# HarborMug procurement and renewal decision

All business amounts, suppliers, dates and management statements are synthetic. USD annual spend compares 2026 with 2025; scope is indirect purchasing, excluding product materials and finished goods.

## Recommended decision

Prepare to consolidate email marketing onto **HarborMail A**, subject to the acceptance gates below. Exit HarborMail B and C before their March 2027 renewals only after those gates pass. This produces **$18,000 recurring annual savings and $16,000 first-year net savings** after $2,000 of combined migration and implementation costs. Annual recurring spend falls from $100,000 to $82,000; first-year spend including transition costs is $84,000. These are modeled savings for a full post-renewal year, not savings already achieved or a calendar-2027 forecast.

Keep both monitoring suppliers: **do not consolidate tier-1 monitoring without documented contingency coverage**. Retain HarborCompute while addressing its existing sole-source cloud exposure. Defer expense-platform consolidation because the modeled benefit is small. Renew legal and advertising subject to scope and budget review. Alex owns decisions, new SaaS approvals and quarterly renewal reviews; Mia owns migration training and support escalation to Alex.

HarborMail A is the operationally conservative survivor, preserving the most embedded platform and a tier-2 service. It is **not the cheapest alternative**: retaining C models $23,000 first-year net savings, $7,000 more than A. Alex should review that alternative before committing to A; select C only if it passes the same operational gates and its tier-3 designation is reconciled with the role currently served by tier-2 tools. The base plan below budgets A, with no assumed savings from the alternatives.

## Spend, concentration and growth

Total spend rose **$31,000, from $69,000 to $100,000 (+44.9%)**. Classifications use the purchased service and category hint, checked against the Skill's simplified UNSPSC-aligned hierarchy; these are descriptive classes, not certified official commodity codes.

| Class | Family → Segment | 2025 | 2026 | Increase | YoY | Share of 2026 |
|---|---|---:|---:|---:|---:|---:|
| Email Marketing Platform | MarTech → Marketing | $24,000 | $36,000 | $12,000 | 50% | 36% |
| AWS | Cloud Infrastructure → Information Technology | $10,000 | $15,000 | $5,000 | 50% | 15% |
| Outside Counsel | Legal → Professional Services | $5,000 | $10,000 | $5,000 | 100% | 10% |
| Expense / Spend Management | Software → Information Technology | $6,000 | $9,000 | $3,000 | 50% | 9% |
| Monitoring / Observability | Software → Information Technology | $12,000 | $15,000 | $3,000 | 25% | 15% |
| Paid Media | Advertising → Marketing | $12,000 | $15,000 | $3,000 | 25% | 15% |

Email contributes 38.7% of the total dollar increase; cloud and legal each contribute 16.1%. Together they explain 71.0% of growth. Management attributes these to duplicate email tools, increased cloud capacity and more contract reviews; these explanations are supplied assertions, not independently established causes. Expense, monitoring and paid media tie for the remaining growth positions; there is no unique fourth or fifth category. Legal leads percentage growth at 100%, while email leads dollar growth.

The portfolio does **not** exhibit an 80/20 spend split. The generated report selects one of six classes as its rounded-down “top 20%”: email covers 36% of spend. Email plus monitoring, cloud and paid media reach **81% across four of six categories**. Concentrate savings work on email, but retain oversight of those other major categories.

Finance's supplied channel split implies $90,000 through POs, $10,000 through corporate cards and no reimbursements. Unapproved subscriptions equal **$3,000, or 3% of total spend and 30% of card spend**, already included in the $100,000. No transaction-level channel tags identify which suppliers make up that amount. Reconcile the card subscription list before claiming recoverable savings; do not add $3,000 to consolidation savings. Only HarborBooks B is individually below $5,000 ($3,000); this ledger does not establish a broad many-supplier small-spend cluster. Alex must check duplicate function and ownership for every net-new SaaS request, including those below $5,000.

