Marketing · Restaurant · Atlanta, Georgia
A $94 sales channel became a $600,000 one.
The short version
Bawarchi Atlanta had a two-month-old website taking $94 of takeaway orders a month. The obvious move was to buy traffic. The data said something more useful: the site was already receiving order intent it could not close, and roughly half of what the website earned in its first year came from fixing that rather than from adding demand. Ten years on, website takeaway runs at $50,000 a month on the same channel — through a pandemic and through three direct competitors opening in the same catchment.
Where it started
Bawarchi Biryanis Atlanta is part of the Bawarchi Group, the largest Indian restaurant chain in the United States. The Sandy Springs location serves North Indian, South Indian, biryani, Indo-Chinese and chaat, and runs a substantial catering business for weddings and large events.
In October 2016 — the month before we started — the website looked like this.
- Website sales
- $94.05
- Value added to carts
- $600.75
- Cart value converted
- 16%
Almost every takeaway order still arrived by phone. Management read the empty order log as a demand problem and wanted campaigns.
What the numbers actually said
They thought nobody knew the website existed. The carts said otherwise: people were finding it, filling a basket, and leaving.
October and November 2016 converted cart value into completed orders at 16% and 35%. Whatever demand the site was already getting, it was losing most of it at the ordering step — a menu organised the way a printed menu is organised, category depth that buried the higher-value items, and a checkout that asked for decisions in the wrong order.
That distinction set the sequence for everything after it. Paying to send traffic into a page converting at 16% would have spent the budget proving the site did not work. So the first money went into the ordering journey rather than the ad account.
Across 2017, cart value grew 22.8× while completed sales grew 44.7×. Hold conversion at the 2016 rate and the year produces $47,917 instead of $93,916 — about half of first-year revenue came from closing better, not from buying more.
Cart conversion peaked at 48% that year, on a small and mostly warm audience. It runs at 25–30% today, on a cart volume many times larger. That is the trade we made deliberately once the ordering journey was fixed: a narrow funnel converts better, a wide one earns more, and the revenue line is the one that pays for the kitchen.
Three decisions
01Rebuild the ordering journey before spending on media
We reworked the home page and the ordering page around order completion instead of presentation: recut the menu categories so higher-value items surfaced early, shortened the path to checkout, and rewrote the on-page copy for someone deciding what to eat rather than someone learning about a restaurant.
02Carry the risk on paid ourselves
When we did propose campaigns in July 2017, the client had no experience of paid marketing and no reason to believe in it. Rather than argue, we took ownership of the ad spend and agreed to be paid on what it produced. The first sales-focused campaign did $1,800 of takeaway orders in a single day.
03Own the repeat customer instead of renting them
Once the ordering base was large enough to be worth something, the work shifted from acquisition to retention: email to existing customers, campaign moments built around occasions, and products designed to bring people back rather than find new ones. Biryani Buckets launched in 2019 and sold out in two weeks.
The first year, month by month
The reason to publish this in full rather than as a headline multiple: the shape is compounding, not a spike. Nothing here is a good month cherry-picked out of a bad year.
Results
The reasonable question about a ten-year engagement is whether the early growth held once the channel got large. It did.
Nine consecutive years of growth on one channel, for one client, with the marketing budget never once scaling in step with the revenue it produced.
What these numbers do not include
Dine-in. Campaigns filled tables as well as order carts, and the restaurant expanded its floor space in 2020. None of that is in the figures above, because the restaurant has no mechanism to attribute a walk-in guest to a campaign. So the channel figures understate what the work produced. We do not know by how much and we are not going to estimate it.
Run rate, not a closed year. The current column annualises the website’s August 2026 monthly figure. The 2017 column is a closed calendar year reported month by month above. We have kept the two visibly distinct rather than presenting the run rate as audited history.
Conversion over time. The 48% figure belongs to 2017 and is not a current claim. Cart conversion today runs at 25–30%, on a cart volume many times what it was then. As the top of the funnel widened to colder, discovery-driven traffic, the completion rate came down while absolute revenue kept climbing. We report both rather than the flattering one.
Profit. Every figure here is revenue, not margin. Apply your own food and labour cost to anything on this page.
Channel attribution. The return figure is blended across paid, organic, email and repeat demand. It is a measure of what the whole marketing programme produced per dollar, not a paid-media attribution claim.
Our fees. We do not publish what we charged. Ask, and we will walk you through the full return calculation on this engagement, our fees included.
Currency. Fees were contracted in Indian rupees and converted at the rate prevailing in each period.
The market did not stay still
Ten years is long enough to stop being a growth story and start being a durability test. Four things happened to this trade area during the engagement.
- Jan 2020A direct competitor opened nearby. Hyderabad House — same cuisine, same catchment, same customer.
- Mar 2020Dining rooms closed. Website order volume doubled within the first two months of the shutdown, because the ordering channel was already built and already converting.
- May 2025A second direct competitor opened in the trade area. We raised the media budget that September and held the channel.
- 2026A third opened. Too recent to draw a conclusion from, and we will not pretend otherwise. Website takeaway is at $50,000 a month.
The claim we would make from this is a narrow one. A first-party ordering channel — your own website, your own customer list, your own data — holds its value when the market gets crowded in a way that a marketplace listing does not. A competitor can outbid you on a delivery app tomorrow. They cannot buy their way into your repeat customers’ inboxes.
Whose customer is it?
