SEO Retainer Overservice Audit with Proof Rows

Audit any SEO retainer in one run by joining quoted scope against actual delivered work to surface overservice hours with evidence rows. Paste two small CSV-shaped lists — quoted hours and actual hours — and get one proof row per overservice client with unbilled dollars, margin impact, and repricing numbers. [Try it — the input opens prefilled with a $0 demo.]

This Actor is the one-run audit only with no weekly tracking in this Actor; it never monitors anything.

Price: audit-completed $0.30 per completed audit (rows or zero-leak quiet) · the fixed demo input runs a $0 trial with no charge · floor $0.31 (charge limits below this skip billing at $0 with counts only). Console Pricing tab is authoritative.

What it does / What you get

Find unbilled SEO retainer hours before they eat your margin. This Actor compares what you quoted against what you actually worked and prices the gap — one run, no logins, no fetching, no monitoring.

• Quoted-vs-actual join — matches each client + service pair across your quote rows and your time-tool rows.
• Overservice proof rows — one dataset row per leaking pair with unbilledAmount, marginErasedPct, and phantomFte.
• Evidence reference on every row — evidenceUrl points at the winning quote-row index plus checkedAt and a check notice, so each number is verifiable.
• Repricing numbers (arithmetic only) — repricingHrs, repricingRate, and repricingFeeUplift suggest the corrected fee; you send it, the tool never pushes anything.
• Quiet suppression — pairs at or inside scope emit nothing; a clean book returns zero rows by design.
• Run summary in KV OUTPUT — counts, quiet flag, and billing keys (billed, chargeUsd, trial, skip) for every completed audit.
• Privacy by design — a person column is accepted and dropped at ingest; person-level data never appears in any row.

Quick start

1. Click Try for free — the input opens prefilled with the $0 demo (5 quote + 5 actual rows).
2. Keep the demo labels (runId, scopeBook, checkedAt) or set your own.
3. Click Save & Start, then open the Dataset tab.
4. Check the 3 overservice rows and the unbilled total of 2280 — this run is the $0 trial, no charge.
5. Replace the rows with your own book (max 5 + 5 rows) and run again for a $0.30 audit.

Input

Minimal JSON (required-only — all five fields are required, so this is also the full shape; values below equal the schema prefill exactly):

{
  "quoteCsv": [
    {"client": "Acme", "quotedHrs": 40, "rate": 90, "service": "SEO-retainer"},
    {"client": "Acme", "quotedHrs": 20, "rate": 120, "service": "PPC"},
    {"client": "Beta", "quotedHrs": 10, "rate": 100, "service": "Web-care"},
    {"client": "Gamma", "quotedHrs": 30, "rate": 80, "service": "SEO Monthly"},
    {"client": "Delta", "quotedHrs": 15, "rate": 150, "service": "Audit"}
  ],
  "actualCsv": [
    {"client": "Acme", "service": "SEO-retainer", "date": "2026-09-01", "actualHrs": 52},
    {"client": "Acme", "service": "PPC", "date": "2026-09-02", "actualHrs": 20},
    {"client": "Beta", "service": "Web-care", "date": "2026-09-03", "actualHrs": 18},
    {"client": "Gamma", "service": "SEO Monthly", "date": "2026-09-04", "actualHrs": 35},
    {"client": "Delta", "service": "Audit", "date": "2026-09-05", "actualHrs": 12}
  ],
  "runId": "demo-audit-001",
  "scopeBook": "demo-book-001",
  "checkedAt": "2026-09-28T12:00:00Z"
}

field	type	required	default	units	limits
quoteCsv	array of objects (client, service, quotedHrs, rate, optional retainerFee)	yes	—	rows	1–5 rows (maxItems: 5)
actualCsv	array of objects (client, service, date, actualHrs, optional person dropped at ingest)	yes	—	rows	1–5 rows (maxItems: 5)
runId	string	yes	—	label	min length 1, echoed in the report
scopeBook	string	yes	—	label	min length 1, echoed on every row
checkedAt	string	yes	—	ISO-8601 date-time	stamped on every row

Job-titled recipes (one-line each):

• demo trial — prefill as-is → 3 rows, $0 trial, no charge.
• agency monthly book — 3 quote + 3 actual leaking pairs → 3 rows, $0.30.
• full-cap boundary — 5 quote + 5 actual rows including at-scope quiet pairs → 3 rows, $0.30.
• quiet check — every pair at or inside scope → 0 rows by design, still a completed $0.30 audit.

Over-cap books (a 6th row on either side) are rejected by the platform schema before the run starts — $0, no dataset, no charge.

Output

Overservice rows go to the default dataset (one item per overservice client + service); the run summary goes to OUTPUT in the default key-value store. Quiet audits and rejections write OUTPUT with zero dataset rows.

