Sugarland Bakery: 92 Days of Ads, and 100% of Reviews Answered
A single Asian bakery in Madison, NJ switched on every INFI surface in five months — Performance Max ads, a self-order kiosk, INFI online ordering, organic search and AI review replies. Here is what each one measurably did, reported separately.
- Client:Sugarland Bakery
- Industry:Asian & Japanese bakery — cakes, roll cakes, pastries, bubble tea
- Locations:1 — 57 Main St, Madison, New Jersey
- Point of Sale:Square
- INFI products:ORDER (self-order kiosk), ONLINE, SITES, GROW (ads, Google profile, AI review replies)
Overview
Sugarland Bakery is one shop on one Main Street in Morris County, New Jersey, selling Japanese-trained pastry — roll cakes, Basque cheesecake, cream puffs, pineapple cakes — alongside a full bubble-tea menu. It is not a chain. It has one counter.
Between April and September 2026 it switched on every layer INFI makes, on top of the Square register it already ran: a self-order kiosk, INFI online ordering, a new website on the domain its Wix site used to occupy, Google Business Profile management, and Performance Max advertising. That makes it an unusually complete picture — and four different kinds of evidence, each of which can be checked on its own terms.
So this page reports them side by side, and does not braid them into a single causal story. Where we cannot separate two effects, we say so.
Five months, one surface at a time
One detail from the migration is worth keeping, because it is structural rather than cosmetic: the Wix home page shipped 1,135,218 bytes of HTML with 21 external script files. The INFI page ships 87,289 bytes and two. That is document weight, not a speed score — but it is the difference between a page assembled in the browser and a page that arrives finished. The old home page's title tag, meanwhile, was still the builder's default: “Home | Online Bakery.”
Paid: what 92 days of Performance Max actually bought
Three campaigns have run for this bakery, all Performance Max, all built and published by INFI. The current one has run continuously since 18 June 2026 on a $5-a-day budget.
| All three campaigns, 1 May – 17 Sep 2026 | |
|---|---|
| Impressions | 146,972 |
| Clicks | 6,635 |
| Click-through rate | 4.51% |
| Average cost per click | $0.092 |
Those clicks did specific things, and the specific things are the point. Counted by the action Google recorded, not by a single blended “conversions” number:
Two things about the delivery are worth an operator's attention. 98.9% of impressions landed in New Jersey — for a bakery whose trade area is a few towns wide, almost nothing was spent talking to people who could never walk in. And 94.5% of impressions, with 99.0% of the spend, went to phones: this is a channel that reaches somebody standing on a sidewalk deciding where to go, which is also why directions and menu views outnumber everything else.
The campaign also got cheaper as it learned. August and July cost within eight cents of each other — and August bought 32% more impressions.
INFI generated all 31 creative assets on the live campaign — headlines, long headlines, descriptions — and supplied every image from the bakery's own Google photos rather than stock. Google approved them all: not one asset was disapproved or limited. The copy also sharpened between campaigns. The first, in May, had three headlines and sold a soft-opening promotion. The current one names the actual pastry case: “Famous Meat Floss Cake,” “Basque Cheesecake & More,” “Fresh Roll Cakes Daily.”
And because Performance Max reports the demand it matched, it doubles as a market survey. “bakery near me” alone drew 5,379 impressions. The campaign also surfaced against competitors' names, and against 545 impressions of Spanish-language searches — panadería cerca de mí, pastelería cerca de mí — which the English-only ads have never spoken to. That is a real, measured demand pocket sitting unserved.
Owned: where the orders come in
The kiosk went in first, in April, and it has been quietly taking a growing share of the counter ever since.
| Rolling 30-day windows | Kiosk share of all orders | Kiosk orders per day |
|---|---|---|
| Window 1 | 21.8% | 9.4 |
| Window 2 | 22.0% | — |
| Window 3 | 21.9% | — |
| Most recent | 27.3% | 12.5 |
The share is highest exactly where a bakery counter hurts most. Between noon and 3pm, and again at 6pm, the kiosk takes 32% to 36% of everything ordered in the shop — a third of the afternoon rush handled without anyone leaving the pastry case.
