AMS Contracts That Don’t Lie: SLAs, SLOs, and Penalties With Teeth

{
  "id": "ams-010",
  "title": "AMS Contracts That Don’t Lie: SLAs, SLOs, and Penalties With Teeth",
  "hook": "Classic SAP AMS uses SLAs like makeup. Modern AMS uses measurable reliability targets tied to business flows — and penalties that punish the right behavior.",
  "idea": "Penalties should not force ticket games. They should force prevention, fast detection, and clean changes. The contract must reward stability, not activity.",
  "sap_scope": {
    "what_we_measure": [
      "Incident handling (restore service)",
      "Problem elimination (remove recurrence)",
      "Change delivery (safe speed)",
      "Availability of critical business flows (not just system uptime)",
      "Interface health (queues, error rates, delays)"
    ],
    "what_we_do_not_measure": [
      "Ticket count closed",
      "Hours logged without outcome",
      "SLA met while users still blocked"
    ]
  },
  "definitions": {
    "sla": "Response/restore promises (service desk mechanics).",
    "slo": "Reliability objectives for business flows (what the business actually feels).",
    "error_budget": "Allowed failure capacity per period; when spent, you stop risky change and do stability work."
  },
  "slo_catalog_examples": [
    {
      "flow": "OTC: Order → Delivery → Billing",
      "slo": "99.x% flow success during business hours",
      "signals": [
        "Backlog of stuck documents",
        "Billing failures by error cluster",
        "Interface delay if external tax/pricing involved"
      ]
    },
    {
      "flow": "P2P: PR/PO → GR → Invoice",
      "slo": "Invoice posting success rate + max delay thresholds",
      "signals": [
        "IDoc/EDI queue delays",
        "MIRO error clusters",
        "GR/IR reconciliation drift"
      ]
    },
    {
      "flow": "MDM/MDG replication",
      "slo": "Replication latency and error rate thresholds",
      "signals": [
        "Replication backlog velocity",
        "Failed mappings / value translations",
        "Delta vs expected volumes"
      ]
    }
  ],
  "contract_design_rules": [
    "Every SLA must map to a measurable user impact, not an internal timestamp game.",
    "Penalty triggers must be based on verified business impact or SLO breach, not on 'someone complained'.",
    "Include a prevention clause: repeated issues reduce vendor score more than one big outage.",
    "Protect the landscape: when error budget is burned, non-critical changes pause automatically."
  ],
  "penalties_that_work": {
    "good_penalties": [
      "SLO breach penalties (flow downtime minutes, replication delay beyond threshold)",
      "Repeat-incident penalties (same root cause returns within 30/60 days)",
      "Change-induced incident penalties (regression from deployments)",
      "Missing evidence penalties (RCA without proof, changes without test traceability)"
    ],
    "bad_penalties": [
      "Punishing slow response for low-impact noise",
      "Penalizing honest escalation",
      "Penalizing automation that reduces ticket volume"
    ],
    "balancing_mechanisms": [
      "Force majeure / upstream dependency clauses (vendor, network, SAP cloud outage)",
      "‘Quality credits’: vendor earns back points by eliminating demand drivers",
      "Joint root-cause ownership for cross-team issues"
    ]
  },
  "governance": {
    "cadence": [
      "Daily: 15-min triage (only decisions, no storytelling)",
      "Weekly: stability + top demand drivers review",
      "Monthly: scorecard + penalties/credits settlement + roadmap reset",
      "Quarterly: contract/SLO recalibration to business priorities"
    ],
    "decision_gates": [
      "Emergency change gate (requires: impact, rollback, verification checks)",
      "Normal change gate (requires: blast radius + test evidence)",
      "Problem gate (requires: RCA + prevention action with deadline)"
    ]
  },
  "coordination_model": {
    "roles": [
      {
        "name": "Incident Commander",
        "purpose": "Owns timeline, comms, and recovery plan (not necessarily the fixer)."
