Outcome
By the end of this lesson, you will be able to calculate the effective discount of a commitment realistically, distinguish published discount from realized discount, and apply the break-even formula to decide whether a commitment is worth it.
| Prerequisites | M5.7.L1: Four commitment instruments |
| Time | 9 minutes |
1. Concept
The number on the tin, “save up to 40%”, is the best you could possibly do.
What you actually get depends on two things. Coverage is how much of your usage the commitment applies to. Utilisation is how much of the commitment you actually use. Fall short on either and the 40% shrinks.
Almost everybody overestimates what they got. The formula below is how you stop doing that.
EFFECTIVE DISCOUNT = coverage × (1 - (1 - d) / utilization)
WHERE: d = published discount, the headline rate (e.g., 40%) coverage = % of eligible usage that runs on the commitment utilization = % of committed hours actually used
WHY: each covered dollar of usage costs (1 - d) / utilization,because you also pay for the committed hours you leave idle.If either coverage or utilization is low, the published discount doesn’t translate to your bill. If utilization falls below 1 - d, the bracket goes negative and the commitment costs more than on-demand.
Walking through scenarios
SCENARIO A: Perfect: 1-yr RI on m5.large at 40% published discount Coverage: 100% (all m5.large hours covered) Utilization: 95% (instance running 95% of the time) Effective: 1.0 × (1 - 0.60 / 0.95) = 36.8%
Nearly the published discount; close to ideal.
SCENARIO B: Imperfect coverage: Same 1-yr RI at 40% Coverage: 80% (some m5.large hours go uncovered) Utilization: 95% (committed instances running 95%) Effective: 0.8 × (1 - 0.60 / 0.95) = 29.5%
Lost about 7 percentage points to incomplete coverage.
SCENARIO C: Imperfect utilization: Same 1-yr RI at 40% Coverage: 80% Utilization: 60% (instance idle 40% of the time) Effective: 0.8 × (1 - 0.60 / 0.60) = 0%
Exactly at break-even: the commitment saves nothing.
SCENARIO D: Bad: Coverage: 40% (committed for less than actual usage) Utilization: 50% (committed instances idle half the time) Effective: 0.4 × (1 - 0.60 / 0.50) = -8%
A net loss: utilization is below the 60% break-even, so the bill is 8% HIGHER than pure on-demand.The publishing-vs-realized gap is what separates “looks like savings” from “actually saves money on the bill.”
Break-even math
Every commitment has a break-even point: below it, you pay more than pure on-demand.
BREAK-EVEN UTILIZATION FORMULA: For a commitment with discount d: Required utilization > 1 - d (utilization = committed hours used / committed hours purchased)
EXAMPLES: d = 40% → required utilization > 60% d = 30% → required utilization > 70% d = 20% → required utilization > 80%
If your utilization is below the break-even, the commitment LOSESmoney vs pure on-demand. You committed for hours you didn't use,and you paid for them.Break-even is a utilization threshold, not a coverage one. Under-coverage never loses money: eligible hours the commitment does not cover simply run on-demand at rack rate. It is under-utilization that loses money. Bigger discounts have looser thresholds: a 40% discount only needs 60% utilization to break even; a 20% discount needs 80%.
Realistic targets by commitment length
COMMITMENT LENGTH SUGGESTED TARGETS──────────────────────────────────────────────────────────────────1-year commit Coverage target: > 80% Utilization target: > 95%
3-year commit Coverage target: > 70% Utilization target: > 95%
Note: 3-yr has more time for workload to change, so coverage target is looser.If you can’t confidently project these levels for the full commitment term, defer the commit or use a shorter term.
Realistic example
WORKLOAD: 100 m5.large instances, 24/7
OPTIONS analysis: Option 1: Pure on-demand: 100 × $0.096/hr × 24 × 365 = $84,096/year (about $841 per instance) Annual cost: $84,096
Option 2: 1-yr RI, all-upfront, 40% published discount: If 100% coverage + 95% utilization: Effective discount: 1.0 × (1 - 0.60 / 0.95) = 36.8% Cost: $84,096 × 0.60 / 0.95 = $53,113 Savings: $84,096 - $53,113 = $30,983 (36.8%)
If realistic coverage 90% + utilization 95%: Effective discount: 0.9 × (1 - 0.60 / 0.95) = 33.2% Savings: 0.9 × $30,983 = $27,885 (33.2%) Cost: $84,096 - $27,885 = $56,211
Difference between "ideal" and "realistic": ~$3.1K/year
Option 3: 1-yr Compute SP, all-upfront, 35% published discount: If 90% coverage + 95% utilization: Effective discount: 0.9 × (1 - 0.65 / 0.95) = 28.4% Savings: $84,096 × 0.9 × (1 - 0.65 / 0.95) = $23,901 Cost: $84,096 - $23,901 = $60,195
Lower discount but more flexible (SP applies to any compute)
Rates are illustrative; check current provider pricing.