Dataset fields (15 per item; duplicateNote only where quote rows repeated the pair):

field	type	meaning
client / service	string	pair identity, as written in your quote
quotedHrs / actualHrs	number	quoted hours vs summed actual hours
unbilledAmount	number (USD)	(actualHrs − quotedHrs) × rate, positive only
phantomFte	number	excess hours over a 160h month, informational
marginErasedPct	number	share of the contracted fee given away, in percentage points
repricingHrs / repricingRate / repricingFeeUplift	number	suggested corrected hours, rate, and fee uplift (arithmetic only)
duplicateNote	string (optional)	records collapsed duplicate quote rows, first row wins
scopeBook / evidenceUrl / checkedAt / notice	string	book label, source-row reference, timestamp, check notice

Real example rows (shortened for viewing — 2 of the 3 demo rows):

{ "client": "Acme", "service": "SEO-retainer", "quotedHrs": 40, "actualHrs": 52, "unbilledAmount": 1080, "phantomFte": 0.075, "marginErasedPct": 30.0, "repricingHrs": 52, "repricingRate": 90, "repricingFeeUplift": 1080, "scopeBook": "demo-book-001", "evidenceUrl": "file:fixtures/demo-audit/input.json#/quoteCsv/0", "checkedAt": "2026-09-28T12:00:00Z", "notice": "checked on record" }

{ "client": "Beta", "service": "Web-care", "quotedHrs": 10, "actualHrs": 18, "unbilledAmount": 800, "phantomFte": 0.05, "marginErasedPct": 80.0, "repricingHrs": 18, "repricingRate": 100, "repricingFeeUplift": 800, "scopeBook": "demo-book-001", "evidenceUrl": "file:fixtures/demo-audit/input.json#/quoteCsv/2", "checkedAt": "2026-09-28T12:00:00Z", "notice": "checked on record" }

Runs finish in seconds on any allowed book size. Download rows as CSV or JSON from the Dataset tab; the repricing dataset view shows the suggested-fee columns together.

How much will it cost?

from $0.30 / completed audit — one flat price on every plan (Free / Starter / Scale / Business all $0.30). Billable event: audit-completed, fired exactly once per COMPLETED audit — with rows or zero-leak quiet alike (per-audit, never per-row). The frozen demo input, rejections, over-cap trips, and pin-skips never charge.

plan	$/completed audit
Free / Starter / Scale / Business	$0.30 / $0.30 / $0.30 / $0.30

Worked math:

• 1 demo-trial audit (prefill as-is) → $0 (no charge call, logged as trial).
• 1 real audit (any 1–5 + 1–5 book, rows or quiet) → $0.30.
• 10-audit monthly loop → $3.00.
• $5 free credit → ≈16 real audits after the $0 demo.

Caps, floor, quiet costs: maxItems: 5 on each CSV is the cost ceiling — bigger books never reach the run and cost $0. Floor $0.31 (minimal max-total-charge): a charge limit below it skips billing at $0 with counts only and no dataset rows; set ACTOR_MAX_TOTAL_CHARGE_USD at or above it for real audits. $0.31 is also the platform's minimum settable limit for this actor, so no settable value falls between the $0.30 audit price and the floor — the two thresholds coincide in practice. No usage add-on (pay-per-event-plus-usage stays OFF), no proxy or AI multipliers, start event at platform default.

event:audit-completed | $/audit 0.30/0.30/0.30/0.30 (Free/Starter/Scale/Business) | includes_usage:false | billed_on_empty:true (zero-leak quiet audits bill) | caps:[quoteCsv maxItems 5, actualCsv maxItems 5] | quiet:[actor-start default]

Trial / Free tier

$5 credit ≈ 16 real audits plus unlimited $0 demo runs. The first run is prefilled and small (5 + 5 rows, the frozen demo envelope): it rides the $0 trial path with no charge call. Any one-byte change to the demo input makes it a real $0.30 audit — that is the anti-farming rule, and it keeps copy-paste agent loops free. Same 5 + 5 caps on free; no paid plan needed for the demo or your first audits.

Click Try for free — input opens prefilled. Start as-is, check rows, then run your own book.

Compatibility

Works with any 2-CSV audit joining quoted scope against actual hours. Verified live on anonymized excerpts, Oct 2026:

• Over-scope excerpt — quoted-vs-actual pair with actual above quoted emitted 1 proof row with evidence reference, Oct 2026.
• At-scope quiet excerpt — pairs at or inside scope correctly suppressed to 0 rows, Oct 2026.
• Frozen demo envelope — synthetic 5 + 5 demo emitted 3 rows with unbilled total 2280 on the $0 trial path, Oct 2026.

Not covered: over-cap books above 5 + 5 rows, login-shaped or malformed input. These are rejected with zero rows — never a guessed row.