On basket size we are going to be more careful than a marketing page usually is. The counter's raw average looks higher than the kiosk's, but that is because a quarter of counter revenue comes from 114 items the kiosk does not sell at all — custom and whole cakes, mooncake gift boxes, gift cards, open-priced specials. Compare only what both channels actually sell and the gap narrows to between 0.4% and 7.3%. The honest summary is that the kiosk takes a quarter of the orders at very nearly the same basket, not that it lifts the basket.
Two smaller facts we can stand behind: kiosk orders are discounted at 0.7% of gross, against 8.5% at the counter — the screen does not talk anyone into a break — and 67.5% of kiosk orders leave a tip. (INFI records no counter tips at all, which may mean the Square terminal's tips never reach INFI's order stream, so we will not claim the kiosk creates tipping.)
Online ordering here runs on INFI, not a third-party storefront: the “Order Now” button on the website lands on the bakery's own INFI ordering page. Between the kiosk and that page, roughly a quarter to a third of all orders now arrive through a surface INFI operates.
Earned: the part anyone can check right now
Open the bakery's Google profile and scroll. As of 17 September 2026 it holds 5.0 stars across 134 reviews, and every one of the twenty most recent reviews has a reply.
All twenty of those replies were written by INFI's AI. We can say that precisely, rather than approximately, because each reply is stored when it is generated and re-read from Google afterwards: 76 replies published, and all 76 match Google's copy character for character.
| Reviews answered | Share |
|---|---|
| Before INFI was connected | 56.5% |
| After INFI was connected | 90.0% |
| Since 1 July 2026 | 100% — 43 of 43 |
Replies typically go up about an hour after the review appears — in practice a fifteen-minute cron picking it up, then a deliberate delay so the reply does not read as a machine. Over the same period INFI pushed 72 menu photos to the profile, and the hours pipeline now runs end to end: when hours change, Google gets them and the website rebuilds itself about fifteen seconds later.
Found: what a one-shop bakery turns up for
One caveat decides the shape of this section, so it goes first. Google's record of this domain begins on 25 May 2026 — three days after the new site did. There is no “before” to compare against, and we are not going to manufacture one. What follows is growth on INFI measured against itself, over matched windows.
A second thing to hold separate: Search Console does not contain advertising clicks. The organic figures here and the campaign figures above are counting different traffic and cannot be added together.
| Organic search | First 56 days | Most recent 56 days | Change |
|---|---|---|---|
| Impressions | 8,385 | 10,903 | +30.0% |
| Clicks | 682 | 840 | +23.2% |
| Distinct queries the site appears for | 808 | 1,075 | +33.0% |
| Queries averaging page one | 637 | 842 | +32.2% |
| Queries averaging the top three | 228 | 291 | +27.6% |
| Average position | 6.27 | 5.90 | improved |
Compare the first thirty days against the most recent thirty and the same direction is steeper: impressions 3,948 → 6,141 (+55.5%), clicks 347 → 444. Over the whole 113 days the property has drawn 1,533 clicks from 19,414 impressions at a 7.90% click-through rate — and 717 of the queries in the recent window drew no impressions at all in the first one. The surface got wider faster than it got busier, which is what you would expect from a single shop in a town of 16,000.
The local cluster — the searches a bakery actually lives or dies on — moved about a position across the board:
| Query | First 56 days | Most recent 56 days |
|---|---|---|
| bakery in madison | 4.35 | 3.16 |
| bakeries madison nj | 4.75 | 3.63 |
| madison nj bakery | 5.05 | 3.89 |
| bakeries in madison nj | 5.27 | 4.13 |
| bakery madison nj | 5.56 | 4.61 |
| best bakery near me | 5.61 | 4.59 |
| coffee shops madison nj | 6.69 | 5.10 |
| matcha near me | 9.48 | 7.75 |
And the one that went the other way, because a table of only good rows is not worth reading: “bakery near me” — the single biggest non-brand term — got worse, 6.51 to 7.20. So did “bakeries near me” and the bare word “bakery.” The broad, high-volume, everyone-competes-here terms did not move for this shop. The specific ones did.
That is the more interesting half of the story. The site now appears for the actual pastry case: basque cheesecake near me at position 6.2, japanese cheesecake near me at 6.2, macarons near me at 6.1, tanghulu near me at 5.3, bubble tea madison nj at 3.3. The click counts on these are tiny — three here, five there — and we are not going to dress them up. But they are the searches of somebody who already knows what they want, and a Main Street bakery that appears for them is in a different business than one appearing for “bakery.”
