Md Tahidul Islam

The practical move is to pull conversion data, identify people who converted but show weak signals—lowest average order value, highest return rates, one-time buyers who never returned—and remove those segments from prospecting and retargeting.

• They recommend reviewing this monthly as an active optimization lever, not setting it up once and forgetting it.
• The reasoning: fewer low-quality impressions mean more budget flows to higher-propensity users, and the algorithm learns from cleaner data.

In one ecommerce account, exclusion lists built from the bottom 30% of converters cut wasted impressions by 19% and lowered cost per acquisition without changing creative or budget. The exclusions targeted weak converters rather than non-converters, the opposite of the usual exclusion practice.

There is a tension across the material: the Facebook algorithm explainer presents the algorithm as finding likely buyers and re-serving ads to previous visitors, while the pruning note argues teams still need to actively exclude weak converters. Read together, they suggest targeting is only half the job.

Read original

Visitors arriving from AI citations such as ChatGPT, Perplexity, and AI Overviews convert at much higher rates because the AI has already answered their questions and compared options before sending the click.

• The failure is routing those visitors into the same generic homepage or blog post built for cold, uninformed traffic.
• The proposed fix is a separate path for AI-referred visitors: proof and specifics above the fold, pricing visible quickly, and one obvious next step instead of several navigation options.
• The metric to watch becomes revenue per visit rather than sessions.

The post reports dashboards showing top-of-funnel organic traffic down 38% year over year while visitors from AI citations converted at 3.2 times the rate of regular organic traffic.

Read original

The Google post describes tracking every click and conversion and using remarketing to bring back people who visited but did not convert.

• The Facebook post describes the algorithm as analyzing user behavior to find likely buyers and optimize delivery, while the advertiser's job is to craft memorable ads and re-serve them consistently.
• The lesson: platform targeting does not remove the need for message discipline and follow-up.

The Google post uses a fictional vegan burger owner named Bob, who uses remarketing to bring back visitors who did not convert. The Facebook post uses a fictional marketer named Jack, who targets young professionals who like coffee and cats and uses retargeting as a friendly reminder.

Read original · Original 2

Start here: it is the most specific piece in the set, naming a lever, a measured result, and a review cadence.

Md Tahidul Islam describes an ecommerce account where wasted impressions fell 19% and cost per acquisition dropped without any change to creative or budget.

• The change was exclusion lists built from the bottom 30% of converters: buyers with the lowest average order value, the highest return rates, and one-time purchasers who never came back.
• The argument is that most teams treat negative audiences as a one-time setup task, adding existing customers and a few job titles, then forgetting them.
• Accounts that improve treat exclusion as an active optimization lever and review it every month.

For brands planning a campaign

For a B2B SaaS paid media campaign, this calls for a standing exclusion review, not a one-off setup: build lists from low-quality converted accounts, such as small contracts, high-churn customers, or accounts with no expansion path; exclude those segments from prospecting and retargeting; and re-check monthly. It gives the campaign a concrete wasted-spend tactic, with wasted impressions and CPA as the stated measurements, instead of a general instruction to tighten targeting.

Start here to learn how to convert that traffic rather than chase volume.

The post reports dashboards showing top-of-funnel organic traffic down 38% year over year, while visitors arriving from AI citations — ChatGPT, Perplexity, AI Overviews — convert at 3.2 times the rate of regular organic traffic.

• Md Tahidul Islam argues that an AI engine has already answered the visitor's questions and compared their options before the click, so they arrive pre-qualified in a way a random searcher does not.
• The post describes a common failure: routing those visitors to a generic homepage or blog post built for cold, uninformed traffic.
• Its proposed fix is a separate path for AI-referred visitors, with proof and specifics above the fold, pricing visible quickly, and one obvious next step instead of several navigation options.

For brands planning a campaign

Directly relevant to the AI-referred traffic half of a brief.

• It supports testing a landing-page variant for AI-referred sessions and changing the reported metric from sessions to revenue per visit.
• The 38% and 3.2x figures are this person's reported observations from dashboards they looked at, not an industry benchmark to quote to a client.

A newcomer can start here for a plain-language explanation of keyword targeting, conversion tracking, and remarketing.

The post walks through Google search ads using an invented scenario: a small business owner running a vegan burger restaurant, choosing keywords, customizing the ad message, tracking every click and conversion, and using remarketing to bring back people who did not convert on their first visit. It is written for business owners with no advertising background and leans on jokes and pop-culture comparisons rather than account data or testing methods.

For brands planning a campaign

Read this for tone and explanatory range rather than tactic.

• If a campaign needs someone who can make paid search legible to non-specialists — in a client explainer, webinar segment, or onboarding series — the style here shows that.
• There are no benchmarks or test design, so it should not be the basis for a performance claim.

Start here for a non-technical explanation of targeting, retargeting, and consistency.

A narrative explainer about an invented marketer named Jack.

