Teamsters Pass Unanimous Resolution to Challenge UPS Over Alleged Use of Roadie Nonunion Drivers, Escalating Labor Tensions

The Analysis

A parcel that gets scanned at 14:05 and misses the 14:20 van does not show up twenty minutes late. It shows up on the next cycle, which is about twenty-four hours later, and then somebody has to explain that to a customer who was promised delivery in two days. Every hour I have ever fought for in this job comes back to that one piece of arithmetic, and it is why I read labour news the way other people read weather forecasts: not to pick a side, but to work out how many minutes the next squall is going to cost me on the dock.

Let me say what did not happen before I say what did, because panic is expensive. Nobody has called a strike. Nobody has announced a lockout or a slowdown. What came out of the International Brotherhood of Teamsters convention in Las Vegas is a resolution, and it passed unanimously. It directs union leadership to confront UPS over what the union calls a breach of the 2023 collective bargaining agreement, specifically the alleged outsourcing of delivery work to nonunion subcontractors. The campaign names UPS's own subsidiary Roadie, a gig platform running final-mile work on lower-paid independent contractors.

The union's argument is that the 2023 contract explicitly forbids UPS from assigning work performed by its members to companies using gig labour, and the evidence it points to is that Roadie runs on UPS labels, UPS tracking and UPS equipment. It also says UPS has spent the past four years scaling Roadie nationwide, and that dozens of Roadie distribution centres and cross-dock facilities have appeared in that time. This is not a new complaint either; the union raised the same one publicly last November.

On the numbers, because the numbers are the only reason this belongs in your week: the union represents roughly 330,000 UPS employees, and about 100,000 of them are drivers. That is the size of the organised workforce whose work is at the centre of the argument. Set against it is a network the union itself describes as dozens of facilities deep, already built, already staffed on a different model, already moving parcels today. A contract argument runs on a calendar measured in months. Delivery runs on a calendar measured in minutes. Those two clocks do not line up, and the gap between them is where shippers get hurt.

Strip away the legal language and what is actually being argued about is who touches the parcel in the last sixty miles. That matters operationally for reasons that have nothing to do with labour law. An integrated unionised delivery network runs on fixed routes, fixed start times and a driver who is obliged to complete the route. A gig layer runs on tender and acceptance. Somebody offers a bundle of stops, somebody else decides whether to take it, and the price and the timing move together. Same brown van, same handheld scanner, completely different behaviour when volume spikes at four in the afternoon on the last Thursday before a holiday.

There are three clocks in any parcel operation, and a dispute like this one touches all three in different ways.

The first clock is the pickup cut-off. Whatever time your carrier says the last van leaves, that is the hour your whole promise hangs from. Everything upstream of it is your own warehouse slack, and everything downstream of it belongs to somebody else's plan. When a network changes shape, this is the number that moves quietly, usually by thirty minutes, usually announced on a portal page nobody reads.

The second clock is the handoff between linehaul and the delivering station. This is where facilities come in. However many of those Roadie cross-docks and distribution centres exist, each one is a place where a trailer has to be broken down, re-sorted and re-loaded onto something that goes to a door. Add one extra touch and you add hours, not minutes.

The third clock is the first delivery attempt, and it is the one almost nobody models. A first attempt that fails does not cost you twenty minutes. It costs you a full return-to-station cycle, a re-sort cycle, and another outbound cycle. In most networks that is twenty-four hours, not two.

Let me build one honest hour model so we have something concrete to argue with. Everything below is assumption, and I am labelling it as mine so you can put your own numbers in. Assume a customer orders at 13:00 on a Tuesday. Assume your own pick, pack and label cycle takes three hours, so the parcel is ready at 16:00. Assume your carrier pickup cut-off is 17:00, giving you one hour of slack. Assume pickup scan to the originating sort facility is six hours. Assume sort plus outbound loading takes five hours. Assume linehaul to the destination metro takes fourteen hours. Assume the destination sort takes four hours. Assume the station-to-van segment, meaning load, manifest and dispatch, takes six hours. Assume the delivery window itself is eight hours wide.

Run it. Three plus one plus six plus five plus fourteen plus four plus six plus eight is forty-seven hours, against a forty-eight hour promise. One hour of slack. That is the entire margin in a two-day delivery commitment, and it is thinner than most people believe. Now assume the first attempt fails because nobody was home. Add roughly twenty-four hours for the re-attempt cycle. You are at seventy-one hours. Assume the second attempt fails too, another twenty-four hours, ninety-five hours. A parcel promised in two days lands closer to four, and every hour of that gap generates a contact to your customer service team.

