Traffic Signal Management Software: Custom Build vs Off-the-Shelf

What the off-the-shelf central systems actually do well

You have two quotes on the desk. One is a central system upgrade from your controller vendor. The other is from a software firm offering to build you a signal performance platform, and it is roughly three times the number. Before anything else: for most agencies reading this, the vendor quote is the right one.

Econolite Centracs, Cubic Trafficware ATMS and Yunex Traffic all do the hard, unglamorous part properly. They speak to controllers over NTCIP (National Transportation Communications for Intelligent Transportation System Protocol), manage controller databases, handle alarms, and sit inside a hardware and liability environment that has been tested for decades. Iteris is strong on detection and the data products around it. None of these is a weak product, and the honest criticism of them is not quality.

There is a second off-the-shelf option that sales conversations skip. Automated Traffic Signal Performance Measures came out of the public sector at Utah Department of Transportation with Purdue research and has been supported by the Federal Highway Administration, and there are open source implementations you can deploy today. If you run one controller make with high resolution logging already enabled, deploy one of those and hire an operations analyst. That is the cheapest correct answer in this category, and any developer who does not mention it is selling.

Where they stop: the mixed estate and the detector nobody is watching

A city of any age has controllers bought across three decades and four capital programmes. Econolite in the corridors done in one era, Cubic Trafficware from a later contract, a Yunex or Q-Free unit where a developer built the intersection, and a spread of firmware versions inside each make. Your central system is strongest with its own vendor hardware. Everything else is supported to the extent that NTCIP 1202, the object definitions for actuated signal controllers, covers it.

That coverage is real, and it is not the coverage you need. NTCIP 1202 gives you status and parameters. The high resolution event log, the fine grained record of detector actuations, phase changes, preemption calls and coordination status, is what performance measures are computed from, and it is not exposed uniformly across makes and firmware versions. You end up with a working central system for control and a blind spot for data across part of the network.

Here is the failure that costs most and appears in no product on the market. A left turn detector goes stuck on. It calls its phase every cycle whether or not a vehicle is there. Nothing goes dark. Nobody calls 311. The signal serves a phase to an empty lane several times a minute, breaking coordination for everyone behind it, and it does that for a year until the next retiming study finds it. Your central system does not alert, because a stuck detector is not a fault the controller reports. It reports a call.

The second thing that stops is timing plan history. Ask an engineer what changed on a corridor eight months ago and you get a memory, a consultant report and a spreadsheet of splits. The controller database has been edited in the field, in the office, and by a contractor during a project. No version, no readable difference between two revisions, no rollback.

The arithmetic: cost to build versus cost per intersection

Do this with your own numbers rather than ours. Take the annual figure on your central system renewal or the performance module quote, add the maintenance line and any vendor data hosting, and divide by your signal count. Use that figure, not a published one.

Now put a build on the same clock. In Digital Heroes delivery experience on device data collection platforms, the monitoring and performance layer runs $70,000 to $160,000 across 14 to 20 weeks. Take the midpoint at $115,000, add migration and four years of support, and you are near $215,000 across five years.

At $250 per intersection per year, 180 signals costs $225,000 over the same five years. That is the crossover. Below roughly 180 signals the vendor module wins outright. Between 180 and 400 it is close enough that money is not the deciding factor and the state of your controller estate is. Above roughly 400 signals a build is cheaper on a five year view and the gap widens annually, because a licence scales with your network and a build does not.

If your per intersection figure is nearer $100, push all of that out and the crossover moves to about 450 signals. At $300 with four controller makes in the field, it arrives closer to 150.

One line nobody puts in the comparison. The staff hours your agency already spends assembling before and after evidence for a corridor grant application, retyping tube counts, and chasing the consultant who holds last year timing files, carries a real hourly rate. In most cities it is larger than the licence.

