Every cell travels the same protocol. The stage list is admin-owned on the master's Control sheet (tbl_Stages) and is the single authoritative vocabulary - notebooks derive it from the table, never hard-code it, and neither should you. Ten stages in three phases:

Each stage has a snake_case code used as a column suffix on the Delivery and Team_Structure sheets: design, eda_ads_prep, modeling_mmm, modeling_pnp, optimization, validation, insights_storyboarding, business_presentation, market_presentation, final_signoff.

The Delivery sheet is the program heartbeat: one row per cell, and for each stage a triplet of columns <stage>_status / <stage>_date / <stage>_notes. Count them and you find 11 triplets, not 10 - the extra one is data_signoff_*, a sign-off gate that is not a stage in tbl_Stages. The triplet sequence is:

data_signoff_*, design_*, eda_ads_prep_*, modeling_mmm_*, modeling_pnp_*, optimization_*, validation_*, insights_storyboarding_*, business_presentation_*, market_presentation_*, final_signoff_*.

Every <stage>_status takes one of six values from tbl_StageStatus: NA, Not Started, In Progress, Blocked, Done, Signed Off. NA is a legitimate value - e.g. a cell with no price-promo scope carries NA on modeling_pnp_status. The _date columns capture stage completion / sign-off timestamps; the cell-level milestones (preparation start, build start, target completion) live on the Cells sheet, not here.

What starts the 8-week clock

The Build clock does not start when someone begins working. It starts when both sign-off dates are set on the Delivery row: the Design sign-off (the design_* triplet reaching sign-off) and the Data sign-off (the data_signoff_* triplet). One without the other keeps the cell in Preparation. From the moment both are in, the cell owes the program six Build stages inside a committed 8-week window.

current_stage is derived, never stored

There is no current_stage column. Reports derive it at render time with one rule:

cell_status is the one manual per-cell input on Delivery, taking one of seven values from tbl_CellStatus: Awaiting Confirmation, In Preparation, In Build, In Presentation, In Support, On Hold, Closed. The middle five form the active progression; Awaiting Confirmation is pre-engagement (SOW/scope not yet locked); On Hold is the exception state.

The nuance that trips everyone, stated precisely per the master schema: cell_status stays "In Build" only while Build-phase stages (EDA & ADS prep through Insights & Storyboarding) are running. It flips to "In Presentation" once all six Build stages are Done or Signed Off and only the three Presentation stages remain. Presenting interim results DURING the 8-week window is not a status change - it lives in the per-stage statuses and surfaces as a derived "Presenting" chip in delivery.html. An automated status_mismatch finding flags any cell whose cell_status disagrees with its Build-phase stage statuses, so getting this wrong is visible.

The Data sheet is the Preparation-phase data-readiness inventory: one row per cell, one column per data type. The 14 feed types come from tbl_DataTypes: sales, kantar_sales, ecomm_sales, distribution, promotion, competition_sales, traditional_media, digital_media, boosted_influencer, organic_influencer, ecomm_media, competition_media, bgs, macro. Each feed cell takes a value from tbl_DataStatus:

• Happy path: Pending → Received → Validated
• Off-ramps: Granularity Issue, Blocked
• Deliberate exclusion: Not in Scope (exact casing - lower-case "in")

A per-cell free-text data_summary one-liner carries the acquisition narrative (blockers being chased, expected receipt window, granularity caveats). The tie-in to the lifecycle: when all 14 feeds for a cell read Validated (or a deliberate Not in Scope), that is the signal the cell is ready for the Data sign-off gate on Delivery - and once both Design and Data sign-off dates are set there, the 8-week Build clock starts.

A synthetic cell (no real cell resembles it) shows the following on Delivery. Read it, then answer the three questions in Check Yourself before peeking at the explanations.

Walk the rule: is any stage In Progress or Blocked? Yes - Optimization is the first such stage, so current_stage = Optimization (the NA on Modeling - PnP is skipped, not counted as pending). Both sign-off dates are set (2026-03-02 and 2026-03-06), so the 8-week clock started - it has been running since the later of the two. And with Build stages still running, cell_status = In Build - even if this cell showed interim results to the market team last week.

The methodology workbook's Process sheet documents a 28-step team-handoff SOP - Data Team → Validations Team → Modeling Team, with per-step ETAs for the model-run loops (and placeholder rows for a Price and Promotions Team section that was never authored). It is genuinely useful color on how the build actually flows day to day: the ADS handoff, the bounds-in / results-back loop between Validation and Modeling at every model level, and the closing "update workbook, then deck" steps.

Caveat: that sheet predates the current tbl_Stages naming and describes team handoffs, not the program's stage protocol. Where they disagree, MASTER_SCHEMA.md wins - treat the SOP as context, not as the lifecycle you report against.