In 2016 this restaurant already had demand. It arrived by telephone, and it was growing. The question facing the business was never whether those orders would happen — it was who the person placing them would belong to.
That is the part of this engagement most easily missed. A marketplace listing would not have created Bawarchi’s takeaway customers. It would have collected them, kept their names, and rented them back to the restaurant one order at a time.
The commission arithmetic is the least interesting part of the argument, but it should be stated accurately — including the fact that it has moved sharply in the restaurant’s favour since this engagement began.
These are like-for-like figures. Marketplace commission is all-in — it already includes card processing — so the own-website row is its true comparison rather than something to be added on top. The difference column is what the restaurant actually kept.
For the first four and a half years of this engagement, the years in which the channel was built, collection through a marketplace would have cost roughly four times what running it directly did. The rate cut in 2021 narrowed that considerably. It did not exist when the decision was made.
Commission is also the floor, not the cost. Visibility inside a marketplace is bought separately. Sponsored listings, funded promotions, and discounts the restaurant pays for itself with a per-order marketing fee charged on top of the discount — independent operators commonly report a blended real cost in the region of 25% to 35% once processing, promotions and paid placement are counted alongside the headline rate. Those figures are delivery-weighted, so they overstate a collection channel; the direction is what matters.
That spend behaves differently from the same money spent on a channel you own. It rents attention inside an application that is showing your customer four alternatives on the same screen, at the exact moment they are choosing what to eat. Stop paying and the visibility stops with it. Search position, a mailing list and a base of repeat customers do not reset when the invoice does.
One thing marketplaces genuinely do provide is discovery: customers who would never have found the restaurant on their own. That is what the commission buys, and it is worth buying. The argument is not about the first order. It is about the second, the twentieth, and who has to pay for each of them.
Ten years of ordering has left this restaurant with something a competitor cannot buy, outbid or replicate: its own list of customers, with their order history, reachable directly and at no cost, for as long as the restaurant wants to reach them. That is the actual output of this engagement. The $50,000 a month is what it produces.
Commission figures are published US marketplace rates and are inclusive of payment processing: collection commission was reduced from 15% to 6% across DoorDash partnership plans in April 2021, with delivery currently at 15% Basic, 25% Plus and 30% Premier. The own-website figure assumes card processing at 2.9% plus 30 cents a transaction on an average ticket around $42, which works out near 3.6%. Blended cost figures for independent operators are drawn from published industry reporting rather than from this client’s accounts. Rates vary by market, platform, ticket size and agreement — check your own before relying on ours. The comparison assumes the same order volume through either route, which is a counterfactual rather than a measurement.
Ten years, in order
Open the full timeline+–
- Nov 2016Engagement begins. Home page and ordering page rebuilt; menu categories restructured around order value.
- 2017Organic social programme starts, pushing traffic to the website.
- Jul 2017First sales-focused paid campaign. $1,800 of takeaway orders in a single day.
- 2018Email programme built on the existing customer base.
- 2019Biryani Buckets launched. Sold out in two weeks.
- Jan 2020Hyderabad House opens in the same catchment — the first direct competitor of the engagement.
- 2020Restaurant expansion, including the Wall of Bollywood. Online order volume doubled within two months of the first COVID closures.
- 2020–22Catering treated as a separate funnel: dedicated campaigns, custom catering tags with caricatures of the bride and groom, an Instagram filter for events.
- 2023New custom-built website. Faster, better structured for search, higher traffic-to-order conversion.
- May 2025A second direct competitor opens in the trade area.
- Sep 2025Media budget raised in response, with spend concentrated on defending repeat order frequency.
- 2026Website takeaway at $50,000 a month. Same client, same channel, tenth year.
Constraints that shaped the work
The client had no experience of paid marketing and no reason to take our word for it, so the case had to be made with results rather than a deck. We put the first campaign on our own books and were paid on what it produced. That set the pattern for the decade that followed: prove it small, then scale what works.
The budget stayed lean by design. Small enough that testing at volume was never an option, which meant the thinking had to be right before the money went out. It is also the reason the media budget never had to grow in step with the revenue.
Attribution was the last piece to fall into place. For the first two years the split between phone and web orders was reconstructed by hand from point-of-sale records, month by month. Instrumenting it properly is the one thing we would now do in month one rather than year three — and it is how we open every engagement since.
In their words
“[Client quote to be sourced and approved. The strongest version speaks to the eight-month wait or to where the restaurant would be without the work — the counterfactual carries more weight from the owner than from us. Attribute by name and title.]”
Name, Title — Bawarchi Biryanis Atlanta
If this sounds like your business
In a food business where most orders still arrive by phone, channel shift is a menu-architecture problem before it is a media problem. Paying to send traffic into an ordering flow that loses two-thirds of its carts is the most expensive way to learn what your site cannot do. Fix the close rate first, then buy volume — the same budget buys roughly twice the revenue.
The reason to do it on your own website rather than a marketplace only becomes obvious later, when a competitor opens down the road and you find out what you actually own.
Read another
Every figure sourced from the client’s own systems, with the basis stated and the exclusions named.Smile Loft Dental
Sanpeggio’s Pizza
Bawarchi Kennesaw
Measurement note. Website figures are drawn from the restaurant’s own ordering platform and point-of-sale records. 2016 and 2017 monthly figures are reproduced from the reporting supplied to the client at the time. Marketing cost figures are Debox invoices for the period, inclusive of ad spend.
