New Resort-Wear Label Scaled from Launch to ₹34.4 Lakh in 66 Days by Arlox.io Using Meta Ads.

At-a-Glance Stats

Benchmark context: The 2026 median Meta Ads ROAS for fashion and apparel brands is 2.18x; the top 25% reach 4.4x and the top 10% reach 6.0x. SEVEN's 66-day blended 6.71x clears the top-decile threshold. (Source)

What the platform would have told you instead. Over the last 30 days of this engagement, Meta Ads Manager reported 247 conversions at 4.72x ROAS. The client's Shopify export recorded 328 orders at 6.42x. Platform attribution under-counted real orders by 25% and real return by 26%. Delayed attribution.

1. Before Arlox.io

Direct answer: SEVEN was a genuine cold start. When Arlox began, the brand had a live Shopify store, a finished catalogue but no media history to begin with. There were six organic orders totalling ₹32,588 across its first month, no paid programme, no pixel conversion history and no audience data of any kind.

Surbhi Gupta built SEVEN as a resort wear label that included coordinated sets, tunics and dresses in bold prints, priced between ₹2,499 and ₹9,999 with matching sets clustered at ₹4,998 to ₹6,998. The product was ready and even the photography was strong. The Shopify store went live in May 2026 and took its first real order on 21 May.

But the momentum wasn't there.

Between 20 May and 19 June, the store produced six orders and ₹32,588 in revenue, roughly one order every five days. It was never a failing business. Just a business with a lot of potential that had not started yet. Products were never an issue in fact even now SEVEN’s products are one of the most consistently virtually tried on daily basis. The virtual try on and its data is explained in Section 8 along with the introduction to Style Me our proprietary one click virtual try on app.

This is a materially different starting position from most agency case studies, and it matters for how the numbers should be read. There was no existing revenue baseline to grow from, no warm customer list to retarget, no lookalike seed with purchase depth and no historical creative to learn from. Every audience signal, every conversion event and every rupee of revenue in this case study was done from zero within a timeline of 66 days.

2. What Made This Hard But Exciting To Work With

Direct answer: A cold start fashion brand faces the single most competitive and highest CPM category on Indian Meta, with none of the algorithmic advantages that make that competition survivable.

Fashion and apparel is the largest and most crowded D2C market in India. It accounts for roughly 22% of India's D2C GMV. It is also the highest CPM vertical on Meta more precisely because so many brands are bidding for the same attention.

A brand entering that auction with an established pixel has advantages: Meta already knows what its buyers look like so the algorithm can find more of them cheaply.

There was a second, less obvious constraint. Seven Shop’s average order value would need to land above ₹5,000 for paid acquisition to work at all. At a ₹2,000 AOV in a high CPM category, the maths simply does not close in India rather acquisition cost eats the margin before the order fulfilment. The brand's price architecture made profitable paid acquisition possible but it also considerably narrowed the buyer pool. Meta would need highly specific signals to find a shopper willing to spend ₹5,000+ on a co-ord set from a label they had never heard of.

And the category's structural problem in India is Cash on Delivery. Fashion and apparel carry the country's worst RTO rates: 26-35% RTO on COD orders. Against 4-8% on prepaid with the national D2C average sitting at 20-30%. (Source) In this case 30% RTO rate would have made every ROAS figure in this study fictional.

3. What We Decided Before Spending Anything

Direct answer: We spent the first two weeks on research and development rather than rushing for campaign launch: making target avatar, CRO audits and payment configuration. One of the most unexpected yet calculated decisions that shaped every number that followed: Seven Shop would sell prepaid only.

Research came before spend. The target avatar and CRO report was produced and delivered on 5 June, fifteen days before the first rupee of ad spend. Shopify access was resolved on 8-10 June. Website changes were audited on 10 June and chased on 13 June. Payment methods were configured across 17-18 June. Campaigns went live on 20 June.

The prepaid only decision. This is the most consequential call in the engagement and it was made before launch. Every one of 667 orders to date has been prepaid, 665 settled through Razorpay by card, UPI, net banking or wallet, and 2 by store credit. None of the orders are COD. And yet zero order cancellation.

The trade off was explicit and we accepted it. Disabling COD in India significantly suppresses the conversion rate because a meaningful share of buyers especially in tier 2 and tier 3 markets will not prepay to an unfamiliar brand. We knowingly gave up those orders.

What we bought in exchange was a clean optimisation signal. Every conversion event Meta received was a paid conversion, not a COD order that might be refused at the door. In a category where COD generates 76-83% of all RTO volume (Source) that decision removed the single largest source of algorithmic pollution before it could start.

