Delamon Rego

To get a figure worth acting on, run a matched-region holdout: split regions into two comparable groups, switch Meta off in one and leave it on in the other, run the test for three to six weeks—or longer at low order volume—then compare each group with its own pre-test baseline.

• The gap between the groups is incremental revenue.
• Divide it by what the platform reported for the same regions and weeks to get a correction factor to apply to every Meta ROAS figure the brand tracks, from campaign tables to dashboards to forecasts.
• That way, budget decisions follow the corrected number.

In one account, $127,500 of incremental revenue against $231,875 reported gives a factor of 0.55x.

• A campaign showing 4.0x should be treated as 2.2x, Rego says, and decisions should be based on that figure.
• Because the factor shifts with channel mix, creative and the platform's own attribution changes, the test is repeated twice a year, with each result checked against post-purchase survey answers.

Rego notes that the direction can reverse. High-average-order-value products with long consideration periods often land between 1x and 3x; in those accounts, Google is usually over-counted rather than Meta.

Read original

That is why Rego starts with a diagnosis, not a tactic.

• Plot every Meta ad that has spent more than $100, with lifetime spend on one axis and cost per acquisition on the other, then draw three lines: target CPA, a kill line at 1.5 times target, and a scale line at 0.7 times target.
• Where the dots fall identifies the problem.
• Almost nothing in the low-spend testing zone points to a creative-volume problem.
• Plenty of testing but few ads reaching the scale line signals a creative-quality problem: study what durable winners had in common and rewrite the production request.
• Money tied up in expensive ads indicates a media-buying problem: kill losers and scale winners daily.
• Sound click-through rates alongside high costs everywhere point to the landing page, offer, product page or checkout.

For the volume problem, Rego recommends launching at least one new ad with three to five concepts for every $10,000 of ad spend.

• Below that rate, they argue, the account will not generate enough winners to absorb new spend at a lower cost, and expensive ads in the kill zone will not be replaced.
• More creative will not fix a media-buying or conversion problem, which is why the diagnoses need to be separated.

Read original

Add a post-purchase survey question asking how long ago the customer first heard about the brand, filter for paid-channel answers, and use the replies to build a payback curve showing how much of each month's spend comes back in later months.

• Measure marginal ROAS separately with a geo holdout: drop spend 30% in one third of the country and raise it 30% in another, then fit the results to a spend-response curve.
• This makes the forecast reflect the return on the next dollar rather than the account average.
• Feed those inputs into a monthly waterfall: extra spend out, revenue in at marginal ROAS spread across later months, gross margin applied, and a running total.
• The lowest point is the cash the brand must have on hand; where the total crosses zero is payback.

In one worked example, adding $100,000 a month to a $1M base at 3.25x marginal ROAS and 52% gross margin produces a $172,000 cash trough in month four, payback in month seven, and $260,000 of profit by month twelve. Rego says this is most critical for brands with average order values above $150 or long consideration periods, when the gap between spending and getting paid is widest.

Read original

Tag every ad on three axes: format (static, vertical video, carousel, user-generated content), concept (product education, benefits, customer story, founder story, problem and solution), and hook (curiosity, problem-first, visual demo, authority).

• Track the tags in UTMs if they are not already there.
• Build one grid per format, with concepts across and hooks down, and put the spend-weighted ROAS of each cell's ads in it.
• Read one grid at a time, since the best and worst combinations differ by format.
• The findings should shape the brief: instead of asking for more statics, request more product-education statics with visual-demo hooks.

Rego says a client's grid showed strong concept-and-hook combinations in one format failing in another.

• Production should stop on combinations a format has ruled out and focus on the top three per format.
• They say this approach has cut client CPA by up to 15%.
• The grid example is attributed to a client and shown in an image, so the result is presented as reported practice, not an audited figure.

Rego separates volume from quality.

• When almost no ads reach the testing zone, the answer is to make more ads.
• The grid helps choose among an existing library; drop any combination that wins in one format but loses in another.

Read original

An ad needs more than $100 of spend to make the diagnostic plot; a grid cell needs five ads or $3,000 of spend before it is read; a geo holdout needs three to six weeks, or longer at low order volume; and a Meta correction factor is retested twice a year because channel mix, creative and platform attribution all shift.