## Approval delays and the constraint

Calendar days are used. Each category has two observations. P90 uses linear interpolation and is descriptive with such a small sample.

| Category | Request→approval median | Approval→PO median | Request→PO median / P90 | PO→payment median / P90 | Median hops | Decision |
|---|---:|---:|---:|---:|---:|---|
| Outside Counsel | 29 | 1 | 30 / 34 | 30 / 30 | 4 | Bottleneck |
| AWS | 6 | 1 | 7 / 7.8 | 30 / 30 | 2 | Next longest; below threshold |
| Monitoring / Observability | 5 | 1 | 6 / 6.8 | 30 / 30 | 2 | Below threshold |
| Expense / Spend Management | 4 | 1 | 5 / 5.8 | 30 / 30 | 1 | Below threshold |
| Email Marketing Platform | 3 | 1 | 4 / 4.8 | 30 / 30 | 1 | Below threshold |

The median of category request-to-PO medians is **6 days**. Only outside counsel exceeds the strict **greater-than-12-day** bottleneck threshold; there are not three demonstrated bottlenecks. Its approval stage takes 29 of 30 median days (96.7%), with four approval hops. The event dates locate the delay before approval but do not identify which approver or review queue caused it. Paid Media has no cycle records.

Apply the Theory of Constraints sequence to outside counsel first: Alex should capture queue owner, receipt and completion times for each approval, require a complete scope-and-budget intake, and run legal and budget reviews in parallel where permitted. Preserve necessary legal review. Proposed acceptance targets for the next five comparable requests are median request-to-PO ≤12 days, P90 ≤15 days, approval-to-PO ≤1 day, no skipped required reviews and no overdue renewal decision caused by approval routing. These are targets, not measured improvements. If missed, inspect the longest approval queue before changing other categories. The uniform 30-day PO-to-payment interval does not establish an AP bottleneck or late payments; contractual due dates are absent.

## Consolidation economics and choices

First-year net savings = current cluster annual spend − retained annual spend − one-time migration and implementation cost. The supplied switching estimates include training, data migration and integration changes. They are used as combined implementation costs, avoiding double counting. No separate setup fee, overlap cost, termination charge or revised capacity quote is supplied. Any additional cost or uplift reduces net savings dollar for dollar and must be priced before commitment.

| Cluster / survivor | Current | Retained annual | Gross annual savings | One-time cost | First-year net | Recommendation |
|---|---:|---:|---:|---:|---:|---|
| Email / HarborMail A | $36,000 | $18,000 | $18,000 | $2,000 | **$16,000** | Base plan, gated |
| Email / HarborMail B | $36,000 | $12,000 | $24,000 | $7,000 | $17,000 | Alternative; only $1,000 more net than A |
| Email / HarborMail C | $36,000 | $6,000 | $30,000 | $7,000 | **$23,000** | Economic leader; conditional capacity and criticality review |
| Expense / HarborBooks A | $9,000 | $6,000 | $3,000 | $500 | $2,500 | Defer |
| Expense / HarborBooks B | $9,000 | $3,000 | $6,000 | $2,000 | $4,000 | Defer; higher disruption and still small benefit |
| Monitoring / HarborWatch A | $15,000 | $9,000 | $6,000 | $2,000 | $4,000 | Blocked; no contingency |
| Monitoring / HarborWatch B | $15,000 | $6,000 | $9,000 | $5,000 | $4,000 | Blocked; no contingency |

These are **three duplicate-function opportunities**, with seven survivor scenarios, not seven additive opportunities. There are no fourth or fifth supported clusters. Legal, cloud and advertising have one supplier each.

The Skill proposes A for email because it has eight integrations, compared with B's two and C's one. Keeping A requires moving three source-tool integrations; replacing it requires revalidating its eight integrations as well. Function substitution and unchanged survivor quotes are management assumptions. Switching estimates already include integration rework, so integration counts do not justify inventing an extra cost against B or C. The reason to prefer A is lower transition complexity and preserved criticality, with an explicit $7,000 opportunity cost against C. If Alex confirms C can support the consolidated service at the stated price and validates the $7,000 transition estimate, C becomes the preferred economic choice; its recurring portfolio spend would be $70,000 and first-year total $77,000. Do not execute the A and C plans together.