It also means the ROAS figures in this case study are how they look. A 6.71x ROAS on a 30% RTO order book is actually a 4.7x ROAS. In SEVEN’s book 6.71x is 6.71x.

What we chose not to do. We did not run a sitewide sale to buy early conversion volume. A standing storewide discount is the fastest way to make a cold start pixel look healthy and the most expensive way to train an audience.

What SEVEN did run is a standing first order incentive: FIRST10 at 10%. It is on 350 of the 667 lifetime orders, approx 52.5% and it accounts for ₹2.09 lakh of the ₹2.62 lakh in total discounts or 7.1% of gross merchandise value. All revenue reported in this case study is already net of it. A second code MITALI15 ran alongside an influencer placement from 11-23 July and carried 17 orders.

The distinction that matters is first order only versus sitewide: a new customer incentive does not teach the existing base to wait for a sale. Worth noting it did not buy basket size either using FIRST10, average ₹5,371 against ₹5,734 for undiscounted orders. The one time a deeper, storewide offer ran, it was time boxed to three days Section 6 covers what that proved.

4. What We Built

Direct answer: A two layer campaign architecture feeding a daily creative testing loop with every asset treated as a falsifiable hypothesis and every decision logged the day it was made.

Layer 1 ABO manual prospecting (cold traffic). New angles were tested against cold audiences at ₹1,000-₹1,500 per day per ad set. The purpose of this layer was not efficiency. It was to manufacture intent, “add to cart”, “checkouts initiated” and “purchase events”. Assets that produced zero conversions on roughly ₹1,200-₹3,400 of spend were paused within 48 hours.

Layer 2 Advantage+ shopping (warm harvesting). Launched 26 June at ₹1,000 per day, targeting cart and checkout abandoners generated by Layer 1. This layer converted intent that the cold layer had already paid to create.

The relationship between the two is the architecture, and it is easy to get wrong. On 8 July the ASC campaign returned 23.6x ROAS and on 14 July, it returned 24x across eight conversions.

The obvious response is to conclude ASC is the winner and trim funds on cold prospecting. The obvious one is not always the right option, it is just a mere reflection of what the majority is doing and as a result what an individual is conditioned for.

We did the opposite: ABO stayed funded, and ASC's budget stayed bounded. When ASC's cost per acquisition rose to ₹1,200 against a ₹900 target on 23 July, its budget was downgraded to ₹1,500/day. The warm pool is a finite resource that depletes faster than it refills, and treating it as a growth engine rather than a harvest is how promising accounts stall in week three.

The creative system. Twelve distinct angles were produced and tested across reels, statics, carousels, UGC, founder led content, and influencer content:

The decision that best explains the method. On 9 July the Pink Guava reel was producing a 22.85% video hook rate, elite by any standard, alongside ₹62.92 clicks and zero cart or checkout actions. It was paused that day. A strong hook attached to a weak angle produces valueless virality: people watch and nobody buys. Protecting that asset because its engagement metric looked impressive would have cost real money for diminishing ROI.

Nine assets were killed across the engagement. It is the cost of the search that produced the winners and the cost was held between roughly ₹1,200 and ₹3,400 per disproved hypothesis.

5. The Turning Point

Direct answer: The account did not spike, it compounded. Daily revenue in the second half of the engagement ran 40% ahead of the first half, on the same architecture, driven by creative replacement rather than budget escalation.

Split the 66 days at the midpoint:

+40% daily revenue, second half against first. That is the difference between an account that found a lucky creative and an account running a system. The winners in the second half are not the winners from the first: Palm Pop, Vacation Edit, and the Cutout Reel had all been retired by then, replaced by founder content, UGC, and bundle-led collection reels.

Fatigue was hunted rather than waited for. Eight separate creative fatigue events were detected and acted on inside the window, each caught on a leading indicator: hook rate, hold rate, outbound CTR, CPM inflation rather than on a ROAS collapse:

The Vacation Edit Reel is the cleanest illustration: video engagement was flagged as deteriorating on 3 July, six days before the return metric actually fell. That six days is what leading-indicator monitoring buys.

The August headwind was real. On 7 August, CPMs inflated over 66% to ₹1,850+, doubling CPCs to ₹29.85, a genuine auction environment shift. A 66% CPM increase halves the traffic a fixed budget buys. Holding a ₹900 CPA target through that requires a conversion rate improvement to offset the traffic cost increase, which is exactly what Section 6 describes.

