Andrew Lolk

The alternative is to run non-brand Search text ads off the same product feed that powers Shopping, so product names, categories, brands, and models fill the ads without hand-built keyword lists.

• Because the feed keeps the ads relevant, Search behaves like a second Shopping channel at similar ROAS, with its volume arriving on top of what Shopping already delivers rather than replacing it.
• Their diagnostic is a budget check: if Search is under 30% of total Google Ads spend, money is likely being left on the table.

On one large ecommerce account, non-brand Search text ads built from the feed produced, year over year, 167% more spend, 180% more conversions, and 830% ROAS — more than double the spend with ROAS still rising, roughly level with that account's Shopping campaigns.

Lolk adds that the feed-driven approach matters more as AI takes over much of the legwork, rather than replacing the feed requirement.

Read original

The rule is too generic: it fits a shopper searching for a specific brand and model, but not private-label goods, unknown brands, or category-led searches where the shopper is not brand-driven.

• Their working rules are a short test: are the search terms category-focused, is the brand unknown, and are other benefits more likely to lift click-through rate?
• A yes to any of these is a reason to move the brand toward the back of the title, not to delete it.
• A Black Friday collaboration with DataFeedWatch supplies the format for the argument, and the post points to that tool's AI mapping and A/B testing features as one way to run title experiments at scale.

Lolk illustrates the point with a shopper searching for a bathing suit with a clear back: whether it is a Nike bathing suit is unlikely to be what they care about, so the brand need not lead. They also report having tested removing the brand from titles entirely, and call that a bad outcome.

Read original

The exception is a category that loses share of spend month after month inside a shared portfolio and can tolerate a lower ROAS target.

• A single shared target lets Smart Bidding shift budget toward whichever product line returns more, slowly starving a line with structurally lower but acceptable returns — typically one with higher repeat purchase rates.
• Giving that category its own tROAS target while leaving the rest consolidated is the fix; they frame it as the exception that proves the rule.

In an account selling dog food and dog accessories, both lines shared one target; accessories returned more, so budget drifted away from dog food.

• On 16 September the portfolio was split by category.
• In Denmark, dog food then ran at 700% and accessories at 850%, category spend rose about 88%, portfolio ROAS landed between 691% and 838%, and weekly net profit ran between 65,000 and 121,000 kr.
• Sweden used lower targets, 600% and 700%, with weekly net profit moving from -4,574 kr in week 35 to +28,243 kr in week 38.

Lolk calls the result preliminary, based on a few weeks of data, and wants another month before settling the split, while keeping consolidation as the default for other accounts.

Read original

The stated logic was sound — higher margin, control of customer relationships and first-party data, and full control of pricing — and Lolk has seen it succeed for some clients and fail for many.

• In their account, the failure comes from treating an established retail network as an expense when it functions as a moat: salespeople in local stores do real selling on a brand's behalf, and because several athletic shoe brands are good enough to recommend, that recommendation power is contestable.
• Neglecting retail forces heavy marketing spend, and the margin gained by going direct is consumed by the cost of discovery.
• Their conclusion is that companies should focus above all on product quality.

Lolk points to the Mbappé move to On, with an equity stake, landing just as Mbappé's stardom peaked, as the culmination of the shift away from Nike rather than a single deal in isolation.

Read original

A newcomer can follow the argument through one measurable account example.

Andrew Lolk describes building non-brand Search text ads directly from a large ecommerce account’s product feed, using product names, categories, brands and models without manual keyword work.

• Compared with the previous year, non-brand text ad spend rose 167%, conversions rose 180%, and ROAS reached 830%—more than double the spend, with ROAS still climbing and roughly level with that account’s Shopping campaigns.
• Lolk says most advertisers treat Search as manual keyword work: they build a few campaigns, leave them untouched, and Smart Bidding quietly cuts spend because ROAS looks bad.
• Search then falls below 10% of the budget.
• With the same clean feed running Search, it behaves like a second Shopping channel at similar ROAS, adding volume on top of Shopping.
• The practical check: if Search is under 30% of total Google Ads spend, money is being left on the table.

For brands planning a campaign

This is the anchor case for a webinar on Shopping and feed-driven Search.