DECISION FACTORS (both options are 1-year terms): Will these 100 instances stay m5.large in this region for the year? If yes, the RI (Option 2) saves ~$4K/year more. Might the instance family or region change? The SP (Option 3) follows the workload. How confident is the 95% utilization? Critical to verify for either option.How to improve coverage and utilization
COVERAGE IMPROVEMENTS: - Match RI specs closely to actual usage (don't buy for "projected growth"; buy for proven floor) - Mix RIs and Savings Plans for flexibility on the edge - Buy on the floor, not the peak - Re-evaluate quarterly
UTILIZATION IMPROVEMENTS: - Don't schedule the committed floor off (defeats the commit) - Run committed capacity 24/7 - Monitor exchange opportunities (AWS lets you exchange Convertible RIs; Standard RIs cannot be exchanged) - Use unused capacity for low-priority workloadsThe single biggest improvement lever for most orgs: don’t schedule the committed floor off. Schedules pause non-committed capacity; committed capacity runs continuously.
Modeling and tracking tools
TOOL PURPOSE──────────────────────────────────────────────────────────────────AWS Cost Explorer RI/SP 1-yr commitment recommendationsrecommendationsGCP Commitment planner 3-yr CUD modelingAzure Reservation calculator Reservation sizingProvider utilization reports RI/SP/CUD utilization per(AWS, Azure, GCP) commitmentZopNight commitment coverage Share of the bill oncard (Cost Reports) commitmentsCommitment automation Automated purchasing andservices (for example portfolio management forProsperOps or Spot Eco) complex estatesZopNight’s commitment coverage card on Cost Reports shows how much of the bill runs on commitments, Its commitment recommendations (an RI or reservation, a GCP CUD, or on AWS EC2 a 1-year Compute Savings Plan) appear only for a resource with 60+ days of history, at least 70% measured uptime and measurable CPU or memory use, and only when the 1-year commitment’s monthly price is below what the resource costs today. Read utilization from your provider’s RI/SP/CUD utilization report; specialized tools provide forecasting if your portfolio is complex.
Common discount-math mistakes
MISTAKE FIX──────────────────────────────────────────────────────────────────Quoting published discount as savings Apply the coverage and utilization formula to get the realistic number
Buying for projected growth Buy for proven floor; incremental commits for confirmed growth
Ignoring utilization Track it monthly; alerts on low utilization
3-yr commit without 3-yr confidence Use 1-yr; revisit annually
Locked into RI when SP would fit An RI cannot be converted to an SP: buy an SP and let the RIs expire
Scheduling the committed floor off Schedules don't apply to committed capacity (the schedule wastes the commitment)2. Demo
A real commitment portfolio review:
PRODUCTION FLEET commitment status (Acme Corp):
CURRENT STATE: 100 instances m5.large × 730 hr/mo = 73,000 hours Commitment: 60% via Savings Plan at 30% published discount Coverage: 60% (60 × 730 = 43,800 hours covered by SP) Utilization: 95% (covered hours actually used)
EFFECTIVE DISCOUNT: 0.6 × (1 - 0.70 / 0.95) = 15.8%
SAVINGS CURRENT: 73,000 hours × $0.096/hr = $7,008/month on-demand $7,008 × 15.8% = ~$1,107/month saved Annual: ~$13,300
OPPORTUNITY: If coverage rose to 80%: 0.8 × (1 - 0.70 / 0.95) = 21.1% effective discount Annual savings: ~$17,700 (+$4.4K vs current)
If coverage rose to 90%: 0.9 × (1 - 0.70 / 0.95) = 23.7% effective discount Annual savings: ~$19,900 (+$6.6K vs current)
DECISION ANALYSIS: Coverage at 80% requires covering another 20 instances × 730 hours = 14,600 hours (about 15,400 committed hours at 95% utilization). If usage shifts off m5.large in 6 months, that commitment may be stranded (still applies if SP, but lower coverage).