The last 30 days, day by day. Volatile at the daily level, strong in aggregate, including the days that went badly.

The weakest day in that table barely returned the money spent. It is explained in Section 6, and it was not hidden from the client on the day it happened.

6. The Three-Day Offer, and What It Actually Proves

Direct answer: A time-boxed 15% discount ran from 14-16 August to offset the August CPM inflation. When it was switched off on 17 August, revenue fell 89% overnight from 16 orders and ₹80,725 to 2 orders and ₹8,997, on the same creatives, audiences and budgets.

The offer was not a calendar promotion. It launched a week after the 66% CPM inflation was logged, and it was engineered to buy conversion rate to pay for a more expensive auction.

For three days the account absorbed a 66%-inflated auction and still held cost per purchase below target. Then the offer came off and the account barely broke even on the day.

7. The Results

Direct answer: ₹5.13 lakh of ad spend produced ₹34.43 lakh of revenue across 623 orders in 66 days, a 6.71x blended ROAS at ₹823.83 cost per purchase, on a 100%-prepaid order book with a 1.8% refund rate.

Revenue trajectory. From six organic orders in the month before launch to a ₹20.1 lakh/month run rate by August, with July delivering ₹17.89 lakh and the first 26 days of August delivering ₹16.88 lakh. The brand did not exist commercially in May. It was a ₹20L/month business by August.

ROAS against category benchmark. SEVEN's blended 6.71x sits above the top 6.0x threshold for fashion and apparel on Meta, against a category median of 2.18x. (Source) That is roughly 3.1x the median brand in the most competitive D2C category in India.

Click through rate against category benchmark. The account's last 30 day CTR was 6.76% against an apparel benchmark of 1.24-2.84% (Source) between 2.4x and 5.5x the category average. Creative quality, not bid manipulation, is what produced the CPA.

The RTO number that isn't there. Indian fashion D2C runs 26-35% RTO on COD orders and 20-30% nationally. (Source) SEVEN's exposure is zero, because there are no COD orders to return. Of 667 lifetime orders: 665 paid through Razorpay (cards, UPI, net banking, wallets; 8 of them part-settled with store credit) and 2 through store credit alone, 0 cancelled, with 12 refunded a 1.80% refund rate by order count and 1.77% by value, against an Indian fashion category benchmark of 8-15%.

Unit economics. ₹823.83 cost per purchase against a ₹5,525.75 AOV means acquisition consumed 14.9% of order value. At the ₹2,000 AOV common in Indian fast fashion, the same CPA would have consumed 41% and the model would not close. The brand's price architecture and the media buying were designed against each other deliberately.

What 6.71x actually means CM1, CM2 and breakeven. A blended ROAS is a platform number. It says the account returned ₹6.71 for every rupee of media, and it says nothing about whether the business made money, because it ignores every variable cost sitting between the sale and the bank. The number that answers that is breakeven ROAS the return a brand needs simply to stand still and it falls out of contribution margin.

SEVEN's COGS is the brand's own figure and not ours to publish, so we are not going to state a CM2 number. What we can state is every other line, because they come out of the shopify orders data:

Two things sit outside this table. Discounts are already netted out of the AOV above, a further ₹2.62 lakh, 7.1% of gross merchandise value, was given up before we get to ₹5,525.75. And shipping is absorbed: across 667 orders the store charged customers ₹200 in total.

The breakeven a brand needs at ₹10 lakh a month is not the one it needs at ₹1 crore, which is why founders pricing their unit economics off current scale so often conclude that scaling is impossible. You can model your own against whatever your COGS actually is with our free BROAS calculator.

One note on vocabulary, because the conventions genuinely differ: we use the Indian D2C convention, where CM1 is net revenue less COGS, packaging, payment gateway, last mile logistics and returns processing, and CM2 is CM1 less acquisition cost. Some operators, particularly outside India, define the levels differently. If your P\&L uses the other convention, translate before comparing.

Catalogue concentration. A 29 product catalogue with 282 active variants sold 28 distinct products inside the window, and the top 10 account for 77.8% of revenue. The electric sangria set alone accounts for 25.5%, ₹9.57 lakh across 172 units. Try-on demand and purchase demand both cluster hard on visually complex coordinated pieces. Section 8 shows how tightly.

Customer base. 637 unique customers, 24.1% of orders containing more than one item, and a 4.24% repeat rate inside a 66 day window early, but real. Orders concentrate hard in Maharashtra (42.7%), and inside it Mumbai alone (38.1%), with Delhi (10.8%), Haryana (8.9%) and Karnataka (7.9%) forming the secondary tier.

