Overage charges and unused SKUs cost more than the gap between a good discount and a great one, because the discount applied to committed usage while the waste billed on demand
The percentage is negotiated once and applies to the spend you planned. The mechanics apply to the spend you did not, every month, for the whole term.
Prepared by Redress Compliance · August 18, 2026 · Datadog negotiations. 12 to 15 renewals advised, 2024 to 2025.
Executive summary
On demand overage ran 15 to 30 percent of total Datadog spend in estates that sized commits once and never revisited them. Usage above commit bills at materially higher rates than committed usage.
Five to eight SKUs were active per estate, and one or two were typically unused beyond a pilot that never ended. Every product line meters separately.
Log spend fell 20 to 40 percent where indexing and retention tiers were redesigned before renewal. A discount on an ungoverned log pipeline is a discount on noise.
Structure compounds and percentages do not. Commit sizing, quarterly adjustment rights and true forward treatment outlast any headline rate.
Why does Datadog spend outgrow the infrastructure?
Because every product line meters separately, and adoption inside engineering adds meters without procurement ever seeing a decision. The full catalog sits on the Datadog pricing page.
- Per host meters: infrastructure and application performance bill on monitored hosts, including autoscaling ghosts.
- Volume meters: logs and network bill on ingested and indexed data.
- Event meters: synthetics, real user monitoring and pipeline visibility bill per test, session or event.
Each meter is rational alone
Together they compound, and the bill becomes the sum of every team's enthusiasm. Procurement discipline starts with an inventory of which meters run and who owns each.
How does on demand overage work against the buyer?
Usage above committed levels bills at on demand rates that run materially above committed rates. An undersized commit therefore converts growth into premium priced overage, documented in Datadog's billing documentation.
- Commits are sized once: usage grows quarterly while commitments reset annually.
- Autoscaling inflates host counts: short lived hosts count against high water marks.
- Nobody owns the meter: engineering adds usage, finance sees the invoice, neither adjusts the commit.
The structural fix is to negotiate quarterly commit adjustments or true forward terms, so growth reprices at committed rates rather than on demand rates.
The Datadog renewal kit
The SKU inventory, the commit structure model and the log tier redesign that lands before the baseline is measured.
Get the brief →What 12 to 15 Datadog negotiations showed
Across roughly 12 to 15 Datadog negotiations advised in 2024 to 2025, overage and unused SKUs decided more spend than the negotiated discount. Three patterns recur.
- On demand overage ran 15 to 30 percent of total Datadog spend in estates that sized commits once and never revisited them.
- Five to eight SKUs were active per estate, and one or two were typically unused beyond a pilot that never ended.
- Log spend fell 20 to 40 percent where indexing and retention tiers were redesigned before renewal.
In roughly 9 of the 12 plus estates benchmarked, the waste billed at on demand rates while the discount applied only to committed usage.
- Your quote benchmarked against 500,000+ real closed deals, adjusted for deal size, region and industry
- Every risky clause flagged with the exact quote, the page, and the replacement language to send back
- Counter emails drafted in your voice, concessions tracked, live coaching on the negotiation call
How do you govern the log line before renewal?
Log cost splits across ingestion, indexing and retention, per Datadog's log management documentation, and the indexing decision dominates. Index what gets queried and archive the rest.
| Lever | What it does | Typical impact |
|---|---|---|
| Exclusion filters | Drop noise before indexing | Largest single saving |
| Flex or archive tiers | Cheap retention for compliance data | Cuts retention spend |
| Sampling | Index a fraction of high volume streams | Preserves signal at lower cost |
| Retention tuning | Shorter index windows per source | Compounds with the filters |
| Rehydration | Pull archived logs back when needed | Makes archiving safe |
Run the redesign before the baseline is measured
Estates that redesigned the tiers cut the line 20 to 40 percent. Doing it after the renewal baseline is set means paying a discounted rate on volume you were about to delete.
Watch the briefing · 6:54Datadog: Negotiate the Billing Mechanics, Not the Rate CardWhy host counting, custom metrics and ingestion tiers decide the invoice.
Which levers move a Datadog renewal, and in what order?
Cleanup and log redesign first, then commit structure, then the discount conversation last. The order is the whole method, because each stage changes the baseline the next one is measured against.
The sequence is the method
- Inventory the active SKUs and kill the pilots that never ended.
- Rebase commits on trailing usage with quarterly adjustment rights.
- Convert overage exposure into true forward terms.
- Trade term length only for rate protection across every meter.
Benchmark one workload on a competing platform with costed effort. The comparable exercise on the other major observability renewal sits in the Splunk Cloud guide, and the meter level detail in the Datadog negotiation guide. Marketplace routing against a cloud commitment is priced in the AWS EDP guide.
What the negotiations measured, 2024 to 2025
Two cuts of the engagement file frame where the money actually leaked.
In estates that sized the commitment once and never revisited it against quarterly usage growth.
Where indexing and retention were redesigned before the renewal baseline was measured rather than after.
Treat the ranges as benchmarks rather than promises. Your adoption curve sets your baseline, and the file describes what disciplined buyers achieved against the same vendor playbook.
Your first five moves
- Inventory the active SKUs against the product catalog and kill the pilots that never ended, since one or two of the five to eight active lines were typically dead weight.
- Redesign log indexing and retention before the baseline is measured, worth 20 to 40 percent of the log line and worth nothing at all if it happens afterwards.
- Rebase the commitments on trailing usage with quarterly adjustment rights, which is the direct answer to the 15 to 30 percent lost to on demand overage.
- Convert the remaining overage exposure into true forward terms, so growth reprices at committed rates instead of premium ones.
- Leave the headline percentage until last, and trade term only for rate protection across every meter. The Datadog practice guide carries the full lever sequence.
Frequently asked questions
Why does Datadog spend grow faster than infrastructure?
Because every product line meters separately and engineering teams add meters without procurement seeing a decision. Each meter is rational alone, and together they compound into the invoice.
How much is lost to on demand overage?
Between 15 and 30 percent of total spend in estates that sized commits once and never revisited them. Usage above commit bills at rates materially above committed rates.
What is the fix for overage?
Quarterly commit adjustments or true forward terms, so growth reprices at committed rates rather than on demand rates. Static annual commitments are the expensive default.
How many SKUs does a typical estate run?
Five to eight active product lines, of which one or two were typically unused beyond a pilot that never ended. The inventory is the cheapest lever on the file.
How much can log costs be cut?
Between 20 and 40 percent where indexing and retention tiers were redesigned. Exclusion filters that drop noise before indexing were the largest single saving.
Why redesign logs before the renewal?
Because a discount on an ungoverned log pipeline is a discount on noise. Redesigning after the baseline is measured means paying a better rate on volume you were about to delete.
Does the discount percentage matter at all?
It matters last. In roughly 9 of the 12 plus estates benchmarked, overage and unused SKUs cost more than the gap between a good discount and a great one.
What order should the levers run in?
SKU cleanup and log redesign first, then commit structure, then the discount conversation. Each stage changes the baseline the next one is measured against.
Do autoscaling hosts really cost money?
Yes. Short lived hosts count against high water marks on per host meters, so an elastic estate can bill as though it ran its peak continuously.
What alternatives move a Datadog quote?
Grafana, Elastic and native cloud monitoring, benchmarked on one workload with costed migration effort. The benchmark works because the growth model prizes committed expansion over list rate defense.
