Teardown: The Notice That Arrives Too Late

The last teardown used a filing from the SEC. This one uses something even easier to check: terms a vendor publishes on its own website, which anyone can read and which govern every customer who signs an order.

I am looking at Datadog's Master Subscription Agreement, last updated 27 July 2023 and published at datadoghq.com/legal/msa. No client of mine is involved. Datadog is a serious company with serious counsel, and nothing below is drafted to trap anybody. That is rather the point.

Everything that follows is in Section 6, which runs to three short subsections and about four hundred words.

The three dates

Section 6.3 tells you how to get out. "If Customer does not want to auto-renew, Customer must submit a request to their Datadog Customer Success Manager at least 15 days before the expiration of the then current Order Term."

So your window to say no closes fifteen days before the term ends.

Section 6.1 tells you when you will hear about a price change. For orders of twelve months or longer, notice goes out "on or around 30 days' prior to the start of the Renewal Order Term," and "such notice will include notice of pricing adjustments, if any." For shorter orders, notice goes out "on or around 15 days' prior to the start of the Renewal Order Term solely in connection with pricing adjustments."

Fifteen days before the renewal starts is the same moment as fifteen days before the current term expires. The two periods are contiguous.

So on an order shorter than twelve months, the email telling you the new price is scheduled to arrive on the day your window to refuse it closes. And it is "on or around," which means it can arrive after.

What silence means

The last sentence of 6.1 is the one that makes the timing matter: "Failure of Customer to take a non-renewal action in a timely manner, shall be deemed to constitute consent to the applicable fee increase."

Not deemed to constitute renewal, which would be ordinary. Deemed to constitute consent to the increase specifically. If the number reaches you on the last day and you do not act, the agreement has already characterised your silence.

The cap that does not cover you

Section 6.2 limits how far the price can climb. Datadog "may increase the pricing in effect at the end of the applicable Order Term for the Services by up to the greater of 9% or" the twelve month change in CPI.

That is a real protection and a reasonable one. It applies, in the clause's own words, "for Orders that have a duration of 12-months or longer."

Then, further down the same subsection: "any renewal for Orders that have a duration of less than 12-months, or where Customer seeks a reduction in the subscription quantity and/or the Order Term will result in re-pricing at renewal and will not be subject to the limits herein."

So the shorter order gets half the notice and none of the cap. The two disadvantages are in different subsections and neither announces the other.

The same line, a third time

Once you have noticed the twelve month boundary, it turns up again outside Section 6 entirely. Datadog may modify the agreement by posting a revised version, and those modifications "become effective as of the first day of the calendar month following the month in which they were first posted; provided, however, that if an Order specifies a fixed term of 12 months or longer, the modifications will instead be effective immediately upon the start" of the next renewal.

A twelve month customer gets changed terms at a renewal boundary they can see coming. A shorter one can get them mid term, the month after they are posted.

Three separate mechanisms, three separate sections, one consistent line. None of it is hidden. It is simply never collected in one place, including by the agreement itself.

Why a linear read misses this

Read 6.1 on its own and it is a vendor committing to tell you about price changes, which reads as a courtesy. Read 6.3 on its own and fifteen days is a normal notice period. Read 6.2 on its own and a nine percent collar is better than most.

Every clause is defensible. The problem lives in the arithmetic between them, and arithmetic is not what a clause by clause review produces. It is what a calendar produces. Put the three dates on one and the shape is obvious in about ten seconds, which is roughly nine seconds longer than it took to read any one of the clauses and considerably less time than it takes to read all three in order and form a view of each as you go.

What I would actually do about it

Nothing adversarial. This is not a clause worth burning negotiating capital on in most deals, and the fix is unglamorous.

Take a twelve month term rather than a shorter one where the commercial difference is small, because the twelve month line is where all three protections switch on. If the shorter term is genuinely better for the business, then put the non-renewal deadline in a calendar the day the order is signed, counted backwards from the expiration date rather than forwards from anything, and set the reminder before the notice is due rather than when it is. The vendor's email is not the trigger. Your own date is.

And read the renewal section first, before the indemnity and the cap, on any agreement that auto-renews. It takes two minutes and it is the section most likely to cost money on a contract that never goes wrong.

Why this is the second teardown and not the tenth

Every finding here came from reading three subsections against each other and against a calendar. The first teardown found the same shape in a different place: an indemnity capped by the clause that followed it, a damages waiver that failed to name the section it was supposed to protect.

That is the work that does not get faster by reading harder, and it is what I built Midly to do. Not to summarise an agreement, which is easy and nearly useless, but to hold the whole instrument at once and price the cross references, so the clause that changes the answer does not sit three subsections away from the clause it changes.