What Meta would have told you instead. Over the same last 30 day window, Meta Ads Manager reported 247 conversions at 4.72x ROAS. The Shopify export recorded 328 orders, and ₹18.87 lakh of revenue against Meta's own ₹2.94 lakh spend figure of 6.42x. Platform attribution under counted real orders by 25% and real return by 26%.

Most agencies report the platform number because it arrives without work. We report the Shopify number because it is the one the founder's bank account agrees with, including on 17 August, where the store's number is the less flattering of the two and our own spreadsheet was wrong.

The one gap the ad account could not close. Across the window a single signature recurs: strong intent generation, weak completion. On 6 July, 28 add to cart produced 2 orders. On 18 August, 19 add to cart and 8 initiated checkouts produced 5. Across 21–22 August, 70 add to cart and 48 initiated checkouts produced 13 orders. 35 people reached the payment step and did not finish.

Traffic that reaches an initiated checkout has already decided it wants the product. Every creative angle in Section 4 attacks the hesitation from the ad side, the founder reel answers it with trust, UGC with peer proof, the 2-in-1 carousel with versatility. None of them are still working at the checkout. That gap is what Section 8 is about.

8. Where StyleMe Fits

Direct answer: StyleMe is Arlox's one click virtual try-on app, embedded directly in the product page. On SEVEN it produced 513 try-on generations across 47 consecutive days at a 98% success rate. It is a friction reduction layer, not a revenue attribution layer and we do not credit it with revenue.

SEVEN sells coordinated sets. The dominant pre-purchase question in that category is not price and it is not delivery, it is "will this actually look good on me?" A flat product photograph is the worst available answer, and a two-piece coordinated set is the hardest case, because the buyer has to imagine two garments on their own body.

StyleMe answers it directly: in one click the shopper uploads a photo and sees themselves wearing the piece on the product page.

Between 11 July and 26 August, on SEVEN:

Try-on demand tracks revenue. Ranking products by try-on volume against their in window revenue rank:

Seven of the top eight try-on products are also top seven by revenue. That is a correlation, the same shoppers who try on are more likely to buy.

The exception. The coastal clash set is third in try-on demand and ninth in revenue. High consideration, low conversion: a large number of people wanted to see themselves in it and comparatively few bought. That is a specific, actionable diagnosis about one product: pricing, fit or variant availability.

Why do we not put a revenue number on this? StyleMe is not an attribution tool and we are not going to pretend it is one. What it demonstrably does on SEVEN is hold people on the product page, give them a reason to come back, and tell the brand something about its catalogue it did not already know, 332 people generated 513 looks across 47 consecutive days, 491 of those looks were saved, and the demand curve across products pointed at a specific merchandising problem.

9. What This Meant for the Founder

Direct answer: Surbhi went from having no idea whether paid acquisition could work for SEVEN to knowing exactly what each rupee buys and being able to plan against it.

The quantitative shift is on the record: zero to ₹20 lakh a month in about ten weeks. The structural shift matters more.

Before this, SEVEN had a product and a store. There was no way to answer the only question that mattered: can this brand be bought profitably? because nobody had ever tried in a way that produced usable information.

Now there is a number attached to every lever. A customer costs ₹824 to acquire and spends ₹5,526. Founder led creative outperforms influencer creative by roughly ten to one on this audience. Bundles lift order value; discounts lift conversion rate and cost margin. A creative has roughly three productive weeks before hook rate starts telling you it is finished. When CPMs inflate 66%, the account needs a conversion rate lever or it needs to spend less.

None of that was knowable in June. All of it is knowable now, and it was produced by 66 consecutive days of running the same loop and writing down what happened.

The question is no longer is this working? It is which lever do we pull next month and that is a fundamentally different position for a founder to be standing in.

About Arlox

Arlox is a performance marketing agency for D2C fashion brands, working on profitable ad scaling, RTO reduction, COD-to-prepaid conversion, and getting brands CM2-positive at scale. This is one of 450+ brands we've worked with, and hundreds of those founders have recorded on-camera interviews about their results. Founded by Varinder Singh Gakhar (Vann Laniakea).

We work out of Gurugram, which, as it happens, is where 4.9% of SEVEN's orders ship. If you would rather hear it from the founders than from us, hundreds of them have recorded on-camera interviews, and they are on our YouTube channel.

Two of the tools referenced above are free to use: the BROAS calculator, which models how the return a brand actually needs moves as it scales, and StyleMe, the store-embedded try-on tool in Section 8.