Fix each reported figure to the method, window, attribution setting and extraction time that produced it, hold the residual against the warehouse of record, and announce a restatement rather than smoothing it.
2Figure context assembledThe campaign, the method behind the number, the window it covers and when it was extracted
3Reconciliation draftedThe figure, its sources, the residual against the warehouse and completeness
4Controls appliedMethod checks, window checks, residual thresholds and completeness confidence
No human action required
Stages 1 to 4 run unaided, and nothing is signed off at any of them — the agent is reconciling, and the analyst lane opens at the completeness gate.
5DecisionSplits at the completeness gate
Evidence sufficient
Goes to the named analyst to sign off.
Anything thin
Adds a second analyst read first.
Analyst review
The number is held with its method, its residual and the window it covers.
Sign off · Append evidence · Send to analyst review
Signed off — by a named analyst▼
6Reporting and warehouse records updatedOnly where write access and records policy allow it
7Outcome evaluatedResidual size, method coverage, analyst corrections and what the read found
Corrections
Each analyst correction is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required8 items
Signing a figure off for a board pack.
Closing or reopening a finance period.
Deciding which attribution model is right.
Restating a number after the period locks.
Automation boundaryAgent acts unaided
✓Hold the method, the window and the extraction time beside the figure.
✓Reconcile platform totals against the warehouse.
✓Label a modelled conversion as modelled wherever that number travels.
✓Flag the number that moved between two decks without.
Nothing is signed off or booked except by a named person, inside the agreed boundaries.
Judging whether a residual is acceptable.
Telling the board what a campaign returned.
Setting the attribution window we report on.
Changes to warehouse tables or finance postings.
Example output
One figure, annotated
Nothing here is a fresh measurement; this record is what one campaign number carried on the day it reached the warehouse.
Reconciliation entry · single figureIllustrative example
Figure
Recorded as
Method
Evidence of record
Confidence
Held for
Paid social conversions, campaign level
Reconciled to the warehouse, residual open
Modelled
Warehouse extract, 3 August 2026
Held unsigned
The reporting analyst, by name
As receivedTaken from the platform export and the warehouse table — it reaches as far as those sources do.
What the record holdsExtraction timeAttribution windowWarehouse residual
Why no sign-off hereWhether a residual may stand is an analyst judgement, not a model output.
ActionSign offAppend evidenceSend to analyst review
What the score decidesBelow the configured threshold a number picks up a second read before the analyst sees it.
Value
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value landsValue 01 – 04
Every figureFrom the source that made it
03Evidence
Where the evidence is used
Our advertising analytics agents answer to a client report that closes with the campaign; this one answers to the warehouse of record and a finance period that locks behind it.
01Approved path
A number needs its method
A figure without its method is a rumour: the window, the attribution setting and the extraction time travel beside it, or it does not travel at all.
02Human review
What was checked, and not found
No platform publishes a reconciliation of its own reported totals against advertiser warehouse data, no shared definition of an attributed conversion holds across platforms, and no vendor figure has been borrowed for this page.
04Build an evidence trail
The figure, the method that produced it and the analyst who signed off stay together.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Platform and channel sourcesAd platforms · web analytics Delivery and conversion exports
Warehouse of recordBigQuery · Snowflake Observed and modelled tables
Finance and planningERP · budget systems Booked revenue and period status
Agent
Campaign analytics reconciliation
Reads the figures Reconciles the sources Holds for the analyst
Reporting and case systemsBI tooling · ticketing Restatement and review records
A campaign-level reconciliation figure can read clean while paid social carries most of the rework. Nestack reports the correction rate by campaign type, not only in total.
