Action & Value Recovery
Proof, not assertion — value recovered after the decision date
Every action carries a decision date. From that day the same metric is tracked against a frozen pre-action baseline, and the gap between the two is the recovered value. Pick an action to see its curve.
Action register
Baseline tracking at the grain the client asked for — how many are done, who owns them, and what each one has returned since its decision date.
Monthly impact ledger
The month-on-month record. This is the artefact that answers "what did the platform actually change?" a quarter after launch.
100-day rolling improvement, split by who owns the fix
The gradual uptick, measured. Total loss run-rate is one line; the two bands underneath are the only split that tells you who to talk to next — upstream is Supply Planning, downstream is the CAM.
Three splits, one budget
The client's question in their own words: how do I split ₹100 between Blinkit and Zepto, then within a platform between promo and search, then across brands. Grey is where the money sits today, colour is where the model puts it.
Competitor ad investment — estimated
Modelled from crawl history: sponsored-slot occupancy × observed slot rates, indexed to our own spend at 100. Directional, not billed spend — and labelled that way on the axis.
Keyword verdicts
Where the search money should move. A keyword we already own organically does not need defending; one we are losing while a competitor bids is where the next rupee goes.
What does the next rupee buy?
The blended return is an average of every rupee already spent. The question on the table is a different one — what the next rupee returns — and the answer is always smaller.
Detected launches
Pick one to open its bio. Everything in the bio is assembled automatically from the detailed PDP crawl plus the ads connector — no manual report.
Text to alert
Type the alert as a sentence. The parser turns it into a rule you can see and edit — so the fluency never costs you the ability to check what was actually set.
Saved rules
Every rule shows what it matches now beside what it fired over 30 days, so drift is visible without opening anything.
Review brand feed
Not a rating number — the themes underneath it. What our products are failing on, and what is working for the competitor set, from the same review corpus.
What shipped this quarter
Delivered in-product rather than as a servicing call. Each note carries a short walkthrough and the result it produced for a client, and a single checkbox to add the feature to your own dashboard.
Initiative register
Every intervention we are running, what it was meant to move, and what it cost somewhere else. A row is only green when both numbers agree — a target metric moving while its counter-metric breaks is not a win.
The trade, measured
Both sides of the transfer on one axis, from the decision date. Share change is shown in points against each geography's own base, and the bold line is the net after weighting each side by the size of its category.
Counter-metric watch
The guardrail set attached to each initiative at the decision date. These are chosen up front, not after the fact — that is the only way a breach means anything.
Pause recommendations
Where the arithmetic says stop. Each card states what would be restored, what it costs to unwind, and who signs it off.
Campaigns
Select a campaign to open its keyword ladder. Every campaign resolves down to the keywords it bought and the SKUs those keywords served — the link back to search that was missing.
Keyword ladder
—
Organic-to-paid handoff
A keyword should graduate. Paid support buys the rank; once organic holds rank 3 or better for six straight weeks, the paid line is doing nothing the page would not do on its own. Grey band = paid support, line = organic rank.
Price compliance
Live price against the approved floor for each SKU × platform. Platform-led discounting shows up here as clearly as our own — the platform funding it does not make the shelf price compliant.
Price history
—
Recommended price change
What the agent proposes, and what it costs. Every candidate is priced out on all four numbers — units, revenue, gross margin and share — because a price that wins share can still lose money.
Approval & write-back
The agent drafts the change and prepares the payload. It cannot approve its own proposal — a named human signs it, and only then does it queue to the platform. Every step is written to the audit trail.