• The 30% threshold gives agency clients a quick diagnostic they can run on their own accounts.
• Because the approach depends on the feed, the pitch naturally pairs feed cleanup with Search activation rather than selling Search in isolation.

It is a concrete case of deliberately breaking a default rule, with numbers and an admission that the data is still early.

Over six months, one product category on a large ecommerce account lost a growing share of spend without anyone noticing.

• At a quarterly business review, the portfolio was split into two tROAS targets by category on 16 September, and the trend turned a week later.
• In the worked example, dog food can run at a lower ROAS than dog accessories because repeat purchase rates are higher; both had shared one target, so Smart Bidding shifted budget toward the better-ROAS accessories.
• In Denmark, dog food now runs at 700% and accessories at 850%, category spend is up about 88%, portfolio ROAS sits between 691% and 838%, and net profit runs between 65,000 and 121,000 kr per week.
• Sweden received the same setup at lower targets, 600% and 700%, with weekly net profit moving from -4,574 kr in week 35 to +28,243 kr in week 38.
• Andrew Lolk calls themselves the biggest advocate for consolidating categories into one portfolio and says it remains the right default for most accounts.
• This is the exception; they want another month of data to confirm it.

For brands planning a campaign

Gives the webinar a segment on when to break the consolidation rule. The diagnostic is simple enough for an agency audience: a category losing spend share month after month inside a shared portfolio, and able to tolerate a lower ROAS target, is a candidate for its own target.

It offers a short, testable set of rules and warns against acting on them too aggressively.

Ahead of Black Friday, Andrew Lolk runs a myth-busting collaboration with DataFeedWatch for ecommerce professionals preparing feeds for peak season.

• Myth one: the brand always belongs first in the title.
• Lolk acknowledges that the template is everywhere, including in their own and DataFeedWatch’s, but argues it is too generic.
• For a search for a specific brand and model, showing that brand makes sense.
• For private label, unknown brands or category-led searches such as a bathing suit with a clear back, the brand may not be what the shopper cares about.
• The rules: Are the search terms category-focused?
• Is the brand unknown?
• Are other benefits more likely to lift click-through rate?
• A yes to any of these is a reason to move the brand toward the back of the title—not delete it.
• Lolk says they previously tested removing the brand and found it a bad outcome.
• The post also points to DataFeedWatch’s AI mapping and A/B testing features for running title experiments at scale.

For brands planning a campaign

The title structure makes this a practical webinar segment with a takeaway checklist. Lolk also spells out both the tool and the method in this sponsored collaboration, making it a useful example of a content-led partner format without implying an endorsement beyond what the post says.

A newcomer should read it after the tactical posts.

Andrew Lolk reflects on Nike’s aggressive move into direct-to-consumer sales through its own site and stores, intended to stop stock declining.

• Four years on, they say it has not worked.
• The logic made sense: higher margin, control of customer relationships and first-party data, and full control of pricing.
• Their explanation for the failure is that companies treat an established retail network as a cost when it is actually a moat: a salesperson in a local sporting goods store sells on a brand’s behalf, and because several athletic shoe brands are good enough to recommend, that recommendation power matters.
• Neglecting retail requires heavy marketing spend, and the margin gained by going direct is consumed by the cost of discovery.
• Lolk argues that every company should focus above all on product quality, and notes that the Mbappé move to On, with an equity stake, landed as Nike’s star was peaking.

For brands planning a campaign

Useful as a webinar’s framing section: performance marketing is one lever, not a replacement for distribution and product quality. For an agency positioning Shopping and feed-driven Search work, this supports selling channel work as part of a broader mix rather than a standalone fix.

In one large account, the team built non-brand Search text ads from the product feed, automatically filling in product names, categories, brands, and models.

• Reported year-over-year results were non-brand text ad spend up 167%, conversions up 180%, and ROAS at 830%.
• They say Search then behaves like a second Shopping channel, adding similar ROAS and volume on top of what Shopping already brings.
• Treating Search as manual keyword work, they warn, can lead Smart Bidding to cut spend when ROAS looks poor.

Build non-brand Search text ads directly from the feed; check what share of Google Ads budget goes to Search and treat under 30% as a potential missed opportunity.

