In short: A 3PL agreement is a set of options with strike prices and notice periods, and the volume band, the cost model split, the notice periods and the rate review mechanics decide what you can do later. Outsourcing the building does not outsource the forecast, since the provider plans labour and space from numbers you supply, in their units, at their grain, across horizons running from weeks for labour to years for automation. A gainshare written on cost per case will deliver cost per case, which is the multitask problem Holmstrom and Milgrom set out in 1991, so the measured set has to cover the dimensions you care about. Priority in a week when the building is short is a commercial position fixed at signature rather than something operational urgency can buy on the day.
The invoice for November arrives in December with two lines nobody was expecting. Overflow storage at an external site, charged by the pallet week. A peak handling surcharge applied to every case above the contracted band, at a rate somebody agreed to eighteen months ago and has not looked at since. The combined figure is larger than the saving that justified the outsourcing decision in the first place.
Nobody in planning knew the band existed. The contract is in a folder owned by procurement, the volume forecast that the provider planned against was sent by a commercial analyst in July, and the person who could have reconciled the two left in the spring.
What the contract actually commits
A 3PL agreement is a set of options with strike prices and notice periods, and it repays being read that way rather than as a rate card. Four elements decide what you can and cannot do later.
The volume band. Rates are quoted against an assumed annual volume, and they hold inside a band around it. Below the band, a minimum volume guarantee or a fixed management fee spreads across fewer units and your effective unit cost rises. Above it, either a renegotiated rate applies or a surcharge does, and the surcharge is almost always worse than the rate you would have negotiated in advance.
The cost model split. Storage is usually per pallet per week, sometimes with a free period after receipt. Handling is per case, per pallet or per order line, in and out. Value added services are priced per unit of work. A management fee sits on top. Each element has its own volume driver, so what the provider needs from you is several forecasts rather than one number.
Notice periods. Adding space, changing a shift pattern and exiting all carry their own notice. The exit notice matters most and gets read least.
Rate review mechanics. Indexation against a published labour cost index or a fuel index, with a review date and a cap if you negotiated one. A contract signed in a low inflation year with uncapped indexation is a different contract three years later.
Peak commitments and the reservation mechanism usually sit in a schedule rather than the main body, which is why they get missed at handover.
You still forecast, and a bad one lands on your invoice
The provider plans labour and space from the volume you give them. Under-forecast and they under-recruit, and the consequences arrive as agency premium, overflow storage, missed cut-offs and a service conversation where you have no position. Over-forecast and they recruit and hold space against it, and you pay through the minimum, the surcharge structure, or the next rate review.
The forecast they need differs from the demand forecast in three ways that make the handoff harder than it sounds.
It is denominated in their units. Pallets received, pallets stored, cases picked, order lines, cartons despatched. Converting a demand plan into those requires case pack, pallet configuration, order profile and channel mix, and that conversion is the piece no job description owns.
It is at their grain. Labour is rostered by day, so day-of-week shape matters as much as monthly volume. A month that is flat in total and lumpy by day costs more to staff than a month with the same total spread evenly.
It runs on their horizons, which are several. Labour is planned two to six weeks out. Space is committed a quarter or more ahead. Automation and racking are years.
The mix point deserves more attention than it gets. The same tonnage arriving as four thousand case picks rather than four hundred pallet moves is a completely different labour requirement, and a volume forecast in tonnes carries none of that. Where the channel mix is shifting, which it is in most consumer businesses, the work forecast diverges from the volume forecast steadily until the productivity figures start missing.
The annual Third-Party Logistics Study led by C. John Langley at Penn State has tracked what it calls the IT gap for well over a decade: shippers rate information capability as essential to the relationship and rate their satisfaction with what they get considerably lower. Most of what sits in that gap is exactly this, the absence of a shared plan in units both sides can act on.
Shared or dedicated
Shared, or multi-user, space means you pay broadly for what you use, capacity flexes inside limits, the provider absorbs peaks by moving labour between clients, and unit costs are lower because the fixed base spreads across several businesses. What you give up is control of priority, layout and the workforce, and you share the building with everyone else's peak.
Dedicated means you pay for the facility whether you fill it or not, you set the layout and the priority, you can invest in automation against a payback horizon because the volume is yours, and you carry the fixed cost through your low season with no relief.
The economics turn on two ratios. The first is your own peak to trough, since a business with a flat profile wastes very little in a dedicated facility and a business with a three to one seasonal swing wastes a great deal. The second is the correlation between your peak and the other clients' peaks in a shared building. If your peak is counter-seasonal to theirs, shared capacity is genuinely cheap for you and genuinely valuable to them. If your peak lands in the same eight weeks as everyone else's, which in consumer goods it does, the flexibility that was sold to you is exactly the thing that will not be available on the day you need it.
Most mature arrangements end up hybrid: a dedicated core sized near the trough, with shared or overflow capacity carrying the seasonal top. Sizing the core is the real decision, and it has the same shape as any capacity decision under uncertainty. Size it at the mean and you pay overflow rates for half the year.
Peak capacity gets reserved months before the peak
The physical constraints are unhelpfully sequenced. Peak labour is recruited and trained weeks ahead of the volume, space is committed a quarter or more ahead, and anything mechanised is years ahead. Meanwhile most of the market wants the same eight weeks.