Now put the union position inside that model. Assume a chunk of your residential volume is being completed on a tender-and-acceptance basis rather than on a fixed route. Assume the tender-to-acceptance handshake costs two hours in normal conditions. Assume that number grows when drivers become uncertain about the rules they are operating under, and assume it goes from two hours to five. The model I just built does not go from forty-seven to fifty; it has forty-five hours of fixed cost in it and only two hours of elastic cost, so a three hour growth anywhere takes it to fifty hours.

A fifty hour actual against a forty-eight hour promise means the promise is broken for every parcel in the system, not just the late ones. You then have two choices, and only two: move your effective cut-off earlier so your own three-hour pick cycle plus buffer absorbs it, or stop publishing forty-eight hours in the affected postal codes. There is no third option where you keep the promise by asking nicely.

Let me put money on it, again with every assumption visible. Assume you ship 25,000 parcels a week into metro areas that could be affected; substitute your own figure, the arithmetic scales linearly. Assume your first attempt success rate is ninety-five per cent. That is 1,250 failed attempts every week. Assume each failed attempt costs you nine dollars in extra handling, re-run and redelivery. Assume one in three failed attempts generates a customer contact, and assume each contact costs you six dollars to resolve, including the labour and the goodwill credit you end up eating. One thousand two hundred and fifty times nine is 11,250 dollars. Roughly four hundred and seventeen contacts at six dollars is about 2,500 dollars. Call it 13,750 dollars a week to fail at your normal rate.

Assume the tension degrades first attempt performance by three points, to ninety-two per cent. Now two thousand attempts fail. Two thousand times nine is 18,000 dollars. Contacts go to roughly six hundred and sixty-seven, call it 4,000 dollars. Twenty-two thousand dollars a week. The delta between a normal week and a tense week is about 8,250 dollars a week on 25,000 parcels, before you count cancelled orders, negative reviews or a retail partner charging you back for a missed appointment.

Run the same numbers through peak. Assume your peak week runs at 2.2 times baseline, so 55,000 parcels. Assume you hold the ninety-five per cent success rate, which is generous, because competing for flexible capacity at peak is exactly when it costs the most. Failures are 2,750 a week, which is 24,750 dollars plus about 5,500 dollars in contacts, around 30,250 dollars in a peak week. Assume an eight week peak window averaging 1.8 times baseline. That is about 198,000 dollars of failure cost at a held service level, and close to 317,000 dollars if the same three point degradation shows up during those weeks. The difference is roughly 119,000 dollars, which is what lobbying nobody can influence ends up costing you if you have no plan.

Now the part people skip: the failure attempts are not evenly distributed, they are concentrated in the same handful of failure modes. Missing unit numbers on apartment addresses. Business addresses where the receiving desk closes at 17:00 and the van arrives at 17:40. Signature requirements nobody asked for. Addresses where the geocode drops the pin three hundred metres from the door because the street was renumbered. Here is a second small model. Assume 1.5 per cent of your address records are missing a unit number, which on 25,000 parcels is 375 pieces a week running into a first attempt that cannot possibly succeed. Assume cleaning that list down to 0.3 per cent, which is 75 pieces. You have just removed 300 failures a week, worth about 2,700 dollars a week or roughly 108,000 dollars across that eight week peak window, and you did it with a form field and a validation script rather than with a contract renegotiation.

The counterweight is that over-reacting costs real money too. Assume you decide today to move twenty per cent of your volume to an alternative carrier as insurance. Assume the alternative is twelve per cent more expensive per piece on your average basket. Assume your average all-in piece cost is eight dollars. Moving five thousand parcels a week at a ninety-six cent penalty is 4,800 dollars a week, call it 240,000 dollars over fifty weeks. Insurance is not free and it is not a moral good; it is a hedge with a price tag, and the right amount of it depends on how much of your book is genuinely time critical. A shipper moving industrial spares on a service level agreement needs that hedge this month. A shipper moving replenishment stock into a warehouse with six weeks of cover does not.

Peak capacity deserves to be dealt with on its own, because it is the single input nobody can buy at the last minute. In parcel, capacity is reserved rather than purchased, and what gets reserved is a pickup commitment, a sort allocation at a named facility and a certain number of van departures out of a named station. Assume your peak planning cycle normally opens eight weeks before your first surge week. Assume integrating a second carrier takes six weeks from signed contract to a clean, tested label service. Assume the work on your side, meaning routing logic, rate shopping and service mapping for the affected postal codes, takes another two.