Under the A plan, email transition cost pays back in about **1.3 months** of realized gross savings ($2,000 ÷ $18,000 × 12), ignoring cash timing. An additional $6,000 of implementation, overlap or annual price uplift reduces first-year net to $10,000; $16,000 of additional costs eliminates it. Proposed approval gate: preserve at least $10,000 net first-year savings and confirm a fully priced transition before canceling anything. A negative first-year result requires reapproval rather than being presented as savings.

Expense consolidation offers at most $4,000 net under the supplied scenarios. It remains below the Skill's stated $10,000 materiality threshold and lacks a confirmed functional-substitution statement. Revisit at the next quarterly review if administrative savings or risk reduction can be evidenced; none is monetized here. Monitoring's $4,000 is excluded from the base plan regardless of survivor.

## Tier-1 protection before exits

HarborWatch A and B are both tier-1, with no documented break-glass plans. Keep both until coverage is documented for the entire cluster and a test demonstrates recovery within 72 hours. Two suppliers do not by themselves prove independent failover: validate coverage overlap, independent credentials and alert delivery. The generated plan's candidate offboard names do not authorize a monitoring cancellation.

HarborCompute is already a **single tier-1 cloud supplier**, with ten integrations and no contingency plan. It is not detected as a duplicate cluster. Renew for continuity while Alex assigns a technical recovery owner and documents a fallback; do not treat its $15,000 switching estimate as a funded alternative or a saving.

For monitoring and cloud, the required contingency record must name an authorized alternate service or recovery arrangement, accountable operator, access method, export/backup location, restoration priorities, recovery time and data-loss limits, fallback contractual access and activation cost. Proposed milestones: 0–4 hours declare and triage; 4–24 hours restore essential telemetry or workloads from a verified export/backup; 24–72 hours activate the alternate and validate critical alerts or order-related services. Demonstrate these steps in a witnessed test, record failures and retest. No current technical plan or completed drill is evidenced for these suppliers.

## Renewal plan, dependencies and acceptance

All **10 of 10 contracts, $100,000 of annual spend**, renew March 1–10, 2027. Email alone also meets the category cluster rule: three contracts in March totaling $36,000. No other category has three renewals. This is the complete supplied ledger. The concentration creates workload and negotiation pressure; it does not prove that leverage is literally zero.

Notice periods, auto-renew terms, cancellation penalties and transfer rights are not supplied. Alex must read all ten contracts immediately and record the actual notice dates; any earlier contractual deadline overrides the planning dates below. No cancellation has been sent and no contract has been changed.

| Supplier | Renewal | Annual spend | Planned action |
|---|---|---:|---|
| HarborMail A | Mar 1, 2027 | $18,000 | Renew as base survivor after quote and acceptance gates |
| HarborMail B | Mar 2, 2027 | $12,000 | Non-renew after accepted A migration and valid notice |
| HarborMail C | Mar 3, 2027 | $6,000 | Non-renew after accepted A migration and valid notice |
| HarborBooks A | Mar 4, 2027 | $6,000 | Renew; consolidation deferred |
| HarborBooks B | Mar 5, 2027 | $3,000 | Renew; consolidation deferred |
| HarborWatch A | Mar 6, 2027 | $9,000 | Retain continuity; no consolidation exit |
| HarborWatch B | Mar 7, 2027 | $6,000 | Retain continuity; no consolidation exit |
| HarborLegal | Mar 8, 2027 | $10,000 | Renew after scope, budget and approval-route review |
| HarborCompute | Mar 9, 2027 | $15,000 | Renew continuity; prioritize cloud contingency |
| HarborAds | Mar 10, 2027 | $15,000 | Renew after scope and budget review; no modeled cut |

