A register, not a diary
0:07
A register, not a diary. Two rules decide what gets in.
0:12
ArcArc records every proposal made from the data at Acme Home Appliances, a demo company, and what happened to it.
0:21
Two rulesA proposal is admitted only if it names a metric the platform computes and a scope it applies to. Without both, nothing could ever be tracked back to it.
0:33
In reviewTwo proposals are in review, waiting on approvals.
0:38
LapsedOne was approved and never carried out, and nobody closed it.
0:44
Being measuredThree are being measured, against a registry of thirty-one metrics.
0:50
MeasurableMeasurable is the tier a scope earns. A synthetic control is strongest, then a comparison group, then movement only.
0:58
Compared againstCompared against is how many comparable units the change leaves untreated. Zero means nothing to compare.
1:06
Movement onlyOne approved decision can never be measured properly: one margin structure for all 470 distributors. There is a real reason to apply it everywhere, and it is still a trade, made now rather than discovered later.
How review made it measurable
1:21
How review made it measurable. The history behind a proposal.
1:26
ARC-014Auto-liquidate channel stock at 75 days, proposed by the analytics engineer after the 60-day price floor worked.
1:34
Two predictionsIt predicts two metrics from the registry: days of cover down six days, and gross margin down four tenths of a point. The downside is predicted as plainly as the upside.
1:47
The first draftThe first draft was ninety days, across all 470 distributors.
1:54
ReviewReview moved the threshold to 75 days on the evidence, and Finance asked for one region so it could be measured at all. It became the West: 120 treated, 350 to compare against.
2:07
The valuable partThe gap between the first draft and the record is where the thinking happened. Most registers keep only the final version.
2:20
Considered, and why notARC-016 records what else was considered: a regional pilot, rejected because the national chains buy everywhere from October.
What the register keeps
2:29
What the register keeps. Rejections, lapses, and what it will not accept.
2:35
Rejected, keptDoubling retailer schemes everywhere was rejected: the competitor lift it cited was confounded, and every region with no comparison group would spend the money and still not know.
2:47
Worth keepingThe reasoning in a refusal is worth as much as an approval, and it is the first thing most registers throw away.
2:59
LapsedA same-day call after every installation was approved in March and never started. The areas were never told, and the measurement window has closed.
3:12
Turned awayAt the bottom, what Arc will not accept: refreshing the brand, improving morale, hiring a CFO. Not bad ideas, just not decisions this register can measure.
Propose something
3:24
Propose something. The gate, live, while it is still a proposal.
3:31
A new proposalPropose something opens the gate itself.
3:41
RefusedName a metric the platform does not compute, and the form says so as you type: nothing could ever be tracked back to it.
3:54
Movement onlyunits_sold is in the registry. Across all 470 distributors, though, nothing is left to compare against, so the best anyone could say later is that the number moved.
4:07
A comparison groupNarrow it to the West, and 350 distributors are left to compare against. The verdict moves to a comparison group before anything is sent.
4:20
In reviewThe proposal is recorded as ARC-017, in review, with its scope, its prediction and the approvals it needs.
4:28
Two approvalsIt needs two approvals, from strategy and from the CFO. Both are pending.
4:35
Before and afterIts baseline is the last full quarter, and it will be measured over the next one, against the 350 distributors outside the West.
4:46
Three in reviewThree proposals are now in review.
What they did
4:50
What they did. Leap reads the decisions that were carried out.
4:57
UnpredictedLeap reads what each carried-out decision did. Every settled one had an effect nobody predicted.
5:05
ValueThree decisions measured, worth ₹4 crore between them after the costs that were found.
5:13
VerdictsEach gets a verdict, from big leap forward to big step back, and nothing at all while the window is still open.
5:23
CertaintyHow well we know comes from the tier the scope earned. A decision applied everywhere can only ever be indicative.
5:34
ARC-010Motor safety stock at Hosur, raised to four weeks on six lines, measured against fourteen lines left as they were.
5:43
PredictedIt predicted four percent more output. It delivered 4.4%, while the untouched lines moved 0.9%. A clean win on its predictions.
5:52
Not predictedIt also added ₹1.8 crore of raw material held, which nobody predicted. It was only caught because the proposal cited raw material as evidence.
6:03
NetNet of that, ₹2.4 crore. The side effect took back about 43% of the gain.
6:13
ARC-008The price floor on ageing channel stock worked: margin rose 0.6 points against 0.1 among the untreated distributors. Days of cover came in worse than planned, which was accepted up front.
6:26
The slipVolume slipped 1.1% in the treated region. Nobody predicted it, and it eats roughly a fifth of the margin gain. Net, ₹1.4 crore.
6:39
The clean oneTwo extra engineers in Coimbatore, modelled against 41 other areas: both predictions landed, and the unpredicted effect was good. First-time fix rose three points.
Back where it started
6:50
Back where it started. Decisions on the home page.
6:57
What we decidedThe home page counts three proposals in review now, and ₹4 crore attributed.
7:06
In flightAnd the decisions in flight, each with its verdict and whether it cost something unplanned.