Night Operators

Where they started

Night Operators is a Vancouver brand run by its founder, Bon Zhan. It sells digital night-vision and thermal goggles (MAX 2.0, PRO 1.0 and the new FUSION 3.0) to airsoft, Nerf and paintball players. The brand grew on organic video; one TikTok did about 40 million views. December 2023 closed at $258,639 in Shopify sales on $1,489 of Google spend. Between 1 November 2023 and 18 July 2024 the store did $1,001,527 in Shopify sales, almost all of it before paid search was switched on.

Then the spike faded. January 2024 did $219,076, February $118,142, March $102,710, April $100,603, May $68,081 and June $64,943. The whole paid setup on Google was one search ad in one ad group, on a $40 daily budget. In Bon's words on the discovery call: "I'm literally not even managing it to be honest with you. It just runs in the background because it's search ads." And: "I can't lie to you. I got search ads and that's it. There was never a single creative in there."

He knew the channel had room. "I know for a fact that Google ads has room for me to keep pushing. I just don't know how." His break-even cost per order was about $97, his average order about $211, and Meta was paused. He planned to restart Meta himself around the new FUSION product. On the audit call three days later the goal was set out plainly: around $70K a month now, $1 million a month the target.

What the audit found

Bon sent the Google Ads invitation on 26 June 2024. Shopify, Merchant Center and the manager-account link followed on 2 to 5 July. We wrote the audit on 27 June, two days before he signed, and walked him through it on a recorded 21-minute call. The document has four sections: Website, Google Ads, Meta Ads and A/B Testing. On the call the Google Ads part took the least time: "there's not so much for me to say because there's literally nothing there." Bon agreed: "It's non-existent. Exactly."

The mechanism behind his scaling problem sat in that one ad group. Brand searches and general searches shared one ad and one budget. Brand searches convert on their own, so the account looked healthy at $40 a day. Every extra dollar spilled into cold searches the ad was never written for. Bon had felt it: "we got a good CPP, but as soon as you increase the budget, then obviously then, um, it throws the money into fire."

The store had its own leak. Add-to-cart was fine at 3.2%. Only 0.78% of sessions reached checkout and 0.32% bought; the June dashboard further down carries all three numbers. The checkout hid the discount the product page had promised and carried no reviews. Below are the Website and Google Ads sections of the audit as written, then the page itself. After that, the control versions from our September test: the homepage and the FUSION product page as they still stood when the CRO work began.

What we did, in order

1. 24 Jun 2024Discovery call, recorded (41 minutes). Bon describes one search ad on $40 a day that 'just runs in the background'. The viral spike has faded. He knows the channel has room: 'I just don't know how.'
2. 27 to 29 Jun 2024Written audit on 27 June and a recorded call to present it. A flat $2,500 a month Google Ads and CRO retainer is quoted on the call. Agreement signed and the first month paid on 29 June; onboarding form and shared Drive folder sent the same morning.
3. 5 to 8 Jul 2024New structure built from keyword research. A brand campaign was kept apart from a non-brand search campaign with eight themed ad groups. They were Fusion, Helmet, MAX 2.0, PRO, Airsoft Gear, Paintball and Airsoft, Night Vision and Nerf Wars. Seven were mapped to a landing page, with three ads each. Brand campaign live Friday 5 July, search campaigns Monday 8 July. Bon on 10 July: 'looks amazing just read everything'.
4. 9 to 17 Jul 2024Shopping feed built through DataFeedWatch with rewritten product descriptions and compare-at prices fixed. The feed carried 29 products when it went live on 16 July, and Merchant Center's July report counted 31 approved. Performance Max built and switched on once Bon's new video creative arrived on 17 July.
5. 22 Jul to 15 Aug 2024Intelligems installed and a FUSION price test run against two higher tiers. Result on 15 August: the cheapest price won on revenue per visitor, $2.36 against $1.38 and $1.41. Test paused, prices held.
6. 6 to 8 Aug 2024Strategy call: reporting walkthrough, GTIN and MPN codes added to lift feed quality, feed-title tests planned, CRO plan agreed (navigation, hero, product page, how-it-works video). Bon shipped the navigation and page changes himself by 8 August and moved to Shopify Plus. He also bought a second retainer for his other brand.
7. 16 Aug to 19 Sep 2024CRO tests through Intelligems: homepage hero and product-page restructure. Checkout gained trust badges, payment icons and two verified reviews on 16 August. The compare-at price was surfaced in checkout on 2 September. 'Military-grade' hero copy was replaced within hours of Bon's objection on 19 September. Triple Whale set up end to end at the client's request, 20 to 29 August.
8. 16 to 25 Sep 2024Course correction. Two-day deep dive after Bon's break-even warning. Broad match removed completely. The brand campaign moved to phrase match with a new bidding strategy, and a $20 a day worldwide brand duplicate was added. The target-ROAS prospecting campaign was paused, a discount asset group added for Q4, and an asset-level reporting script installed. Two new campaigns live on 25 September.
9. 24 Sep to 28 Oct 2024A FUSION full-kit offer page was briefed to a design studio; first draft on 7 October ('overall the structure looks a lot better than our current product page'). Search Console set up on 24 October. On 25 October an outside audit shared by Bon was reviewed and the duplicate purchase conversion action was fixed. On 28 October Bon paused the Google work and moved it to a cheaper agency.