The thresholds serve different purposes in the same account: $1,000 of spend marks the testing zone in the scaling chart, $3,000 or five ads gates a cell in the creative grid, and the holdout length scales with order volume rather than staying fixed.

Read original · Original 2 · Original 3 · Original 4

It answers the first question a newcomer asks: when a DTC brand's customer acquisition cost stalls, what is actually broken? Instead of a list of tips, it offers one diagram and five named diagnoses, letting you place any brand you are evaluating on it.

Rego's method plots every Meta ad the brand has ever run with more than $100 in spend on a single chart: lifetime spend on the horizontal axis, cost per acquisition on the vertical.

• Ads under $1,000 in spend form a testing zone on the left; three horizontal lines mark the target CPA, a kill line at 1.5x target and a scale line at 0.7x target.
• The dot clusters identify the bottleneck.
• Too few ads in the testing zone signals a creative volume problem, solved by launching at least one ad with three to five concepts per $10,000 in spend.
• Plenty of ads testing but few under the scale line signals a creative quality problem, solved by studying what durable winners shared and rewriting the creative direction.
• Money tied up in expensive ads signals a media buying problem.
• If every ad is too expensive despite good click-through rates, the problem is conversion, which belongs to the landing page, offer, product page and checkout—not more content.
• Their argument is that these five failures look similar in a dashboard but require completely different spending decisions.

For brands planning a campaign

Use the five diagnoses to scope a paid media engagement before proposing tactics. A creative brief fixes a volume problem, not a media buying problem, so the audit output—not the pitch—should determine which work gets bought.

It is the shortest of the four posts and most directly challenges numbers a brand already trusts, making it a useful first read for anyone auditing reported ROAS.

Rego describes running a geo holdout rather than trusting platform attribution.

• Regions are split into two matched groups; Meta is turned off in one and left on in the other.
• The test runs three to six weeks, longer when order volume is lower.
• Each group is compared with its own pre-test baseline; the gap between the groups is incremental revenue.
• That figure is then divided by what Meta reported for the same regions and weeks to produce a correction factor.
• In the worked account, $127,500 divided by $231,875 gives 0.55x, meaning a campaign reported at 4.0x is really 2.2x and should be judged at 2.2x.
• They note that products with high average order values often run the other way, landing between 1x and 3x, with Google usually the over-counted channel.
• The factor is retested twice a year because it shifts with channel mix, creative and Meta's own attribution changes; each result is checked against post-purchase survey answers.

For brands planning a campaign

Budget and forecast using corrected ROAS, and treat the correction factor as a maintained figure to be retested on a regular cadence, not a one-off audit.

It answers the question a brand asks after deciding to increase spend: what will this do to cash, and when will it come back? The worked numbers make it the most concrete of the four posts.

Rego starts with the post-purchase survey, adding the question "how long ago did you first hear about us?" with answers bucketed from last week to more than four months ago, alongside the existing first-touch question.

• Filtering responses to paid channels produces a payback curve showing how each month's spend returns over subsequent months.
• Marginal ROAS is then measured with a geo-holdout test: spend drops 30% in one third of the country and rises 30% in another.
• A spend response curve is fitted so the number reflects the next dollar rather than the average.
• Those inputs feed a monthly waterfall: extra spend out, revenue in at marginal ROAS spread across later months, gross margin applied, and a running total whose lowest point is the cash required and whose zero crossing is the payback month.
• The example adds $100,000 a month on a $1M base at 3.25x marginal ROAS and 52% gross margin, producing a $172,000 cash trough in month four, payback in month seven and $260,000 of profit by month twelve.
• They flag brands with average order values above $150 and long consideration periods as needing this most.

For brands planning a campaign

When a brand proposes a spend increase, ask for the marginal ROAS test and the cash trough before agreeing to the budget—not after.

It answers a narrow but common question: when a brand already makes plenty of ads, how do you find out which ideas and openings are worth repeating? It is a useful companion to the $100M scaling post because it addresses the creative quality branch of that diagnosis in detail.

Rego tags every ad on three axes—format (static, vertical video, carousel, user-generated content), concept (product education, benefits, customer story, founder story, problem and solution) and hook (curiosity, problem-first, visual demo, authority)—and tracks them in the URLs.