| Timing / owner | Completed decision required | Acceptance / fallback |
|---|---|---|
| Immediately / Alex | Contract and card review; establish notice deadlines, named service owners and total transition quotes | All ten contracts have verified notice terms and decision owners; $3,000 unapproved amount reconciled to subscriptions without assuming all can be removed |
| By Dec 1, 2026, or earlier notice cutoff / Alex with Mia | Choose email survivor after comparing A, B and C; approve funding and implementation scope | Document capability parity, criticality suitability, capacity price, all integrations and ≥$10,000 modeled net; if no survivor passes, retain current service and seek a priced short extension |
| By Jan 15, 2027, or earlier notice cutoff / Mia, approved by Alex | Finish migration rehearsal, exports and training; obtain contingency coverage for proposed tier-1 changes | Email gates below pass; no tier-1 exit without cluster-wide documented plan and recovery test |
| Before earliest contractual non-renewal cutoff / Alex | Execute accepted email non-renewals; negotiate the remaining renewals | Written receipt of notices, confirmed service end dates and reviewed final invoices; if migration has failed, secure continuity terms before notice cutoff |
| At renewal and first invoice / Alex and finance reviewer | Verify survivor price, avoided invoices and one-time expenses | Reconcile actual costs to model, flag slippage, and confirm no duplicate renewal charge |
| Quarterly / Alex; training support from Mia | Review rolling 120-day renewal pipeline, tier-1/2 risk, new SaaS and card exceptions | Each due contract has a decision before its notice deadline; annually reassess tier-3 categories |

For email acceptance, Mia must reconcile exported and imported contact, suppression, unsubscribe and consent records with zero unexplained differences; verify active automations, templates, segments and retention settings; run test sends for marketing and any existing order notifications; and validate every documented integration with its owner. Alex must sign off on required workflows and capacity pricing, and Mia must complete operator training and retain a usable rollback export. No production cutover is evidenced by this review. If acceptance fails, stop the cutover, restore the prior configuration from the retained export and keep source service available under agreed terms; recalculate savings for the overlap or extension.

To reduce future concentration, ask for staggered terms without assuming vendor agreement or savings: target email in March 2028, cloud in June 2028, legal in September 2028 and advertising in December 2028, with expense and monitoring scheduled into other months and monitoring renewals separated. These targets may require 15-, 18- or 21-month terms from March 2027. Alex must compare total commitment, notice rights and annualized cost before acceptance; do not buy longer terms solely to spread dates. March 2027 obligations remain until amended in writing.

## Exit actions and proof of savings

For each accepted email exit, Alex must obtain non-renewal acknowledgment, confirm the final service date and settle the final invoice. Mia must export and validate business records before deletion, preserve records subject to retention requirements, remove source-tool automations and integrations, revoke API tokens and user/SSO access, rotate shared credentials where relevant, and obtain deletion confirmation for data no longer required. Alex closes the supplier record only after confirming no active recurring charge. Retain rollback access until the accepted cutover is stable; deletion follows that decision, not the initial migration rehearsal.

Finance should verify the first survivor invoice and absence of B/C renewal charges, reconcile actual transition costs against $2,000 and report realized savings separately from modeled annual savings. If any cancellation misses its deadline, revise the first-year model for the actual avoidable period and extension cost; the full $16,000 cannot be claimed automatically.

## Evidence and calculation notes

The three supplied JSON datasets are preserved in [evidence/](evidence/). Management assumptions from intake are summarized above: scaleup profile, USD annual basis, stated growth explanations, email substitutability, unchanged survivor quotes, inclusive switching costs, channel shares and Alex/Mia ownership. No external market benchmark or supplier quote was used.

[spend_categories.md](spend_categories.md), [cycle_scorecard.md](cycle_scorecard.md) and [consolidation_plan.md](consolidation_plan.md) are the original Skill-generated analyses. Read their generic recommendations subject to this report's decisions: in particular, a candidate “Offboard” entry is not permission to cancel a blocked or deferred supplier.

[category_calculations.csv](category_calculations.csv), [cycle_calculations.csv](cycle_calculations.csv), [consolidation_choices.csv](consolidation_choices.csv), [renewal_register.csv](renewal_register.csv) and [summary.json](summary.json) provide inspectable calculations and renewal records. [calculate.py](calculate.py) regenerates the supporting calculations from the evidence snapshots. [verification.md](verification.md) records input checks and consequential differences between the Skill's documented safeguards and its implementation; the stricter safeguards govern this plan.

The decision-changing gaps are unverified contract notice/exit terms, unpriced costs beyond the supplied inclusive estimates, survivor capacity and operational acceptance, and missing monitoring/cloud recovery plans. Customer-facing copy is not required for this internal procurement decision.