Two more shapes worth naming. The category is covered — japanese bakery near me (7.2), japanese bakery nj (5.5), asian bakery near me (6.8), korean bakery near me (2.6) — and the intent is overwhelmingly local and mobile: 742 different “near me” queries produced 3,847 impressions, and 83.4% of all impressions were on a phone. That is the same trade area the advertising found from the other side, measured independently.
“Where do we rank on Google” and “does the machine that answers the question know this bakery exists” are no longer the same question. This site tells fourteen named AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the rest — that they may read everything, publishes an llms.txt index and a sitemap, and serves its menu as plain text that needs no JavaScript to read. Common Crawl, the open corpus behind many AI models, took four pages on 11 August 2026 and five on 13 July — four of those five being posts INFI wrote. And on 17 September 2026, asked “best asian bakery near Madison NJ”, an answer engine named Sugarland first — and was the only source in that result that got the right Madison. Every directory listing it retrieved alongside had drifted to Madison, Wisconsin.
What this page does not claim
Four things, deliberately, and they are the reason to trust the rest.
There is no “before INFI” trading baseline. This bakery's first INFI order and its first meaningful volume arrive on the same day, so there is nothing honest to compare against. You will not find “orders up X% after switching” on this page, because that number does not exist.
We do not connect the advertising to the till. Impressions, clicks and order starts are measurements of the ad platform. Revenue is measured at the register. Dividing one by the other produces a number that looks like return on ad spend and is really just two things that happened in the same month. INFI's own code carries that warning; so does this page.
And we are not claiming the Google profile growth. Views of this bakery's Google profile rose steeply over the same five months — but paid impressions over that period are of the same order of magnitude as the total profile views, and the advertising alone reports more direction requests per day than the profile's entire daily increase. The two cannot be separated with the data we hold, so we are not going to bank the larger number and quietly credit the free half of the work.
And we are not claiming anyone arrived through an AI assistant. There is no analytics anywhere that could separate a visitor sent by ChatGPT or Perplexity from any other, and no server log that records which AI crawlers came or how often. What is checkable is that the site invites them in writing, that its content is readable without running a line of JavaScript, that an AI-corpus crawler took it twice this summer, and that on one dated test one engine answered with it. Not more than that.
- A $5-a-day Performance Max campaign, running unbroken for 92 days, delivered 146,972 impressions and 6,635 clicks at a 4.51% click-through rate and $0.092 a click.
- The measurable outcomes are local ones: 910 direction requests, 708 menu views, 522 site visits, 122 order starts.
- 98.9% of the reach stayed in New Jersey and 94.5% of it was on a phone — the right shape for a one-shop trade area.
- The kiosk takes 21.8% to 27.3% of all orders, and a third of the afternoon rush, at a like-for-like basket within a few percent of the counter.
- Kiosk orders are discounted at 0.7% of gross versus 8.5% at the counter.
- 5.0 stars over 134 reviews, the last twenty all answered, 76 AI replies matching Google character for character, and 100% coverage since July.
- Organic search grew 30% in impressions and 33% in distinct queries between matched 56-day windows, with the local Madison cluster up about a position across the board — while the broad term “bakery near me” went the other way.
- The demand the ads surfaced but nobody is serving yet: 545 Spanish-language searches for a bakery nearby.
Advertising figures are read live from the Google Ads API for this account, 1 May – 17 September 2026, with per-day and lifetime queries cross-checked against each other. Order figures are summed order by order from the merchant's own order records in the bakery's local time zone, across four consecutive 30-day windows, and reconciled to the point-of-sale sales summary to the cent. Review figures come from the reply queue joined back to the replies Google returns, plus an independent pull of the public profile on 17 September 2026. Search figures are pulled live from this domain’s Google Search Console property, 25 May – 14 September 2026, comparing two matched 56-day halves; they are organic only and contain no advertising clicks. Crawler records come from Common Crawl’s public index; the answer-engine observation was a single dated query on 17 September 2026 and is reported as an observation, not a measurement.
Most restaurants have one of these three working and two dormant. If you want to know which of yours is which — what your ads would actually reach inside your own trade area, what share of your counter a kiosk would take at your busiest hour, and how many of your reviews are sitting unanswered right now — book a demo and we will measure all three before you commit to anything.
See what INFI can do for your restaurant
Add self-ordering, online ordering, loyalty, and marketing on top of the POS you already run.
Book a Demo