Slice performance — reported separately, not only in aggregateIllustrative example
Slice
Failure rate
Lift
Lift vs. threshold
Status
Paid social campaigns
9.5%
3.6×
Review
Retail media placements
6.8%
2.6×
Review
Programmatic display
4.2%
1.6×
Watch
Search and brand campaigns
2.3%
0.9×
Normal
Bar: correction-rate lift vs. search and brand baseline · scale 0–4.0× · tick marks the 2.0× review threshold2 of 4 slices over threshold
Evidence-linked improvement
What a silent restatement costs
A loop ends when the silently restated metric has become a case the next release must pass. That suite is what the next report issued is measured against.
Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect
Correction rate rises on paid social campaigns.
02Diagnose
The number that changed between two decks and never said it had is read back until one cause remains.
03Improve
Number the change; the figures that drove it are filed underneath it.
04Verify
Each touched figure case is run once more, and one red holds it back.
05Learn
It stays on as a standing test, and the restatement rules move with it.
Learn → DetectThe return edge. The next report is measured against a suite one case longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, reconciliation, evaluation, integration, then production validation and handover.
WorkstreamWeek 1Week 2Week 3Week 4Week 5Week 6
01Method discovery and automation-boundary scoping.
02Platform, warehouse and finance sources.
03Figure-to-method and residual-threshold mapping.
04Figure and extract ingestion.
05Campaign, method and window binding.
06Completeness scoring and review routing.
07Analyst sign-off workflow.
08Warehouse and finance integration.
09Reconciliation and restatement cases.
10Guardrails and sign-off controls.
11Figure-trail instrumentation.
12Deployment, documentation and Agent Care handover.
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallelFinal scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
Capability✓ in scope · — not at this tierPilotOne campaign type, one quarterProductionProduction reporting workflowAdvancedMultiple brands / markets
Introduced at Pilot
Reconciliation to your warehouse✓✓✓
Named analyst sign-off✓✓✓
Reportable-metric baseline✓✓✓
Introduced at Production
Reporting by campaign—✓✓
Sign-off workflow in your systems—✓✓
Approved write-back—✓✓
Warehouse-and-platform integration—✓✓
Introduced at Advanced
Multi-brand estates——✓
Cross-market figure packs——✓
Large campaign estates——✓
Multi-source reconciliation controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build priceConfirmed after discovery based on integrations, workflow complexity, reporting volume, approval controls and deployment requirements.
Separate from buildBuild pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bringWe build with it
01Your campaigns and the platforms behind them→Source mapping and figure captureWeek 1
02Representative exports, warehouse tables and postings→Record binding, method logic and the reconciliation baselineWeek 2
03Your method, window and labelling rules→Figure mapping, method binding and the automation boundaryWeek 1
04Access to relevant APIs, feeds or exports→Platform, warehouse and finance-source assessment, then integration setupWeek 2
05Reports you would not want defended→Restatement cases and failure testingWeek 4
06What no attribution model may prove→Completeness scoring, review routing, guardrails and release controlsWeek 3
07A named analyst to sign the figure off→Sign-off workflow, then pilot and production validationWeeks 5–6
Nothing else is requiredDeployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Read the widths as weeks truly worked rather than as layout, which forces two phases into the fifth.
PhaseW1W2W3W4W5W6
DiscoveryW1
BuildW2 – W3
EvaluateW4 – W5
Pilot & LaunchW5 – W6
Week focusW1Reporting workflow discovery, method mapping and the automation boundaryW2Source integration and the reconciliation baselineW3Figure assembly, method logic and sign-off controlsW4Evaluation suite, restatement cases and failure-mode testingW5Warehouse integration, pilot figures and targeted correctionsW6One reporting year run under the analytics lead, then Agent Care handover
Reading the bandA bar spans only the weeks its own work is named in. Two phases land in the fifth because that is where they fall.
At the end of W6Validation closes on live reports, and Agent Care picks up the watch.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Marketing AI agent
Build a campaign analytics agent around the number your board pack has to defend.
Bring us one figure and both decks it appeared in. If your platform exports and your warehouse tables are squared by hand once the period closes, then that residual is carried on trust. Attribution and media-mix modelling stay with your analysts; the record is ours.