For brands planning a campaign

For a B2B ecommerce agency hiring paid search and feed expertise, this offers a practical webinar angle: how a clean product feed can generate and scale non-brand Search text ads, and how to review budget share. The 30% budget check gives the session a simple diagnostic.

Examples: Source 1

They argue that rule is too generic for private-label or unknown brands and for category-focused searches where shoppers may not care about brand.

• They suggest moving the brand later in the title when search terms are category-focused, the brand is unknown, or other benefits are more likely to lift click-through rate, without removing the brand entirely.
• They also mention testing original titles against AI-generated ones and scaling title structures with rules.

Check whether search terms are category-focused, whether the brand is unknown, and whether other benefits are more likely to increase CTR; then test title order instead of applying one default.

Lolk notes that removing the brand from the title entirely was tested and did not work well.

For brands planning a campaign

Useful for a feed optimization webinar: a brand can evaluate whether Lolk's title tests fit its catalog and how to set up controlled title experiments across product groups. The practical rules give the agency a concrete framework for a feed audit.

Examples: Source 1

In mid-September, the team split the portfolio into two tROAS targets by category.

• In a dog food and accessories example, dog food could run at a lower ROAS because repeat purchase rates were higher, while accessories were more one-off.
• In Denmark, dog food ran at 700% and accessories at 850%; spend on the category rose about 88%, portfolio ROAS was between 691% and 838%, and weekly net profit was between 65k and 121k kr.
• Sweden used lower targets and moved from a weekly net loss to profit.
• Lolk still calls consolidation the default and says only a few weeks of data were available.

Monitor category share of spend over time; if one category keeps losing share and can run at a lower ROAS, give it a separate tROAS target while keeping consolidation as the default.

Lolk presents the split as an exception that proves the rule and says more time is needed before treating it as settled.

For brands planning a campaign

This offers the agency a case-study topic for a Shopping and feed webinar: how shared tROAS can quietly starve a repeat-purchase category, how to diagnose it, and when a category-level split is worth testing. Ask for longer-term results before treating the early findings as settled.

Examples: Source 1

The logic was higher margin, first-party customer data, and pricing control, but they argue companies often treat the retail network as a cost when it is a moat.

• A salesperson in a local sporting goods store can steer customers to a brand, and those recommendations carry weight.
• Going direct gains margin but can lose it to the cost of discovery.
• They conclude that product quality should be the central focus and mention Mbappe taking an equity stake in On.

For brands planning a campaign

For an agency assessing Lolk as a paid search and feed expert, this shows a broader commercial-strategy perspective beyond campaign mechanics. It could support a wider webinar segment on channel mix, retail influence, and why paid search and feeds are not the whole demand picture.

Examples: Source 1

They name a default practice, explain when it should change, and show the account-level result. They also flag early evidence, as with the portfolio split, and retain the default recommendation rather than generalizing from one case.

For brands planning a campaign

For a webinar or agency assessment, Lolk's material is specific enough to turn into worked examples, with claims framed around conditions and tests rather than universal promises.

Examples: Source 1 · Source 2 · Source 3

The post describes using the company’s AI mapping and A/B testing features for title experiments. It does not specify payment or other terms, so ask about compensation and the collaboration’s terms.

Original disclosure

“I’m kicking off a collaboration with DataFeedWatch”

View example

View example

Fry’s piece on Conversion Lift arriving for Search and Performance Max makes a similar argument to Lolk’s feed-driven Search post: did reported performance cause the sales, or merely take credit for them? Fry’s post on Google’s Target ROAS calculator also builds a target from profit margin, as Lolk does from a category’s acceptable ROAS.

Fry focuses on measurement and profitability, including incrementality testing and tying ROAS back to the profit and loss account.

• Lolk focuses on structure and inputs: what goes into the feed, how ads are built from it, and when to split a portfolio.
• Readers checking whether a channel works will get more from Fry; those deciding how to configure a feed or bidding structure will get more from Lolk.

