Daniks.AI vs Manual Amazon PPC: StellaHive Case Study

Over six weeks in spring 2026, a premium candle brand on Amazon US ran two Sponsored Products tracks side by side: Daniks.AI on autopilot and an experienced in-house operator managing campaigns manually. The AI side returned 6.2 ROAS at 16.1% ACoS. The manual side returned 3.5 ROAS at 28.3% ACoS. The gap was not subtle — and unlike the previous Daniks.AI case studies published on this site, the mechanism was almost entirely bid-math efficiency, not conversion rate. Both sides converted clicks to orders at roughly 10%. The AI simply bought those clicks for 34% less.

Ownership disclosure. I am Ekaterina Rubtcova, the founder of Daniks.AI, the AI-native Amazon PPC automation that managed the auto side of this test. I built it for my own listings first — the Daniks cookware brand that reached Top-1 in Germany and is currently Top-20 in the USA — and only then made it available to other Amazon sellers. StellaHive is a customer brand running on Daniks.AI, not my brand. They sell premium soy candles and home fragrance products on Amazon US. The dashboard data below is from their seller account with permission to publish.

What this case study tested

The StellaHive test ran from mid-April through late May 2026 — six full weeks covering the spring home-refresh season and Mother’s Day, both strong demand windows for home fragrance products.

• Track A (Auto): Daniks.AI managing Sponsored Products campaigns end-to-end — campaign creation, keyword harvesting, bid adjustments, negative keyword management, and target-ACoS optimization with an AI agent.
• Track B (Manual): the brand’s in-house operator running parallel Sponsored Products campaigns on the same product catalog, using weekly search-term-report reviews, manual bid adjustments, manual negative keyword harvesting, and placement modifiers.

The operator on the manual side was not a beginner. A 28.3% ACoS and 3.5 ROAS on a home-fragrance brand is a solid result — many Amazon sellers would be happy with those numbers. This is a case where the AI won against a competent baseline, and the mechanism that drove the gap is unusually clean.

Key Takeaways

• Daniks.AI delivered 6.2 ROAS at 16.1% ACoS vs 3.5 ROAS at 28.3% ACoS for manual management — a 77% ROAS advantage over six weeks.
• The AI’s edge was bid efficiency, not conversion rate. CPC was $0.40 vs $0.61 — the AI paid 34% less per click. CVR was 10.65% vs 9.87% — only 8% higher.
• This is the cleanest bid-math story across all published Daniks.AI case studies. Fornel won on CVR. Tropeza won on both CPC and CVR. StellaHive won almost entirely on CPC.
• Volume scaled cleanly: 693 orders and $16,133 in sales on the auto side vs 95 orders and $2,065 on manual. The AI managed 4.4× the spend and produced 7.3× the orders.
• Revenue per click: $2.48 on auto vs $2.14 on manual. The AI extracted 16% more revenue from every click despite the nearly identical conversion rate, because it entered the auction at a lower cost basis.

StellaHive on Amazon US

StellaHive is a US Amazon brand in the premium home-fragrance category — hand-poured soy candles, wax melts, and reed diffusers. Mid-to-premium price point ($25–45 per set), year-round demand with clear spikes around Valentine’s Day, Mother’s Day, and Q4 gifting. The product is consumable with genuine repeat-purchase behavior: a customer who burns through a candle set in six weeks often reorders the same scent or tries a new one.

Home fragrance is a category where CPC discipline matters more than it does in durable goods. The average order value is moderate, margins are healthy but not enormous, and the buyer journey is short — someone searching “soy candle gift set” is usually ready to buy today. That makes the cost-per-click the dominant lever in PPC profitability, because the conversion rate on a well-optimized listing is already high and hard to move further through advertising alone.

The headline numbers from the Daniks.AI dashboard

The view from the Daniks.AI Amazon App, mid-April through late May 2026, Amazon US marketplace:

Metric	Daniks.AI (Auto)	Manual (Human)	Edge
Impressions	586,401	104,011	Auto +464%
Clicks	6,505	963	Auto +575%
Spend	$2,591	$585	Auto +343%
CPC	$0.40	$0.61	Auto -34% (cheaper)
Orders	693	95	Auto +629%
Sales	$16,133	$2,065	Auto +681%
ROAS	6.2	3.5	Auto +77%
ACoS	16.1%	28.3%	Auto -12.2 pp
CVR	10.65%	9.87%	Auto +8%

The first thing to notice: both sides converted at roughly 10%. On the same product catalog, in the same buyer window, with the same listings, both tracks turned about one in ten clicks into an order. The conversion rate gap is 8% — real, but small compared to the 44% CVR gap on Fornel or the 73% CVR gap on Tropeza. The story on StellaHive is elsewhere.

CPC: where the AI actually won this one

Manual PPC came in at $0.61 per click. Daniks.AI came in at $0.40. That 34% gap is the single largest driver of the ROAS difference in this case study.