Confidence in 80% coverage over the next 12 months: HIGH (workload has been stable for 6+ months).
DECISION: Increase SP coverage to 80% next renewal cycle. Track utilization; if drops below 90%, re-evaluate.The math turns a vague “should we buy more SP?” into a specific decision. The 30% SP rate here is an example rate; L1’s table puts a 1-year Compute SP nearer 35%.
3. Hands-on (5 min)
Calculate effective discount for one of your team’s commitments:
COMMITMENT: __________ (RI / SP / CUD)PUBLISHED DISCOUNT: ____%
COVERAGE: ____% (% of eligible usage running on the commitment) How to find: cloud provider's commitment report
UTILIZATION: ____% (% of committed hours actually used) How to find: cloud provider's RI/SP utilization report
EFFECTIVE DISCOUNT: ____ coverage × (1 - (1 - ____ discount) / ____ utilization) = ____%
ANNUAL SAVINGS: Total spend on this workload class × Effective % = $______
COMPARED TO publicly quoted savings: $______ less than the headline number
IMPROVEMENT OPPORTUNITIES: □ Increase coverage (commit more) □ Increase utilization (don't schedule covered capacity off) □ Exchange to better-fit instrument □ Defer renewal if usage likely to change
BREAK-EVEN check: Required utilization: 1 - ____ (discount) = ____% Current utilization: ____% Above break-even? Yes / NoIf your utilization is below break-even, the commitment is losing money vs on-demand. Investigate.
Ask for it instead. The same task you just did in the console, asked in one sentence.
BEFORE A ZopNight account with one cloud connected. A year of history if you have it, and existing commitments visible in the bill.ASK "Show cost by provider over the last year, and any commitment recommendations we have."CHECK the schedule-first rule from L4. Commit to the floor you keep after scheduling, not the one you have today.Tools behind it: get_cost_by_provider (read, Cost), get_cost_trends (read, Cost), get_recommendations (read, Optimize). Connection details are in the MCP integration docs.
4. Knowledge check
Q1
Coverage 80%, utilization 95%, published discount 40%. Effective discount is:
A. 29.5% (0.8 × (1 - 0.60 / 0.95))
B. 40% (the published rate)
C. 76%
D. 38% (40% minus utilization loss)
Show answer
Correct: A. Each covered dollar saves 1 - 0.60 / 0.95 = 36.8%, because you also pay for the idle 5% of the commitment. Only 80% of usage is covered, so 0.8 × 36.8% = 29.5%. Quoting the published rate as savings is fiction.
Q2
1-year RI at 40% discount. Break-even utilization:
A. 40%
B. 100%
C. 71.4%, which is the inverse figure
D. 60%: calculation: 1 - 0.40 = 0.60
Show answer
Correct: D. Break-even utilization = 1 - discount. Below 60% of the committed hours used, the RI loses money vs pure on-demand. Bigger discounts have looser thresholds: 40% needs 60%, 20% needs 80%. It is a utilization threshold, not a coverage one: under-coverage never loses money.
Q3
The biggest improvement lever for effective discount:
A. Increasing coverage (commit more of the proven floor) AND increasing utilization (don’t schedule covered capacity off)
B. The cloud provider raising its published discount rate, which is outside of anyone’s own control entirely
C. Reducing the discount percentage
D. Negotiating a longer commitment term
Show answer
Correct: A. Both sit in the formula. Coverage is usually the bigger lever for under-committed customers; utilization is usually the bigger lever for over-committed ones.
5. Apply
Track coverage and utilization quarterly: coverage on the commitment coverage card in ZopNight’s Cost Reports or your provider’s coverage report, utilization in your provider’s RI/SP/CUD utilization report. Re-baseline commitments before renewal; verify break-even is well above the required threshold.
For purchase candidates, ZopNight’s commitment recommendations (on the Savings recommendations page) cover RIs and reservations, GCP CUDs and, on AWS EC2, Compute Savings Plans, once a resource has 60+ days of history and steady measured uptime; cross-check them with your provider’s own RI/SP recommendations.
Related lessons
- L1: The four commitment instruments
- L3: Over-commitment is worse than under-commitment (next)
- L4: Schedule first, commit second
- L5: Share-of-savings vs flat-fee vendors
Glossary terms touched
Effective discount · Coverage · Utilization · Break-even coverage · Realized savings