• It covers how Facebook's algorithm analyses user behaviour to find likely buyers, optimizes delivery and targeting, and supports retargeting people who have already interacted with the brand.
• The audience example is deliberately simple — young professionals who like coffee and cats — and the lesson is consistency rather than any single test result.

For brands planning a campaign

Like the Google search ads post, this offers explanatory range rather than proof.

• It shows the creator can carry a concept through a story with a lesson attached, making it suitable for awareness-stage content or a client-facing primer.
• For a paid media brief, read it alongside the negative-audience and AI-traffic posts to judge whether the creator can work at both introductory and tactical levels.

The explanation follows Bob, a fictional vegan burger owner, and relies on film and superhero metaphors rather than campaign data.

Uses a fictional business scenario and pop-culture comparisons instead of walking through a real account.

For brands planning a campaign

A plain-language primer for clients new to paid search. It gives no spend, impression or conversion figures, so don’t use it as benchmark evidence in a media plan.

Examples: Source 1

The practical levers are consistent creative and retargeting. The explanation follows Jack, a fictional small-budget business owner selling to young professionals who like coffee and cats.

A fictional character walkthrough frames the algorithm as a helper that finds the audience.

For brands planning a campaign

A useful introduction to algorithm-driven targeting for novice or sceptical clients. It gives no audience sizes, budgets or results, so use it to explain the concept, not to build a media plan or forecast performance.

Examples: Source 1

The team removed those segments from prospecting and retargeting.

• The stated result was 19% fewer wasted impressions and a lower cost per acquisition, with no change to creative or budget.
• The post argues that teams should review negative audiences monthly as an active optimisation lever, rather than set them up once and forget them.

Exclusion lists built from conversion data and refreshed monthly.

The exclusions target weak converters, not non-converters—the opposite of the usual exclusion practice.

For brands planning a campaign

Addresses the goal of reducing wasted ad spend and offers a testable starting point. Before replicating the approach, ask for the agency’s conversion and return-rate data, plus the account, category and measurement period.

Examples: Source 1

The post argues that AI engines have already answered visitors’ questions and compared options before sending them to a site, so they arrive pre-qualified.

• It recommends a separate landing path for AI-referred visitors, with proof and specifics above the fold, visible pricing and one obvious next step.
• Track revenue per visit rather than sessions.

Route AI-referred visitors to a dedicated page experience rather than the generic homepage or blog post.

For brands planning a campaign

Addresses the goal of tracking AI-referred traffic and offers a concrete hypothesis and metric. The percentages are not tied to named dashboards or accounts; test them against a client’s own data before using them as a benchmark.

Examples: Source 1

Bivas Kumar Roy works through purchase-count differences between Shopify and Meta Ads, covering attribution windows, duplicate Purchase events sent by both Pixel and Conversions API, and consent limits on measurement data, and lays out an audit sequence of tracking, deduplication, data quality, consent and attribution. That is the same habit Md Tahidul Islam shows when treating wasted impressions and weak-converter exclusions as a data question rather than a creative one, and both write in numbered steps a reader can work through.

Bivas Kumar Roy focuses on measurement plumbing, such as event IDs, deduplication and consent mode, while Md Tahidul Islam writes more about audience selection and landing paths. Roy builds posts as structured checklists, while Islam's explainers use fictional business owners to carry the explanation.

Read original · Original 2

Josh Gregory describes what breaks once budgets cross a monthly threshold: shrinking ROAS, climbing acquisition costs, funnel gaps for cold traffic, Google and Meta overlapping, and no lifetime-value or offer strategy behind the spend. His remedies, including audience signals and exclusions to keep targeting sharp, align with Md Tahidul Islam's argument that teams should prune weak converters instead of only building lookalikes.

Josh Gregory writes from an agency angle for brands at a specific spend level, using bulleted problem-and-remedy lists and sometimes offering free Google Ads audits. Md Tahidul Islam reports single-account observations with a named figure, such as 19% fewer wasted impressions, and uses narrative scenarios in explainers rather than audit offers.

Read original · Original 2

Nicolò Pauletto argues that the platform's multimodal model reads creative and effectively sets distribution, making weak hooks and cluttered openings a distribution problem rather than a performance one: creative is now the targeting.

• Md Tahidul Islam makes a related argument from a different angle: the algorithm learns better from cleaner audience data once weak converters are excluded, and AI-referred visitors should land on a page built for their intent.
• Both mix direct instructional posts with first-person accounts of hard-won experience.

Nicolò Pauletto focuses on creative as the targeting signal and on platform changes such as Muse Spark and conversational shopping, while Md Tahidul Islam focuses on exclusions, tracking-style pruning and AI-search landing paths. Pauletto also writes more personal narrative posts about building skills over years of failed ventures, unlike Islam.

Read original · Original 2

Both posts work through the same buyer decision that stalls a SaaS agency's reporting: Meta Ads shows 103 purchases while Shopify or GA4 shows 82, and the first instinct — the platform is overcounting — is usually wrong.