What happened

Structure was the missing piece. Demand was already there. Ten days after launch the account read 7.85x, mostly retargeting. On 24 July the seven-day read was $2,100 of Google spend against $23,700 of Google-attributed conversion value. July closed at $126,705, up 95% on June, on $5,680 of Google spend. On 26 August we wrote in the client channel: "When we started working together 2 months ago we were at $14k/week in rev and last week we hit the $53k/week." The tracker rows for the four weeks before launch are $14,740, $14,924, $18,240 and $11,317; week 34 is $53,869. August closed at $221,828 on $19,223 of spend. September did $246,963.

One thing needs saying plainly. In the same window Bon ran his own Meta ads with a coach. He also ran 15%-off email promos and a 60%-off sale, sold on Amazon with buy buttons on the product page, and launched FUSION 3.0. The store total is not a Google-only result.

On the site, two A/B tests ran through Intelligems. The homepage hero test put a new headline and three bullets above a stronger call to action. It lifted desktop conversion from 1.21% to 2.07% and revenue per visitor from $5.50 to $8.58. It was cut to desktop only after the mobile version performed worse; the all-device read was +4.33%. The product-page test moved the benefits above the buttons and made the title smaller. In the 9 September report it showed conversion up 12.4% and revenue per visitor from $1.88 to $2.03; the final read on 17 September was +13.28%. A price test on FUSION found the cheapest tier won on revenue per visitor: $2.36 against $1.38 and $1.41.

What did not move: store-wide conversion. It sat between 0.29% and 0.33% from June to September as sessions doubled from 86,911 to 181,606 and got colder. It reached 0.40% in October. Average order value did double, from $220.82 to $470.58. The campaigns pointed at the $489 to $649 FUSION line rather than the $180 to $290 MAX and PRO line. The founder also moved FUSION's prices up in mid-July and back down by August, so the mix is what carried it.

What the ROAS numbers count

In October an outside audit found two purchase conversion actions, both set as primary in Google Ads. Google's own ROAS counted some sales twice. Triple Whale, which we installed in August and managed to, reads outside the platform and was not affected by that duplicate. It was not clean either: on 3 September the founder told us Triple Whale was picking up "only 50% of the net profit that it should be". Neither number was perfect; we optimised to the third-party one.

That applies to the attribution share on this page. The 35% and 44% figures for July and August are Google's own conversion value over Shopify revenue, so they are overstated. We cannot date when the second conversion action was created, because it was only found in October, so we cannot prove June's 7.9% is clean. If it is, the jump is smaller than it looks; if it is not, both ends are inflated. Treat the size of the change as uncertain and the direction as solid.

The ramp was not a straight line. In-platform ROAS went 7.94x, 7.88x, then 3.60x as weekly spend rose from $1.3K to $3.5K in three weeks. By mid-September Triple Whale read 2.85x and Bon wrote that Google was "borderline losing money now." Broad-match keywords were pulling in people researching $5K to $10K military kits. We removed broad match completely, moved the brand campaign to phrase match, paused the target-ROAS prospecting campaign and added a discount asset group for Q4. Week 41 (7 to 13 October) was the highest in-platform return of the ten weeks we tracked side by side. It ran $3,624 of spend, $64,070 of store revenue, 6.58x in-platform, 5.12x on Triple Whale and store conversion 0.49%.

Did it clear his break-even?