Why does a 21-cent CPC gap appear on the same marketplace, same category, same six-week window?

A skilled manual operator sets bids based on their last review of the search-term report. Between reviews — typically weekly — the auction moves. New competitors enter, dayparting patterns shift, Amazon’s own algorithm reweights placements. The bids the operator set on Monday morning may be too high by Wednesday and too low by Friday. The average CPC the operator pays is the average of these drift periods: some clicks bought cheaply when the bid happened to be right, some bought expensively when it drifted above the clearing price.

An AI agent does not have a review cadence. It adjusts bids on a near-continuous loop, responding to live auction signals, time-of-day conversion patterns, and placement-level performance data. The result is a CPC that tracks the auction’s clearing price more tightly. The AI does not consistently underbid — it bids accurately, which on average means lower than a drifting manual bid.

On StellaHive, that accuracy translated into $0.40 vs $0.61: the AI bought the same marketplace’s clicks for a dollar while the manual side paid $1.53 for the same dollar’s worth of clicks. Over 6,505 clicks, the auto side saved roughly $1,366 in CPC compared to what the manual side would have paid for the same click volume at its average CPC.

For the manual playbook the human side was running, my Amazon PPC strategy guide covers the three-phase campaign structure that makes manual PPC competitive in the first place.

Conversion rate: nearly identical, and that is the point

CVR on the auto side was 10.65% vs 9.87% on manual — an 8% advantage. In the Fornel case study, the CVR gap was 44%. In Tropeza, it was 73%. On StellaHive, it barely moved.

That is actually the most interesting finding. It means the keyword mix on both sides was comparable in quality. The manual operator was sending clicks from search terms that converted at nearly the same rate as the AI’s search terms. The human’s keyword research and negative-matching were working — the manual side was not wasting clicks on irrelevant terms.

The AI did not win by finding better keywords. It won by buying the same quality of keywords for less money. That is a pure bid-optimization story, and it is the cleanest separation of the two mechanisms (keyword quality vs bid efficiency) across all published Daniks.AI case studies.

ACoS: 12.2 percentage points is the gap between good and great

The ACoS gap was 16.1% vs 28.3% — a 12.2 percentage-point difference. Both numbers are profitable on a candle brand with healthy margins. But the gap between them determines how aggressively the brand can scale its ad spend.

At 28.3% ACoS, the manual side was profitable but constrained. After referral fees (~15%), FBA fulfillment, storage, and the cost of goods, a 28% ad tax leaves a thin margin. The brand can maintain current volume but cannot scale spend without compressing profit.

At 16.1% ACoS, the auto side was deeply profitable. Every additional dollar of ad spend returned $6.22 in revenue. The brand could double its ad budget and still maintain ACoS well below the breakeven threshold. That is the difference between a PPC program that sustains a business and one that grows it.

Across $16,133 in AI-driven sales, the 12.2-percentage-point ACoS gap represents roughly $1,968 in recovered margin compared to running the same volume at the manual side’s ACoS. Over a full year at this run rate, that is $17,000+ in margin recaptured through better bid math alone.

For the underlying fee-stack reasoning on why even small ACoS differences compound into significant P&L impact, the breakdown applies directly.

How StellaHive compares to the other Daniks.AI case studies

This is the fourth public Daniks.AI case study on TheFBAGirl. Each one has surfaced a different mechanism:

Case Study	ROAS Gap	Primary Mechanism	Manual Baseline
Tropeza (Aug 2024)	3.1 vs 1.8 (+72%)	CPC + CVR	Weak (56.6% ACoS)
Basecamp Roasters (2026)	2.4 vs 1.3 (+85%)	CPC + CVR	Very weak (77.7% ACoS)
Fornel (Dec 2025)	5.8 vs 5.2 (+12%)	CVR only	Strong (19.2% ACoS)
StellaHive (Spring 2026)	6.2 vs 3.5 (+77%)	CPC only	Moderate (28.3% ACoS)

The pattern across four case studies:

• When the manual baseline is weak (Tropeza, Basecamp), the AI wins on everything — cheaper clicks, better keywords, wider gap.
• When the manual baseline is strong (Fornel), the AI wins on keyword quality alone, with a narrow gap.
• When the manual baseline is moderate (StellaHive), the AI wins on bid efficiency alone, with a wide gap.

StellaHive is the proof that the AI’s value is not monolithic. The tool adapts to where the opportunity is. When the human’s keywords are already good but their bids drift, the AI closes the bid gap. When the human’s bids are tight but their keyword inventory is narrow, the AI closes the keyword gap. The common thread across all four: the AI finds margin wherever the manual process leaves it.

Honest caveats

Same transparency standards as the previous case studies, plus notes specific to StellaHive:

1. Single brand, six weeks, spring season. The test window covers Mother’s Day — a peak demand period for home fragrance. Both tracks benefited equally from elevated buyer intent, so the relative gap is meaningful, but the absolute ROAS numbers (6.2 / 3.5) would likely compress in a lower-demand window like July or August.