What a custom build actually costs, start to steady state

Bands first. A monitoring and performance layer covering event log collection across mixed makes, detector health alerting, measures by approach and corridor, and a timing plan repository with version history runs $70,000 to $160,000 over 14 to 20 weeks. A full platform adding preemption and transit signal priority reporting, 311 and work order integration, corridor grant reporting and multi agency access runs $200,000 to $500,000 phased across 9 to 18 months.

Data migration runs 10 to 25 percent of the build figure and almost none of it is software. It is reconciling your controller inventory against what is physically in the cabinets, mapping historian tags to intersections because nobody ever did, and importing timing plan databases that exist in three formats. Agencies underestimate this because it looks like typing. It is archaeology.

Year two and every year after runs 15 to 20 percent of build cost annually if you want a partner reachable when a corridor goes wrong at 07:30. Add infrastructure, which on this workload is driven by event log retention rather than compute, plus a few days each year when a firmware update changes a retrieval path.

Then the line that arrives in month seven on almost every one of these projects. You will find that high resolution logging is not enabled on part of your fleet, or that some units are too old to produce it at all. That is a controller and detection capital item, not a software one, and it belongs in a different budget. Verify it in six cabinets before you sign anything.

The four situations where building wins

Regulatory and grant fit. Before and after evidence on the same approach, computed automatically around a timing change, is what a Highway Safety Improvement Program or Congestion Mitigation and Air Quality Improvement Program application asks for. Pedestrian delay and push button actuation counts are what answer an accessibility complaint under your Americans with Disabilities Act Title II obligations instead of arguing with it. A vendor module reports what the vendor decided a report should be.

Scale economics. Past roughly 400 signals on a per intersection licence, the arithmetic above stops being close, and the licence keeps growing with the network you are trying to expand.

A workflow that is your actual advantage. Agencies retime corridors on a three to five year cycle because retiming is a consultant engagement. With continuous measures you stop restudying forty approaches that are fine and fix the six that are failing. Split failures and arrivals on green tell you which six. No product shapes that targeting around how your engineers work.

Integration sprawl across three or more systems. A complaint lands in your 311 platform, a work order is raised in Cityworks or Cartegraph, the asset record lives in Esri ArcGIS, controller data sits in the central system and loading history is in a historian. Answering why a signal was bad last Tuesday requires all five. That join is the build.

How to decide in a week, then buy a specification rather than a build

Run this. Monday, pick one corridor of eight to twelve signals carrying at least two controller makes. Tuesday, send a technician into three cabinets to confirm whether high resolution logging is enabled, retrievable and time synchronised. Wednesday, pull one week of event data off two controllers of different makes. Thursday, hand it to whoever in your agency is best with data and ask for arrivals on green on one approach, by hour. Friday, count the hours it took.

That number, multiplied across your network and across a year, is your business case, and it belongs to you rather than to a vendor. If the data came off cleanly and the calculation took an afternoon, deploy an open source performance measures implementation and stop reading. If two of the three cabinets could not produce usable logs, your first project is a controller upgrade programme and no software helps this year.

If the data came off from one make only, and the join across your systems is the part that hurt, that is the shape a build fits. The next step is a paid discovery phase rather than a build. At Digital Heroes that means a signed product requirements document covering the data model, the controller configurations in scope, permissions and acceptance criteria, written before any code exists. You own it whether or not we build the thing, so take it to three other firms and get quotes that are finally comparable.

We are wrong for you if you want a replacement for signal control, conflict monitoring or anything in the safety critical path inside the cabinet. We decline that work. We are also wrong if you have no permanent technology staff and no plan to acquire any, because an operations system needs an owner inside the agency. What we bring is more than fifty specialists, over 2,000 projects delivered, a named team you meet before signing, and India LLP, United States LLC and United Kingdom LTD entities so intellectual property assigns under your own law. Check Clutch, Trustpilot, Fiverr Vetted Pro and our D-U-N-S listing first.

Book a 30-minute call with Digital Heroes and get a written plan and a fixed quote within 48 hours.