The founder gave his margins twice: "roughly 50% margins", then "sometimes its 40% margins depending on country of shipment", with "ideally 2.5x roas plus to be safe" and "anything lower is grey area could be unprofitable". At 50% margins break-even is about 2.0x; at 40% it is 2.5x.

Triple Whale's Triple Attribution reading ran 2.51x to 5.23x across weeks 35 to 44. All ten weeks cleared 2.0x, and weeks 35 to 38 (2.51x, 2.89x, 2.75x, 3.14x) sat at or just above his 2.5x line. But the model he called "the most accurate" was Triple Whale's linear paid model, which ran about 20% lower. Our own 9 November message put week 43 at 4.21x linear against 5.23x, and week 44 at 2.81x against 3.47x. On that model weeks 35 to 37 do not clear his 2.5x line.

Three limits on that. Triple Whale was only set up between 20 and 29 August, so July and early August have no third-party reading. These are the founder's own margin statements, not a profit and loss we measured. And he did not agree the channel was profitable. On 15 September he wrote that 2.85x "would mean break even after costs with your agency". On 1 October, looking at the linear model: "so looks like we arent breaking even ... as of september overall". After the handover he described his Google-era margins as "9% average days".

What the business became

At the start, paid search was one ad and 7.9% of revenue. At the end it was a structured account. A brand campaign sat apart from eight themed non-brand ad groups, seven of them pointed at a matched landing page. The shopping feed grew from 29 products and 31 approvals in July to 115 approved by September, with 20 not approved and 10 limited. Performance Max ran with asset groups by product, and YouTube ads were live. Google's own attribution put 35% to 44% of store revenue on the channel in July and August. The store got back to its January level, and September beat it, this time on a paid engine.

Measurement changed too. Before, Bon's method was this: "I literally just hop on a Google ads account and look, how much did we spend today? And just subtract it from the total profits. That's it." From July he had a weekly tracker with spend, revenue, MER, new customers, cost per new customer, 90-day value and conversion rate. Then came Triple Whale and Search Console. On the 6 August call: "Very well done on the organization there. This is something I've never done before."

He moved fast on his side, which is part of why this worked. He rebuilt the navigation, FAQ and About pages within 48 hours of the 6 August call. He moved to Shopify Plus on the projected volume: "considering the numbers I am projected to do this month with ads, I save $2000 in fees." He published dormant reviews, shot new creative, hired a 3D animator, and five and a half weeks in bought a second retainer for his other brand.

The economics stayed thin by his own account: "roughly 50% margins", and he wanted "2.5x roas plus to be safe." The retainer was a flat $2,500 a month for Google Ads and CRO. In late October, four months in, he moved the Google account to another agency, which he said "has a cheaper going rate". The last Night Operators fee was charged on 30 September and the work stopped on 28 October 2024. He wrote the next day: "if their results do not exceed my expectations beyond a reasonable doubt, I will happily switch back."

Weekly Google spend on the account was under $300 by mid-December. We do not know whether that was a strategy decision or an account problem, and we did not run the account. The brand itself is bigger now: its About page, checked 7 September 2026, lists 17,000+ orders shipped since 2023, 5,000+ verified reviews and 90+ countries.

What you can take from this

Verify this case study

• Every store figure here is Shopify total sales, copied into our weekly tracker. It reproduces $64,943.18 and $221,828.33 to the cent, and both dashboards are shown above.
• Every founder quote carries its date and comes from four recorded calls (24 Jun, 27 Jun, 2 Jul, 6 Aug 2024, on Read.ai) or the client Slack channel. Spoken quotes are machine transcriptions. We named the brand and the founder on our own decision. Nothing here was written for publication as a testimonial, and we have not asked him for one.
• The audit bullets are verbatim from the Google Doc created 27 June 2024, 18:31 UTC. Google's Merchant Center emails for July and August are shown above: 432 and 3.27K ad clicks. The September email reports 3.59K clicks and 410.62K impressions, and is available on request.
• ROAS is labelled by source throughout. Google's in-platform figures from July to October 2024 are overstated by a duplicated purchase conversion action, fixed 25 October 2024. Triple Whale sits next to them; HoloGrowth is a Triple Whale partner.
• See the brand's own search and shopping ads in the Google Ads Transparency Center. The Meta Ad Library covers the founder's own Meta ads, not the Google work described here.
• Ask for a reference on your Strategy Call. Numbers on this page last updated 2026-09-08.