Liability caps are among the most negotiated provisions in commercial contracts, and among the most poorly benchmarked. Most attorneys know what they want from a liability cap, but very few have visibility into what counterparties are actually accepting across the market right now.
Over the past several months, we analyzed anonymized liability cap data across a set of reviewed agreements. The dataset runs to roughly 800 commercial contracts, heavily weighted toward SaaS and professional services agreements, with meaningful coverage in IP licensing and M&A ancillary agreements. This post shares what we found, and what it means for how you should read and negotiate liability cap language.
A note on the data: this is internal review data, not a formal academic study. We are not claiming statistical rigor. What we are sharing is a working practitioner's view of the market based on actual contracts reviewed on behalf of actual legal teams. Treat it accordingly.
Where SaaS liability caps cluster
The most common SaaS liability cap formulation, across the contracts we reviewed, ties the cap to fees paid in a trailing period, almost always twelve months. The clustering around the "fees paid in the prior twelve months" formulation is strong: roughly 65 to 70 percent of SaaS agreements in our review data use some variant of this structure.
What varies considerably is the multiplier. Standard vendor paper almost universally caps at 1x twelve-month fees. Customer-negotiated positions that succeed in getting movement typically reach 2x or 3x. Movement above 3x is rare outside of agreements where the vendor is handling sensitive data, where data breach exposure has pushed customers to seek higher caps on a specific carve-out basis.
Agreements valued at $50,000 or less annually almost never produce negotiated cap changes in our data. The economics of negotiating a liability cap on a $25,000 SaaS contract rarely pencil out for the customer's legal spend. Agreements at $500,000 and above show much higher rates of negotiated cap language, with customers more frequently achieving 2x and occasionally 3x multiples.
The implication for practice: if you are a GC reviewing a SaaS MSA at $100,000 annual contract value, a 1x cap is market-standard. Asking for 2x is a reasonable negotiating position. Asking for 5x unlimited will produce friction without proportionate benefit in most contexts.
Uncapped carve-outs: the market is moving
The more interesting finding is in uncapped carve-outs. The proportion of SaaS agreements that contain at least one unlimited or separately-capped carve-out has grown notably over the past eighteen months of our review data. The carve-outs appearing most often are: IP infringement indemnification, data breach obligations, and gross negligence or willful misconduct.
IP infringement carve-outs appearing uncapped in vendor paper are not surprising; vendors do not want a contracted liability cap to limit their exposure for indemnifying the customer against patent or copyright claims. That is a known and generally accepted market position. What has changed is the scope of the data breach carve-out.
Several years ago, data breach carve-outs in vendor contracts were relatively rare and, when present, usually capped at a fixed dollar amount. In the contracts we have reviewed more recently, a significant proportion of vendor agreements include data breach obligations that are either uncapped entirely or capped at a separate, higher multiple than the general cap. We have seen provisions capping data breach exposure at 3x annual fees when the general cap is 1x, and provisions that remove any dollar ceiling on data breach liability entirely.
For customers in highly regulated industries, these uncapped data breach provisions are sometimes the most important risk item in the contract, not the general liability cap. We flag them specifically and distinctly in our output because they often get lost in the surrounding cap language.
What GCs actually renegotiate
We also pay attention to what redline activity looks like on liability cap provisions: where do customers mark up the vendor paper, and where do those markups succeed?
The most common customer redline on a liability cap provision is not moving the multiplier. It is adding a mutual cap structure where the vendor's draft had imposed a cap only on its own liability, leaving the customer's liability uncapped or governed only by general law. This asymmetric structure, which appears in a meaningful portion of vendor-drafted agreements, is the first thing most experienced GCs mark up, and it is also among the most consistently successful redlines. Vendors accept mutual caps more readily than they accept higher cap multiples.
The second most common customer redline is tightening the exclusions from the cap. Many vendor agreements exclude consequential damages from the cap along with or instead of setting a dollar ceiling on them. Customers who understand the consequential damages issue mark up those exclusions, because in many SaaS arrangements the damages from a service failure are predominantly consequential. A liability cap that excludes consequential damages from coverage is, for many breach scenarios, functionally uncapped from the customer's exposure perspective.
A note on mutual vs. asymmetric structures
We want to be clear about something: we are not saying that asymmetric liability caps are uniformly inappropriate. Vendor paper that caps the vendor's liability at 1x annual fees while leaving customer liability uncapped reflects a considered position from the vendor's risk perspective, and in many straightforward services agreements that position is commercially reasonable. The vendor is providing a defined service for defined fees; capping its own exposure at those fees is defensible.
The problem arises when the asymmetry is not visible. When a GC is reviewing fifty vendor agreements and each one has slightly different cap language, the asymmetric structures tend to get lost in the volume. Our benchmark data is useful here not because it tells you that asymmetric caps are bad, but because it tells you what is normal. When a vendor's paper deviates significantly from normal, in either direction, you should want to know that.
Contract type makes a significant difference
The benchmarks differ substantially across contract types in our data. Professional services agreements, where the vendor is delivering services rather than software, tend to have higher negotiated caps on average and more varied multiplier structures. IP licensing agreements are the most variable category: cap structures in IP licensing range from very low to uncapped depending on deal specifics and negotiating positions.
M&A ancillary agreements, such as transition services agreements and standalone IP licenses associated with deals, are particularly heterogeneous. These agreements are often drafted under time pressure with less standardization than a commercial SaaS MSA, and the liability cap provisions reflect that. They also tend to carry higher actual risk exposure than routine vendor agreements, which makes careful clause reading more important, not less, even when deal pressure suggests moving fast.
How to use this for your practice
Benchmark data on liability caps is most useful as a calibration tool, not a negotiating script. Knowing that 1x annual fees is the standard SaaS vendor position tells you when a vendor offering 2x in their standard paper may be signaling flexibility elsewhere, or when a customer asking for 10x is starting from an uninformed position.
The other practical application is internal consistency. Many in-house legal teams have inconsistent cap positions across their vendor portfolio, not because of deliberate negotiating decisions but because each contract was reviewed at a different time by different people working from different baselines. Running a batch review of your vendor MSA stack against a consistent cap standard reveals that inconsistency and lets you decide which agreements are worth reopening on their next renewal cycle.
That is precisely the kind of review Clausebeam was built to support: not replacing the attorney judgment about what to accept, but ensuring that the information those judgments should be based on is actually visible.