That makes peak capacity an options problem rather than a purchasing one. You can pay a reservation fee now for the right to a volume later, or take spot capacity at whatever the market charges when you need it. Barnes-Schuster, Bassok and Anupindi analysed this contract structure in Manufacturing and Service Operations Management in 2002, and the arithmetic transfers directly: the reservation is worth buying when the spread between the reserved rate and the expected spot rate, multiplied by the probability you exercise, exceeds the fee.
The practical version is to commit the base, take options on the middle, and leave the tail deliberately exposed, having decided in advance what happens if the tail arrives. Turning away volume, moving a promotional window, shipping from a different node, or accepting a service drop on a named customer segment are all legitimate answers. Choosing between them in October is a plan. Choosing on the Tuesday of peak week is a scramble that costs more than the option would have.
Give the provider a range rather than a number for peak, with the shape by week and by day. A single figure gets planned to exactly, and the conversation about what happens at the high end never occurs.
Gainshare, open book, and what they cost in transparency
Three commercial structures show up, and each one puts the incentive somewhere different.
A fixed rate closed book contract gives you price certainty and the provider every incentive to reduce cost, since all of the saving is theirs once the rate is set. It also gives them every incentive to reduce cost in ways you would not have chosen.
Open book means you see the cost base and pay cost plus an agreed margin. Their incentive to reduce cost weakens considerably, because the margin is calculated on the cost.
Gainshare sets a baseline and splits the savings against it. The design problem is entirely in the baseline. Set it from last year's actuals and you reward a bad last year. Reset it annually and you remove any incentive to invest in something with a payback longer than the reset period, which is most of the things worth doing.
The failure that recurs across all three has a clean theoretical description. Holmstrom and Milgrom, in the Journal of Law, Economics and Organization in 1991, set out the multitask principal-agent problem: when one dimension of performance is measured and rewarded and other dimensions are not, effort moves toward the measured one. A gainshare written on cost per case will deliver cost per case, and part of the delivery will come out of pick accuracy, damage rates, receiving discipline and the willingness to absorb your exceptions, none of which appear in the formula.
The correction is a measured set covering the dimensions you actually care about, with the gainshare gated on the unmeasured ones being maintained rather than simply hoped for. Open book then demands real access: labour hours by activity, agency premium separately identified, pallet positions occupied by day, and the productivity assumptions behind the establishment. Without the ability to audit those, cost plus is a margin applied to a figure the other party computes.
Exit risk grows with tenure
Over a long relationship the provider accumulates three kinds of relationship-specific investment. Knowledge, meaning your order profile, your customer exceptions, your seasonality and the workarounds. Configuration, meaning the warehouse system setup, label formats, EDI maps and the reports that have accreted over five years. Sometimes physical assets, meaning racking, handling equipment or an automated system funded across the contract term.
Klein, Crawford and Alchian described the general mechanism in the Journal of Law and Economics in 1978, and Williamson developed it at length in 1985: once relationship-specific investment exists, whoever holds it acquires bargaining power at renewal that they did not have at the original tender. The rate you get in year six reflects how hard you would find it to leave.
Four things reduce the exposure without pretending it away. Own your data contractually, including master data, order history and configuration documentation, extractable in a stated format on a stated timeline. Agree a transfer value for physical assets at exit while you are still negotiating from strength. Know your transition window, since moving a warehouse takes a quarter of planning and can only happen in the low season, which means one opportunity a year. And keep an alternative credible, whether that is a second provider carrying part of the volume or a site you could stand up, pricing the unit cost difference as insurance.
The trade underneath is real and worth stating plainly. A long relationship produces better service, because the provider learns your business in ways no transition document transfers. It also produces worse commercial terms, because your alternative gets less credible every year. Re-tendering every three years buys price and costs service, and there is no arrangement that gets both.
The limit
A 3PL optimises across its whole client base. In a week when the building is forty people short and three clients all want their volume out, the allocation gets decided by contract terms, by revenue at risk, and by the relationship, in roughly that order. Your operational urgency is an input to that decision and a weak one.
Your priority in that week is a commercial position, and it was fixed at signature. The terms that create it are specific. A service level with a financial consequence large enough to change the provider's arithmetic. A dedicated labour pool with a stated headcount rather than a best endeavours clause. Contractual priority on named order types, with a written sequence for what ships first when capacity is short. A named escalation route with a person, a timeline and an authority level attached.
A vague service level with a token penalty buys nothing on the day the building is full. And one position cannot be bought at any price, which is priority ahead of a client who is larger, longer-tenured and paying more. If your volume is a small share of that building, plan on being third in the queue and hold the buffer that implies.
There is a second limit worth naming, which is that no forecast improvement solves a structural shortage in the market. When regional industrial space and warehouse labour are both tight, a better forecast buys a better position in a queue rather than more capacity, and the response has to be physical: a second location, an earlier build, or a range held further upstream. Where the goods sit inside the building is X5's subject, and the labour plan running against it is Y2's.
Take last year's invoices, split every line into storage, handling, value added and surcharge, and check which of those moved with volume and which moved with something nobody ever forecast.