That is sixteen weeks of lead time against a dispute nobody knows will still be open in sixteen weeks. Which is another way of saying the decision to build a second channel is not really a decision about this dispute at all. It is a decision about whether you rebuild your network on your own timetable or on somebody else's. The arithmetic above gives you the downside. Sixteen weeks gives you the calendar. Put them side by side and the question tends to stop being interesting.

There is also a version of this where nothing happens at all, and it deserves the same weight as the alarming one. Assume the argument gets settled quietly, or assume it grinds on for a year without touching a single van departure. In that world every dollar spent on contingency looks wasted in the quarterly review. That is simply the arithmetic of insurance, and it is the reason I would rather you built a second channel carrying twenty per cent of your volume permanently, priced and tested, than a panic plan sitting in a drawer. Permanent diversification is a cost you control. A drawer plan is a cost you pay as well, and it does not work on the night you need it.

What none of us can control is the legal outcome, and it would be dishonest to pretend otherwise. The union says the 2023 agreement bars this. UPS is described as running these arrangements deliberately. Somebody will eventually rule on it, and that will take far longer than your peak season. What you control is the number of variables in your own network that depend on someone else's argument being settled first. That is the number to drive down over the next few weeks, while there is still slack in the calendar to do it without heroics.

Start with visibility, because you cannot manage a clock you cannot read. Pull your own tracking data and look at the scan events between pickup and final delivery for the last thirty days. What you are looking for is a handoff you did not expect: a scan that names a facility you have never heard of, a tracking status that changes vocabulary halfway through, a final-mile scan that comes from a different operator identity than the one on your label. Write down how often it happens, split by destination postal code. If it is two per cent, note two per cent. If it is twenty per cent in three specific markets, that is the map of your exposure, and it took you one afternoon to build.

Once you have that map, turn it into a routing decision rather than a report. Take the postal codes where the exposure concentrates and build one simple rule: every parcel going to those codes is quoted at the honest transit time, cannot be labelled until the address passes validation, and carries a different delivery promise from everything else. Assume this touches fifteen per cent of your volume. Assume half of that fifteen per cent can absorb a longer stated promise without any measurable effect on conversion. The other half is where you spend money, on a faster service, on inventory pre-positioning, or on access points. Building the rule is one day of work. Deciding which half is which needs your own data, and nobody can do that part for you.

Then get the cut-off question answered in writing, and I mean your specific account's cut-off, not the published one. Ask what time the last scan has to happen, whether the clock starts at pickup scan or at station arrival, what the rule is on weekends, and what the rule is the week of a public holiday. Published network cut-offs and account cut-offs diverge, and the divergence is usually twenty or thirty minutes, which is exactly the size of your entire delivery margin.

Alongside that, decide what you are actually promising. Most shippers publish a delivery range that somebody in marketing picked because it looked good on a checkout page. Pull every published promise, every email template, every marketplace listing, and check it against the hour model above with your real volumes. If your model says fifty hours and your page says two days, you are already lying to customers and you have been for months. Change the page, or change the model. Both are legitimate; pretending is not.

The alternative routes are worth pricing now, while nobody is desperate. A regional carrier that covers four states well may take twenty per cent of your volume in those four states at a price you can live with, and having it live and tested in September is worth ten times having it theoretically available in late November. Access points and lockers are another option for the slice of your book that keeps failing at the doorstep, and they convert a variable outcome into a fixed one. Commercial-address redirection works for part of your residential base, though you need customer consent and a clean way to explain it. Pre-positioning inventory into a second node is the heaviest option and the most reliable one, because splitting your fulfilment footprint means a disruption in one metro never touches the whole book.

Watch the traps on the way through. Read your own service guarantee before you rely on it, because most of these documents contain language about events outside the carrier's control, and a labour dispute lands squarely in that territory. Do not assume one rate card covers both networks; surcharges, accessorials and delivery-area adjustments are billed per piece per service, and a blended assumption in your budget will be wrong. Do not accept verbal confirmation of a pickup time from a driver; get it from your account team in an email. Do not move volume based on a headline and then discover your dimensional data does not match the new carrier's billing.

And then there is the thing everybody forgets until it is eleven at night. Whatever happens in this dispute, the parcels currently in your building still have to get out tonight. Somebody has to own the exception queue, somebody has to have authority to spend money to fix one, and that somebody should not be a person who needs three approvals to authorise a re-delivery. Give one named person a threshold and a budget, and let everybody else sleep.

↑ Back to the key points

— By Nadia Pryce