2. Budget allocation was heavily weighted toward auto. The brand spent $2,591 on the AI side vs $585 on manual — a 4.4× difference. The allocation logic routed more spend to campaigns hitting their target ACoS. Per-unit metrics (CPC, CVR, ACoS, ROAS) remain fair; volume comparisons reflect budget allocation, not efficiency.

3. No A/B isolation. Auto and manual ran on different campaign sets. A clean A/B would alternate the same keywords between strategies. This is a side-by-side comparison, not a controlled experiment.

4. Halo effects. The AI side drove 5.6× more impressions than manual. Increased brand visibility likely lifted the manual side’s baseline. The manual side’s 3.5 ROAS might have been lower if the AI were not running alongside it.

5. I built the tool. I am the founder of Daniks.AI and I am publishing this because the numbers are strong. Weight that accordingly. Run your own parallel test before making a decision.

What this means if you are evaluating Amazon PPC automation

A practical read for sellers considering AI PPC tools, from someone who founded one:

• If your current CPC feels too high and your ACoS is in the 25–35% range despite decent conversion rates, you are likely in a StellaHive-shaped situation. The AI’s first win will be CPC reduction through better bid timing. Expect the ROAS improvement to show up within two weeks.
• If your CVR is already above 8–10% on a well-optimized listing, there is less room for the AI to win on keyword quality. The win will come from bid efficiency — buying the same quality of traffic for less money.
• If you are in a consumable or repeat-purchase category (candles, coffee, supplements, pet supplies), high CVR is the baseline expectation. The PPC lever that moves is CPC, not CVR. This is where Daniks.AI’s continuous bid loop earns its keep most clearly.
• If you are still running bids on a weekly review cadence, the drift between reviews is where your margin goes. Seven days of unoptimized bids on a $2,500/month ad budget can easily cost $300–500 in excess CPC.

For the full feature breakdown, the Daniks.AI review covers the pricing tiers, strengths, and where the tool falls short.

Frequently asked questions

What is a good ACoS for candles and home fragrance on Amazon?

Most operators in the home-fragrance category target 15–25% ACoS. Margins on soy candles are healthy enough to sustain ACoS in the mid-20s, but brands that want to scale aggressively need to push below 20%. StellaHive’s 16.1% on the AI side is in the aggressive-growth band; the manual side’s 28.3% is in the maintenance band.

How did Daniks.AI achieve 34% cheaper CPC than the manual operator?

The AI adjusts bids on a near-continuous loop based on live auction signals, time-of-day conversion patterns, and placement-level data. The manual operator reviewed bids weekly. Between reviews, bids drifted above or below the optimal clearing price. The average CPC reflects the average of those drift periods — and drifting up costs more than drifting down saves.

Why was the CVR gap so small compared to other case studies?

The manual operator on StellaHive was doing good keyword work. Their search-term targeting produced clicks that converted at 9.87% — close to the AI’s 10.65%. The gap was in bid execution, not keyword selection. This is actually the strongest possible endorsement of the manual operator’s keyword research skills.

Is 6.2 ROAS realistic for an Amazon candle brand?

On a mid-premium candle brand with strong listings and clean keyword targeting, 5–7 ROAS is achievable during peak-demand windows like Mother’s Day season. Year-round average would likely settle in the 4–5 ROAS range. The 6.2 ROAS benefits from the spring gifting window — but the relative gap vs manual holds.

How long should I run a parallel test to see results?

In my experience, the CPC advantage shows up within the first week — bid optimization converges fast when the AI has live auction data. The full ROAS picture stabilizes by week three. StellaHive’s six-week window is longer than strictly necessary for the pattern to emerge, which makes the numbers more reliable.

Does Daniks.AI work on Sponsored Brands and Sponsored Display too?

Yes. Daniks.AI manages Sponsored Products, Sponsored Brands, and Sponsored Display campaigns. This case study covers Sponsored Products only, because that is where the parallel test was structured. The bid-optimization mechanism applies across all ad types.

What to do this week

If you run an Amazon brand in a consumable or home-goods category, pull your search-term report and your average CPC from the last 30 days. If your CPC is above $0.50 and your CVR is above 8%, you are likely in a StellaHive-shaped situation — your keyword targeting is working but your bids are drifting between reviews. Run a parallel two-week test: keep your existing campaigns running, set up a clean campaign set on Daniks.AI or any AI PPC tool you can trial, and compare daily CPC and ACoS. The bid-efficiency pattern will be visible by day seven.

For the operator-level walkthrough of these metrics and the PPC framework I run alongside any automation tool, subscribe to @AmazonFBAGirl on YouTube. I break down real dashboards, real numbers, and real mistakes — no theory, no hype. Drop a comment with your category and current ACoS, I read every one.