules of every country on the route. Each change is work someone has to do and test.
  6. Data migration. Order history, customer and carrier master data, tariffs, warehouse locations, inventory. Often spread across several systems and spreadsheets.

How to estimate the cost of custom modules. The simplest way is parametric: team size times duration times rate. An illustrative example, not a quote: an integration layer with its own data model, connectors to several key carriers and customers, and handling for one national monitoring regime, built by two experienced developers and one tester over six months. ARDURA’s software development services page states that a typical three-person team (two developers and QA) costs about EUR 14,500–28,700 per month at 20 working days, depending on seniority. Six months of such a team is therefore about EUR 87,000–172,000, before internal costs: the time of dispatchers, warehouse managers and your IT team.

How to compare with a packaged system. A packaged TMS or WMS has licence or subscription fees, implementation, integration and support. Two lines are most often missing from the comparison: the cost of each additional integration performed by the vendor, and the cost of manual work during the period when the system does not yet support a new requirement. Estimate both from your own history: how many partners have we onboarded in the last three years, how many regulatory changes have we had to absorb, how many hours do dispatchers spend re-keying data?

Run the comparison over five years, because logistics systems live long and the cost of integration and change is spread over time. The option with the lowest first-year cost is rarely the one with the lowest total cost.

How do you keep operations running when the system is down and trucks are moving?

In logistics, a system outage does not stop the world. Trucks keep moving, goods keep arriving at the dock, customers keep waiting. The question is what people do when the system is down and how to return to a consistent state afterwards.

Fallback modes. The driver app should work offline. The warehouse should have a procedure for working from printouts or a local data copy, with operations entered later. Dispatch should have access to today’s order list even when the main system is unavailable. These modes must be designed and rehearsed, not just written down.

Replacing a system without downtime. The riskiest moment is the switch from the old system to the new one. A big-bang weekend cutover is tempting in its simplicity, but in logistics, where every day of disruption means penalties and lost customers, it is better to move traffic gradually: one lane, one warehouse, one client first, then the next. We described this approach in detail in the article on the strangler fig pattern. Logistics adds one rule to it: never cut over during peak season or just before it.

Data ownership and exit. Whatever the option, the contract should give the company the right to a full data export in a usable format, interface documentation and, for custom-built software, the rights to the code. Owning the integration layer is an extra safeguard here: if the TMS vendor goes under, raises prices or stops developing the product, replacing the core does not mean re-wiring every partner.

Contract model. When building custom modules, fixed price is tempting because it makes the budget predictable. But in logistics, scope changes during the work: a new customer arrives, a rule changes, a carrier turns out to use a different status format from the one it promised. In our article on fixed-price software contracts, we explained who really pays for transferring risk to the vendor. In logistics projects, where integrations are hard to predict, that analysis matters even more.

How do you make the decision in six to eight weeks without risking operations?

Choosing a logistics system can take a year: RFPs, demos, reference visits, negotiations. Most of that time does not improve the decision, because demos show what the vendor wants to show, not what causes your company trouble. It is better to spend six to eight weeks on work that answers the right questions.

Weeks one and two: a map of processes and exceptions. Describe transport and warehouse processes, but above all the exceptions: what happens when a consignee refuses delivery, when a carrier runs late, when goods are damaged, when the vehicle changes mid-journey on a monitored load. Exceptions are where packaged systems most often fall short and where dispatchers spend most of their time.

Week three: an integration inventory. List every partner the system exchanges data with: customers, carriers, telematics providers, authorities. For each, record the format, frequency, volume and who maintains the connection today. Add the partners you plan to connect in the next two years. This list often surprises the board with its size.

Week four: the split table. Fill in the table from the previous section for your company. For every module where you decide differently from the default, write down the reason. If the reason is “we have always done it this way,” go back to the process map.

Weeks five and six: a test on your own data. Ask two or three packaged-system vendors to show how they handle the three hardest exceptions from your map and how they connect one partner from your list, using your data rather than a demo dataset. In parallel, if you are considering building some modules, commission a short discovery phase: integration layer architecture, data model and a cost estimate.

Weeks seven and eight: decision and pilot plan. Based on the results, decide the split and plan a pilot on one lane or one warehouse, with measurable success criteria: on-time performance, number of manual interventions, time to connect a new partner.

This process does not guarantee a perfect decision, but it does guarantee that the decision is based on what actually costs money in logistics: exceptions, integrations and regulation, not the length of a feature list.

What are the key takeaways for a COO?

A logistics software decision is not about one system. It is about a dozen or more functions of different kinds. If you remember one rule, make it this one: buy the physics, earn on the commerce, own the integration.

Remember six things. First, do not make one decision for the TMS and the WMS: the warehouse leans towards buying, transport more often justifies building some functions. Second, buy the software that drives automation, and require open communication standards from the vendor so you are not locked to one manufacturer. Third, build where the margin is: pricing, carrier allocation, the customer portal and planning rules. Fourth, own the integration layer for carriers, customers, telematics, e-CMR, eFTI and national regimes, with your own data model, because it decides the cost of every change and whether you can ever replace a system. Fifth, in route optimization, buy the engine and invest in data quality and in measuring plan against actual. Sixth, replace systems gradually and never during peak season.

If you are considering building your own integration layer, customer portal or pricing module alongside a packaged TMS or WMS, see how we work in custom software development, or book a call. We start with a discovery phase that takes one to two weeks and ends with a scope, an architecture and a cost estimate you can take back to the board.