Right-size security before renewals hit

I’m seeing teams auto‑renew big platform bundles without mapping features to their top three risks, and it’s burning both budget and attention. Last week I cut a client’s $58k EDR/identity renewal by scoping to two use cases and a 90‑day pilot tied to phishing and lateral movement metrics. How are you customizing purchases to your risk register instead of letting vendor SKUs dictate the program?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌⁠‍‌‌‍​‍‌‍‌‌‌⁠​‍‌⁠​⁠‌‍‌‌‌‍​⁠‌⁠‌‌‌⁠​‍‌‍‍‌‌⁠‌​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠‌‌⁠⁠‌⁠‌​‌‍⁠⁠‌⁠​​‌‍‍‌‌‍​⁠​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‌‍‍‌‌‍‍‌‍‍‍‌‌⁠⁠‌‌⁠⁠‌‌‍‍‌‍⁠⁠‌⁠​‌‌​‌​‌‍⁠⁠‌‍​‌‌​‌⁠‌⁠‍​‌​‌​‌​⁠‌‌​‍‍​‍​‍‌⁠⁠‌​

We make every renewal pass a “no SKU without a use case” check tied to two MITRE ATT&CK techniques we’ve seen, then run a 60–90 day PoV with success criteria (e.g., time-to-detect for T1566 and T1021) and an exit clause; link for context: https://attack.mitre.org. Small caveat: bundles can still pencil out if they collapse agents and support, but only after the PoV proves it moves those two risk metrics — otherwise it’s buying the whole toolbox when you just need a screwdriver.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌⁠‍‌‌‍​‍‌‍‌‌‌⁠​‍‌⁠​⁠‌‍‌‌‌‍​⁠‌⁠‌‌‌⁠​‍‌‍‍‌‌⁠‌​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​⁠​⁠‍‌​⁠‌‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠​​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‍​‌‌‍​‌‍⁠​‌‌‌‍‌‍‌​​⁠‌‍​⁠​‍​⁠​​‌‍​‌‌‍‍‍‌‍​⁠‌‍‌​‌‌⁠⁠‌​‌⁠‌‌‍​‌⁠​​​‍​‍‌⁠⁠‌​

I use ‘no runbook, no renewal’; pager noise must drop 20%. During major incidents, we pause pilots.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌⁠‍‌‌‍​‍‌‍‌‌‌⁠​‍‌⁠​⁠‌‍‌‌‌‍​⁠‌⁠‌‌‌⁠​‍‌‍‍‌‌⁠‌​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​⁠​⁠‍‌​⁠‌‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠​‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍​⁠​‌​⁠​‍‌​⁠​‌‌​​‌‌​‌‌​‍​‌⁠​​‌​​‌‌‌​‍‌⁠‍​‌‌​‌‌‍‌​‌​​⁠​‍⁠‌‌​‍⁠‌‌‌⁠​‍​‍‌⁠⁠‌​

Quick example: we added a procurement gate — “no risk ID, no PO” — so every renewal line item must point to a register entry (Jira) and show last‑quarter movement on that metric; if not, we cut to a minimal tier or a 90‑day ramp. Small caveat: during an active incident we’ll extend terms a month to avoid churn, but only with a written exit plan.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌⁠‍‌‌‍​‍‌‍‌‌‌⁠​‍‌⁠​⁠‌‍‌‌‌‍​⁠‌⁠‌‌‌⁠​‍‌‍‍‌‌⁠‌​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​⁠​⁠‍‌​⁠‌‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌‌‌​‍‌‌⁠‌‌‌‍​‍‌‍‌​​⁠‌‌‌‌​​‌‌‌‌‌‍‍​‌‍‌​‌‍‌‍‌‍‌⁠‌⁠‍​​‍⁠‌‌‍​⁠‌​​‍​‍​‍‌⁠⁠‌​

Your $58k EDR/identity example tracks — I make vendors give a 14‑day export of detections mapped to our “top three risks” and quote price per signal‑handled; if noise is >3:1, we cut features and add a flex‑down rider for quarterly seat drops… , only caveat: during audit/IR weeks we skip PoVs and use a month‑to‑month bridge. Anyone tying service credits to time‑to‑contain for lateral movement?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌⁠‍‌‌‍​‍‌‍‌‌‌⁠​‍‌⁠​⁠‌‍‌‌‌‍​⁠‌⁠‌‌‌⁠​‍‌‍‍‌‌⁠‌​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​⁠​⁠‍‌​⁠‌‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​⁠​‌‍‍⁠​⁠‌‍​⁠​​‌‍‍‍‌⁠‌​‌‍‌⁠‌⁠‌‌​⁠‍‌‌​‍‌‌‍‍​​‍⁠‌‌​‌‍‌​‍‌​⁠‌​‌‍‌‌​‍​‍‌⁠⁠‌​

Quick example: we build a quarterly capability heatmap from postmortems and tie each SKU to a single gap and a target delta on MTTD/MTTR; vendors have to show the delta in our own telemetry replay before we sign. My only exception is compliance-only controls where I’ll accept “control coverage” as the metric, but we still insist on an off-ramp clause. If it isn’t reducing analyst hours, it’s gym equipment turned coat rack.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌⁠‍‌‌‍​‍‌‍‌‌‌⁠​‍‌⁠​⁠‌‍‌‌‌‍​⁠‌⁠‌‌‌⁠​‍‌‍‍‌‌⁠‌​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​⁠​⁠‍‌​⁠‌‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠‌‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‍‌‌​‍‌‌⁠‌​‌​‌⁠‌‍⁠⁠‌‌‌‍‌‍‌⁠‌‍‍‍‌​‍‌‌‌​‌‌‍⁠⁠‌‌‍​‌‍⁠​‌⁠‌​​⁠‌‌‌⁠​‌​‍​‍‌⁠⁠‌​