Proof excerpts are anonymized; no affiliation, endorsement, or sponsorship implied.

What people use it for

• Agency owner month-end check — join this month's quote against the time-tool export, find every leaking retainer.
• Retainer repricing evidence — take the suggested fee uplift into the renewal conversation as arithmetic, not assertion.
• Scope-creep proof — show the client which service grew past its hours, with a source reference per pair.
• Clean-month verification — run a quiet book to prove nothing leaked, still one flat $0.30.

What this Actor does NOT do

• Does not export a repricing CSV — one-run audit only; proof rows stay in the dataset.
• Does not keep weekly history or benchmarks — no weekly tracking in this Actor.
• Does not fetch pages, take logins, or monitor anything — owner-assembled rows in, evidence rows out.
• Does not emit person-level verdicts — suggested numbers are arithmetic only; verify before you act.

Use with schedules / webhooks / API / MCP

Schedules — run the audit weekly from the Console (Schedules tab → fixed book input). Each real run bills $0.30; keep a $0 demo run alongside as the free proof it still works.

Webhooks — alert-only-when-rows-leak pattern on your own endpoint: fire on run SUCCEEDED, read KV OUTPUT, and page only when rowsEmitted > 0:

{ "watch": "OUTPUT.rowsEmitted", "alert": "rowsEmitted > 0", "quiet": "0 rows means clean book, no alert" }

API — start a run and poll it (runs finish in seconds; never block a call-actor wait longer than ~45s — poll get-actor-run instead):

$curl -X POST "https://api.apify.com/v2/acts/steadyworks~seo-retainer-overservice-audit/runs?token=<APIFY_TOKEN>" -H "Content-Type: application/json" -d "@your-book.json"

// Node.js (apify-client)
await client.actor("steadyworks/seo-retainer-overservice-audit").call({ quoteCsv, actualCsv, runId, scopeBook, checkedAt });

# Python (apify-client)
client.actor("steadyworks/seo-retainer-overservice-audit").call({"quoteCsv": quote_csv, "actualCsv": actual_csv, "runId": run_id, "scopeBook": scope_book, "checkedAt": checked_at})

MCP — agents reach this Actor through the hosted Apify MCP server (search-actors → fetch-actor-details → call-actor):

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com"
    }
  }
}

Connect with your Apify account, search seo-retainer-overservice-audit, and call it with the Input JSON above. No standby server, no extra permissions — runs only.

Troubleshooting

symptom	cause	fix
0 rows on a real book	all pairs at/inside scope — quiet audit by design (still $0.30)	read KV OUTPUT report.leakingCount; add one over-scope pair to verify the join
HARD-FAIL, zero rows	over-cap (>5 rows a side), malformed, or login-shaped input	trim to 5 + 5 rows with the required keys; never send credentials
$0 billed with counts only, no rows	charge limit below the $0.31 floor (PIN_BELOW_PRICE skip)	raise max total charge to ≥ $0.31 and re-run
demo input billed $0.30	input differs ≥1 byte from the frozen demo envelope	re-paste the prefill exactly; any edit makes it a real audit
person column missing in rows	by design — dropped at ingest	nothing to fix; per-person analysis stays in your time tool

FAQ + legality

Does this change my data? No — read-only. It joins the rows you paste and writes proof rows to your own dataset; your quote and time tool are untouched.

How much does one audit cost? $0.30 per completed audit (rows or zero-leak quiet). The frozen demo is a $0 trial; rejections and cap-trips are $0.

What counts as a completed audit? Any run that joins your book — leaking rows or a quiet zero-row result. Only rejections (bad shape), over-cap trips, and floor-skips are not completed and never bill.

Do empty results really bill? A quiet zero-leak book is a completed audit and bills $0.30 — the value is the verified-clean answer, not the row count.

Does it handle logins or JavaScript pages? There is nothing to log into — it never fetches. It reads only the rows you supply.

Personal data / GDPR? The person field is dropped at ingest and never emitted. Process only books you own or have the right to process; each row carries a check notice — verify before you act on it.

Terms? Run audits on your own client books. Suggested repricing figures are arithmetic, not legal or pricing advice.

Other Actors you might need

need	where to go
repricing CSV export	Not in this Actor — one-run audit only; proof rows stay in the dataset
weekly history + benchmark drift	Not in this Actor — no weekly tracking in this Actor
books above 5 + 5 rows	Not in this Actor — over-cap books are rejected at $0

Nothing else is needed for the one-run audit itself — two small lists in, proof rows out.

Changelog + support

./CHANGELOG.md — 0.2 TEST-RUN $0.30 cohort (2026-10-02) · 0.1 adoption baseline (2026-10-02). Issues tab on the Store listing, or the GitHub repo (lukegor/13-agency-overservice-ledger).