• The discrepancy can come from attribution windows and click-through versus view-through credits; a Purchase event sent through both the browser Pixel and the Conversions API without working deduplication; Purchase firing more than once on a refresh or repeated checkout; test, staff or $0 orders; wrong event values and currencies; external payment flows such as PayPal and Klarna; or consent state affecting what is measurable.
• The useful part is the method: separate tracking errors from attribution errors first, then audit in a fixed order — tracking, deduplication, data quality, consent, attribution.
• Fixing duplicate events will not resolve a difference caused by attribution windows, and changing attribution settings will not fix a broken Purchase event.
• The second post adds practical checks a Shopify agency can run the same week: confirm Event IDs match across browser and server events, verify deduplication in Events Manager, test unique transaction IDs, walk the full journey from landing page through external payment redirect and cross-domain configuration to the confirmation page, and review consent-granted, consent-denied and consent-changed scenarios.
• For a SaaS agency whose client disputes the dashboard before any optimisation decision, this is a better starting point than a media-buying method: the measurement cannot be judged until it is settled.

The tradeoff is scope.

• The posts cover measurement and tracking engineering for ecommerce and EU advertising: Meta CAPI, Google Ads, GA4, GTM and Consent Mode V2.
• They do not address the agency's need to convert AI-referred traffic, and say little about campaign structure, creative or targeting efficiency.
• An agency that only needs to know which platform is right would get the diagnostic it needs; one that needs a campaign plan would still have to source media-buying and AI-search work elsewhere.
• Both posts are general troubleshooting guides, not results from a named account, so the audit order is a method to apply to the agency's own data, not a proven outcome.

Examples: Source 1 · Source 2

The posts name the threshold and symptoms: once budgets cross roughly £25–30K a month on Meta or Google, return on ad spend starts to decline, acquisition costs climb and results swing unpredictably.

• The default responses — changing bids, testing random creatives, duplicating campaigns — are treated as noise because they do not fix the system behind the ads.
• At that stage, money is lost through funnel gaps for cold audiences as warm buyers are exhausted by scaling, Google and Meta working in isolation so the channels overlap and pay for the same clicks, and the lack of a lifetime-value or offer strategy to support paying more to acquire a customer.
• The proposed structure is operational rather than a single lever: weekly creative testing cycles, margin and customer-acquisition-cost checks before budgets are raised, offers built for first-time buyers, a funnel progression mapped from awareness to purchase, and landing pages designed for cold clicks.
• The second post narrows to Google Ads and the common complaint that more spend is not moving sales.
• Its diagnosis: increasing the budget too quickly broadens targeting, dilutes the best audiences and drops conversion rate, so campaign structure, audience signals and exclusions, search terms, product feeds and creative have to evolve alongside the spend.
• For an agency asked to scale a client account rather than diagnose a dashboard, this answers the same buyer decision at the level of account structure and cross-channel planning.

The tradeoff is evidence depth and the commercial shape of the offer.

• These are structured account diagnoses written at a general level, with no named client, reported before-and-after figures or measurement window, so the agency gets a framework to apply rather than a tested result.
• The posts do not cover AI-referred traffic or AI search, so the agency would need another source for that goal.
• The second post also ends with a free Google Ads audit offer for ecommerce brands, making an audit conversation the practical next step rather than a published method the agency can run alone.
• For a B2B SaaS client, the ecommerce framing around product feeds and offers would need translating.

Examples: Source 1 · Source 2

The post on Meta's Muse Spark covers three changes with direct consequences for the agency's paid-media needs.

• First, the algorithm is natively multimodal and processes images and video the way it processes text, so it reads creative before serving it: weak hooks, cluttered frames and slow first three seconds are now distribution inputs, not just performance variables.
• Second, a conversational shopping layer inside Instagram, Facebook and WhatsApp pulls from creator content, community posts and brand storytelling already on the platform, so a brand absent from those conversations is absent from a discovery surface.
• Third, the post concludes that creative is the new targeting — bad creative has become a distribution problem rather than a performance problem — a shift that supports an agency's case for creative production budget instead of more media spend.
• That maps to the agency's paid-media needs and the broader question of how buyers now find products.
• It is the only one of the three alternatives that treats platform-level AI changes as the main subject.

The tradeoff is that this is forward-looking platform commentary, not a method with a measurable result, and the two posts pull in different directions.

• The Muse Spark post describes a rollout and its implications but offers no account data, test or figure the agency could verify, so it is best used to frame a creative strategy rather than justify a spend decision.
• The second post is a five-year personal account of failed businesses — dropshipping, affiliate marketing, self-made products — that ends with a focus on specialising in media buying, numbers and scaling systems.
• It shows persistence and offers an origin story, but contains no operational paid-media or AI-search framework.
• An agency looking for a diagnostic or repeatable process would get less practical detail here than from the other two alternatives.

Examples: Source 1 · Source 2