Niti AI / Industries / D2C & ecommerce

D2C budget & growth decisions · outcomes measured

Your agency reports 3.2× ROAS. Your Shopify revenue confirms 1.9×.

Niti is a revenue decision system for D2C and ecommerce: it reconciles channel-reported ROAS with store-confirmed contribution, then ranks budget, pricing, and inventory actions — each with estimated impact, margin gates, and outcomes measured at T+7 to T+30. For India and US founders, performance heads, and CMOs spending ₹15L–1Cr+ (or $150K+) a month, where the weekly deck and the bank are quietly 60–70% apart.

The review nobody wants to run

A funded D2C brand was spending ₹80L a month across Meta and Google. The agency's blended ROAS: 3.2. Independent server-side computation against Shopify: 1.9. Scale decisions had been made for months on numbers 68% higher than what the store actually confirmed.

Meanwhile a fashion brand scaled a denim campaign to ₹6.2L over eight days. The SKU went out of stock on day three. Nobody found out until day eight, when returns started spiking — ₹6.2L spent sending traffic to a page that couldn't convert for five of those days.

And the new Head of Performance who just joined? They paused the WhatsApp retention campaigns to "test the channel properly." The previous team had already run that exact test. The finding was in a deck nobody could find. The repeat cost ₹8L.

What it costs

None of these show up as a single red number. That's why they survive.

68%
Agency-reported ROAS overclaim against server-side, confirmed on a ₹80L/month account. Not an estimate — a measured gap.
₹6.2L
Spent scaling a single campaign into an SKU that went out of stock on day three and kept running for five more days.
35%
Share of one ₹100Cr brand's marketing spend concentrated on SKUs under 20% gross margin, discovered at the CFO's first budget review.

Each of these was invisible inside the tool that was supposed to catch it — the ad platform dashboard, the agency deck, the Shopify analytics view. None of them are built to check spend against margin, stock, or a second data source. Niti is built to do only that.

What changes

A decision queue, not another chart.

"Which ROAS number is real?"

Server-side attribution, per SKU, per channel

True ROAS computed on eligible days only — out-of-stock days and sale-distorted days excluded — so the number isn't quietly inflated by periods your ad platform can't see the difference on. The overclaim, per channel, becomes a number your CMO can put in a board deck.

"What do I actually do about it?"

SCALE / REDUCE / HOLD / PAUSE, with a number attached

The decision queue is constraint-gated: nothing is recommended to scale without checking margin and stock cover first. Every recommendation carries an estimated ₹ impact — not a suggestion, a specific number you can hold the system to.

"Is this SKU about to waste my spend?"

The out-of-stock gate

Campaigns are automatically blocked from scaling once a SKU has under seven days of velocity-estimated cover. Named campaigns, estimated daily waste, before the ₹6.2L mistake happens — not eight days after.

"Am I spending against my own margin?"

SKU contribution grading

Every SKU graded A+ through NR on contribution (average ROAS × effective gross margin, on eligible days). The inversions — heaviest spend on your worst-margin SKU — become visible in one screen instead of a CFO discovering it at a budget review.

"Is my ad spend actually driving marketplace sales?"

Marketplace halo, made visible

A media-mix model with a channel-contribution Sankey: ad spend on one channel to revenue on another, including the Meta-to-Flipkart halo effect most tools can't see because they only read one settlement file.

"Was that a bad week, or an external one?"

The variance defense report

One click: the external factors — CPM shifts, competitor promos, demand index moves — behind an anomalous week. Built for the CMO who needs to defend a number in a leadership meeting without guessing.

The gate that matters here

Hard gate · stock & margin

Nothing scales on a SKU you can't fulfil or can't afford to sell.

The decision queue checks two things before it will ever suggest SCALE: velocity-estimated stock cover, and contribution margin. A campaign that clears both still has to earn its recommendation on marginal ROI. One that fails either is suppressed and named — not left for someone to notice on day eight.

Proof, not promise

Every decision is tracked to an outcome. Nothing is graded on a demo.

On a live account, 53 decisions have been tracked end to end: 12 favourable, 38 neutral, 3 unfavourable. That's the real distribution — including the ones that didn't work, which is the part most vendors don't show you.

  • ₹2.33L in revenue confirmed from scaled adsets, attributed and tracked, not estimated after the fact.
  • ₹21.5K/day in budget freed from 24 adsets that were reduced or paused on evidence.
  • +₹3.24L in scale opportunity identified across 14 adsets in a single review — headroom that was sitting unused.
Outcome classCountWhat it means
Favourable12Confirmed positive result against the estimate
Neutral38No material change either direction — held or logged
Unfavourable3Didn't work. Kept in the record, not hidden.

Live account data, illustrative labelling. The three unfavourable outcomes are the ones that make the other fifty credible.

FAQ

D2C & ecommerce — common questions

What decisions does Niti make for ecommerce brands?

Budget reallocation, scale/reduce/pause calls, discount and pricing moves, and inventory-aware spend gates — each with an estimated contribution impact and a measured outcome window.

How do you handle platform ROAS vs Shopify reality?

Niti reconciles channel-reported performance against store-confirmed contribution so scale decisions are made on the number that hits the bank, not the number in the agency deck.

What is the 45-day pilot?

$1K activation + $2K/mo. If no confirmed favourable outcome lands within 45 days, month two is credited. Activation covers real setup work and is non-refundable.

What if our catalog data is messy?

Use Services for onboarding and COGS/identity cleanup, or start with the free margin audit.

Explore

Where to go next

The 45-day Platinum Pilot

$1K activation. $2K a month. If nothing's confirmed favourable in 45 days, month two is free.

Activation is real analyst work — margin config, API connections, campaign-to-SKU mapping, and your first tracked decision — which is why it's non-refundable. The guarantee sits on the subscription, not the activation: if no confirmed favourable outcome is logged within 45 days, month two is credited. We charge full price because we know what this does; the guarantee handles your risk, not ours.

Show me the decision queue on my data