Read original · Original 2

Wyatt’s post on running Performance Max and Shopping in parallel covers campaign segmentation and custom-label logic for feed items, much like Lolk’s argument that the product feed should drive more than one campaign type. Wyatt’s follow-up on AI campaign analysis also lists the kind of performance-dip diagnosis Lolk discusses when Smart Bidding shifts spend away from a good category.

Wyatt gives step-by-step build instructions tied to a specific feed platform’s custom labels and tools, with the platform’s features in the foreground.

• Lolk focuses on decisions and account outcomes, offering conditional rules rather than numbered steps.
• They are also more likely to look beyond campaign mechanics to distribution and product strategy.

Read original · Original 2

Daniv’s post on plain, practical SEO deliverables is the organic-search counterpart to Lolk’s warning that teams treat paid search as a few campaigns nobody maintains. Both also write about what a client or agency should put in place month by month.

Daniv focuses on SEO program structure and the client-agency relationship, including who holds decision authority and why work stalls in approvals. Lolk stays within paid search and feed mechanics, using harder numbers on individual accounts and giving less attention to how the working relationship shapes delivery.

Read original · Original 2

Fry's two posts take a good-looking number and ask what it actually proves.

• One covers Google opening Conversion Lift testing to Search and Performance Max without a rep: hold back a group of users who never see the ads, then measure the gap.
• The point is that a brand search campaign at 10.1x ROAS and $242K of revenue may have collected sales that would have happened anyway.
• The other walks through a beta Target ROAS calculator that turns a profit margin into a breakeven figure — 667% at a 15% margin — and warns that the margin entered is usually too generous once shipping, returns and fees are counted.
• For an agency selling measurement as part of the engagement, or a webinar segment on whether brand-search ROAS means anything, Fry covers ground Lolk's posts do not.

This is measurement and bidding arithmetic, not feed structure, product titles or how to build Shopping and Search campaigns.

• A webinar built around Fry would need someone else to cover feeds and campaign builds, and the paid Search expansion case Lolk makes would be left out.
• The incrementality test also has floors — at least 1,000 observed conversions, at least $5,000 in campaign budget, and availability limited to some accounts — so most small and mid-size client accounts cannot use it.
• The method suits agencies with larger clients better than a universal offer.

Examples: Source 1 · Source 2

The rationale is that Performance Max focuses on warm traffic and leaves low-impression items with almost no exposure, so a parallel Shopping campaign picks up revenue rather than competing for it.

• A second post describes a campaign deep-dive summary that flags products that dropped off, budget pacing, out-of-stock items and channel-mix shifts.
• For an agency already running Performance Max and trying to recover Shopping volume, this is closer to the working problem than Lolk's Search budget argument.

The method relies on one feed and reporting platform, and parts of the post read as a product walkthrough as much as a technique.

• A webinar built on it would therefore include a tool demonstration, with the terms to be settled separately.
• Wyatt also says the typical results appear in the image rather than the post text, so ask for the revenue-gain figures before presenting the method to clients.
• This also trades away the centre of Lolk's pitch: the claim that a clean product feed can be reused to run non-brand Search text ads as a second Shopping channel.
• The webinar premise would shift from expanding Search to expanding Shopping.

Examples: Source 1 · Source 2

Daniv's first post boils an SEO programme down to four monthly deliverables: fix product and category pages built for commercial-intent searches; build internal links with descriptive anchor text; create supporting content that answers buying questions; and acquire links through monthly outreach.

• It argues against pitches led by long technical audits, dashboards and brand-mention reports with no next action, and includes a worked example of a "standing desk" cluster feeding a commercial page.
• The second post explains why campaigns stall: pending account and website access, feedback arriving one message at a time, no one holding final decision authority, and every change pushed to the safest option, while the agency gets blamed for a slower timeline.
• For an agency that loses weeks between recommendation and launch, that second argument is the rarer one.

Daniv's posts do not touch paid Shopping or feed-driven Search.

• A webinar built around them would not meet the paid search and feed need; it would be an organic search and client-process session, a different skill from the one the agency named.
• As a hire, the posts point to SEO and account-management strengths rather than paid media execution.
• The four-deliverable list is a framework, not the account-level figures Lolk's posts include, so a client asking what the work returns would get a method, not a result.

Examples: Source 1 · Source 2