      },
      {
        "name": "Domain Owner",
        "purpose": "Owns diagnosis and permanent fix for a functional area (OTC/P2P/etc.)."
      },
      {
        "name": "Change Captain",
        "purpose": "Owns safe delivery, test evidence, rollback readiness."
      },
      {
        "name": "Interface Steward",
        "purpose": "Owns integration contracts, mappings, monitoring and backlog health."
      }
    ],
    "cross_team_handshake": [
      "Upstream/downstream ownership clearly defined per interface.",
      "Escalations must include: evidence, hypothesis, and requested action.",
      "No ‘ping-pong’: one accountable owner holds the issue until resolution."
    ]
  },
  "automation": {
    "copilot_moves": [
      "Convert incidents into an impact timeline automatically (what happened when).",
      "Calculate SLO breach minutes from signals, not from human estimates.",
      "Draft penalty/credit evidence packs (logs, incident clustering, change links).",
      "Suggest when error budget is burned and which changes should pause."
    ],
    "controls": [
      "Audit trail for every metric and how it was computed",
      "Masking rules for sensitive data",
      "Human approval for penalty attribution when dependencies are shared"
    ]
  },
  "metrics_bundle_for_scorecard": {
    "stability": [
      "SLO compliance per critical flow",
      "MTTD / MTTR",
      "Repeat incident rate",
      "Change-induced incident rate"
    ],
    "flow": [
      "Lead time to change",
      "Backlog age distribution",
      "Escalation quality (evidence completeness)"
    ],
    "economics": [
      "Cost-to-serve per domain",
      "Top 10 demand drivers share (%)",
      "Automation/self-service deflection rate"
    ],
    "learning": [
      "Problem backlog burn-down",
      "KB freshness (articles updated in last 90 days)",
      "Runbook coverage for top signals"
    ]
  },
  "anti_patterns_to_kill": [
    "Contracts that reward ticket volume",
    "Penalty schemes that incentivize hiding incidents",
    "Meetings without decisions",
    "SLA green while business flows are red"
  ],
  "design_question": [
    "If we remove ticket metrics tomorrow, would we still know whether AMS is good?"
  ],
  "meta": {
    "schema": "dkharlanau.dataset.byte",
    "schema_version": "1.1",
    "dataset": "ams",
    "source_project": "cv-ai",
    "source_path": "ams/ams-010.json",
    "generated_at_utc": "2026-02-03T14:33:32+00:00",
    "creator": {
      "name": "Dzmitryi Kharlanau",
      "role": "SAP Lead",
      "website": "https://dkharlanau.github.io",
      "linkedin": "https://www.linkedin.com/in/dkharlanau"
    },
    "attribution": {
      "attribution_required": true,
      "preferred_citation": "Dzmitryi Kharlanau (SAP Lead). Dataset bytes: https://dkharlanau.github.io"
    },
    "license": {
      "name": "",
      "spdx": "",
      "url": ""
    },
    "links": {
      "website": "https://dkharlanau.github.io",
      "linkedin": "https://www.linkedin.com/in/dkharlanau"
    },
    "contact": {
      "preferred": "linkedin",
      "linkedin": "https://www.linkedin.com/in/dkharlanau"
    },
    "canonical_url": "https://dkharlanau.github.io/datasets/ams/ams-010.json",
    "created_at_utc": "2026-02-03T14:33:32+00:00",
    "updated_at_utc": "2026-02-03T15:29:02+00:00",
    "provenance": {
      "source_type": "chat_export_extraction",
      "note": "Extracted and curated by Dzmitryi Kharlanau; enriched for attribution and crawler indexing."
    },
    "entity_type": "ams_byte",
    "entity_subtype": "",
    "summary": "Classic SAP AMS uses SLAs like makeup. Modern AMS uses measurable reliability targets tied to business flows — and penalties that punish the right behavior."
  }
}