MAC clauses in tech acquisitions: patterns across our 2024 matter index

How many carve-outs does a material adverse change definition need before it stops being a condition and becomes a decoration?

We do not have a normative answer. We do have 480 executed share purchase agreements in the technology sector, signed during 2024, sitting in the Recall indices of accounts that have consented to anonymised research use. This note describes what their MAC clauses look like. It is an observation across our own index, not a market study, and the caveats at the end are not a formality.

Method and sample

Recall indexes a firm’s matter history for search. For accounts that have opted into research use, our team can run structured queries across the index with party names, amounts, dates and identifiers removed at the tenant boundary before anything reaches us. We see clause text and metadata tags (sector, deal type, governing law, buyer or seller side), not the deal.

The sample: 480 SPAs tagged technology or software, executed between 1 January and 31 December 2024, from 41 accounts. Governing law: 46% English, 22% a continental European law (German, French, Dutch and Swiss predominating), 19% a US state, 13% other. Buy-side representation 54%, sell-side 46%. The sample is European mid-market by weight. It is not representative of the large-cap US market, and we do not claim it is.

MAC clauses were located and decomposed by Review’s clause classifier into definition, carve-outs, carve-backs (disproportionate-effect qualifiers) and usage (condition, warranty, termination right). A random 10% (48 agreements) were checked by hand by two of our lawyers. Classifier agreement with the human check was 94% on carve-out counts and 89% on carve-back presence; disagreements were mostly nested carve-outs counted as one or several.

What we saw

Carve-out counts. The median MAC definition carried nine carve-outs. The interquartile range was seven to twelve. The lowest was two; the highest, in a US-law deal, twenty-one. Counts were higher in agreements where the seller’s counsel held the pen on the definition, which is unsurprising but useful to see in numbers.

The usual list. General economic conditions, industry-wide conditions, changes in law, changes in accounting standards, acts of war or terrorism, natural disasters, announcement or pendency of the transaction, and actions required by the agreement or consented to by the buyer appeared in more than 80% of definitions. Failure to meet projections appeared in 71%, almost always paired with a carve-back preserving the underlying cause.

Pandemic language is leaving. An explicit pandemic or epidemic carve-out appeared in 38% of 2024 agreements. In the 2022 agreements in the same indices (a smaller set of 210), it was 79%. Where it remains, it is increasingly folded into a general “force majeure events” item rather than standing alone.

Cyber language is arriving. A carve-out or, more often, an express inclusion addressing cyber-attacks, data breaches or system outages appeared in 29% of definitions. Half of those treated a cyber event as not excluded, that is, as something that could constitute a MAC, which inverts the usual structure. In software targets this makes sense: a breach is the business.

Disproportionate-effect carve-backs. 84% of definitions with industry or economic carve-outs qualified them with a disproportionate-effect test. Of those, 61% measured disproportion against “other participants in the industry in which the target operates”; the remainder used narrower comparators, which favour buyers.

“Prospects”. Only 18% of definitions included the target’s prospects in what a MAC could affect. English-law agreements were least likely to (11%); US-law agreements most likely (34%).

Usage. The MAC was a closing condition in 77% of agreements, a termination right in 68%, and a warranty (no MAC since the accounts date) in 91%. Agreements with a MAC condition but no bring-down of the no-MAC warranty at completion were rare, at 6%.

What we did not see

We did not see any agreement in the sample where a MAC clause was invoked. Recall indexes documents, not disputes, and most MAC arguments never reach a document that would be indexed. So nothing here speaks to enforceability or to how any tribunal reads these words. The clause is drafted far more often than it is tested.

Caveats

• The sample is our accounts. It over-represents European mid-market firms with knowledge-management functions mature enough to connect a DMS to Recall. Treat every percentage as “in our index”, not “in the market”.
• Executed, not negotiated. We see the final text. We do not see which side proposed the pandemic carve-out or who conceded “prospects”. Negotiation dynamics are invisible here.
• Classifier error is non-zero. The 10% hand check found disagreements in about one agreement in twelve on at least one field. The headline percentages carry a few points of noise.
• Sector tagging is the firm’s. “Technology” means whatever the firm’s matter-opening form meant by it.
• Year-on-year comparisons use a smaller 2022 base, and the account set changed between the two years. The pandemic trend is clear enough to survive that; the cyber trend is suggestive, not established.

Why this is in Recall

Review uses observations like these to calibrate what counts as an unusual MAC definition for a given governing law and side, so that a Note flag (“two carve-outs is well below what we typically see in English-law tech SPAs”) carries a basis rather than an opinion. The underlying figures are visible in the trace view whenever such a flag appears.

Nine carve-outs, a disproportionate-effect test and no mention of prospects: if a 2024 English-law tech SPA crossed your desk, that is most likely what it said.