• One grid per format puts concepts across and hooks down, with each cell showing the spend-weighted ROAS of the ads in it.
• Read the grid only after five ads or $3,000 in spend have accumulated.
• It reveals the best and worst concept-and-hook combinations for that specific format; they report that strong combinations in one format frequently fail in another.
• The briefing change follows: make more of the top three combinations per format, stop producing combinations a format has already ruled out, and replace "we need more statics" with "we need more product education statics with visual-demo hooks."
• They attribute the grid example to a client and close with an offer to help analyze someone's content.

For brands planning a campaign

Ask for the tagged creative grid in a creative review; it turns a debate about taste into a decision about which concept-and-hook pairs to fund in each format.

Where the dots cluster reveals the problem.

• A healthy account has a decent number of ads in the testing zone (under $1,000 of spend), a handful of heavily scaled ads carrying most of the spend, a couple of promising ads picking up budget, and nothing meaningful spending above the kill line.
• They map five failure shapes onto that picture: a creative volume problem (too few ads and almost none in testing, with a suggested launch pace of one new ad carrying three to five concepts per $10k of ad spend); a creative quality problem (plenty of testing but few ads ever come under the scale line, so the fix is raising the hit rate by identifying what durable winners share); a media buying problem (money sitting on the wrong ads, losers not killed, winners not scaled); and a conversion problem (sound creative and buying, but traffic that does not convert, pointing to the landing page, offer, product page or checkout).

One graph, three reference lines and a five-way bottleneck checklist applied to ads the brand has already run.

For brands planning a campaign

A brand can hand Rego an existing account and get back a structural read, not a list of tactics: which of the five bottlenecks the account has, and whether the answer is more creative, better creative, better buying or a conversion fix. For a paid media measurement brief, the recommendation ties a specific observable pattern in the ad data to a change the brand can check against its own dashboard before adjusting spend.

Examples: Source 1

Tag every ad on three axes — format (static, vertical video, carousel, UGC), concept (what the ad is about: product education, benefits, customer story, founder story, problem/solution) and hook (how it opens: curiosity, problem-first, visual demo, authority) — using UTMs if the tags are not already in place.

• Then build one grid per format, with concept across and hook down, and spend-weighted ROAS in each cell.
• Read a cell only once it holds five ads or $3K of spend.
• Analyze one format at a time, because winning and losing concept/hook combinations differ by format.
• The output guides a campaign aimed at specific audiences: instead of asking for "more statics," ask for "product education statics with visual-demo hooks."

Three-axis tagging, one grid per format, minimum cell thresholds of five ads or $3K before reading a cell, then rewritten creative briefs naming the top concept/hook combinations per format.

Rego notes that a combination strong in one format may fail in another; drop it for that format rather than repeat it.

For brands planning a campaign

For a brand briefing creative work, this turns an opinion about what to make into a rule about what to stop making.

• The brand can require concept, hook and format tagging in UTMs, then use the grid to decide which combinations get more production budget.
• Rego claims the outcome is a CPA reduction of up to 15%, illustrated with an unnamed client example.

Examples: Source 1

You risk short-term cash losses" sets out a measurement chain for brands whose ad spend is ahead of the cash it returns.

• Add a "how long ago did you first hear about us?" question to the post-purchase survey alongside the existing "how did you very first hear about us" question, then filter to paid-channel answers to build a payback curve showing how much of each month's spend comes back in later months.
• Next, estimate marginal ROAS with a geo holdout — drop spend 30% in a third of the country and raise it 30% in another third — and fit a spend-response curve so the forecast uses the return on the next dollar rather than the account average.
• Finally, build a monthly cash waterfall crossing extra spend, revenue at marginal ROAS, gross margin and the timing from the payback curve.
• The lowest point of the running total shows the cash the brand must have on hand; the crossing point shows payback.
• Rego works through an example: $100K/month added to a $1M base, 3.25x marginal ROAS, 52% gross margin, a $172K cash trough in month four, paid back by month seven, and $260K profit by month twelve.

Post-purchase survey question, paid-channel payback curve, 30% up/down geo holdout for marginal ROAS, spend-response curve, and a month-by-month cash waterfall across several spend scenarios.

For brands planning a campaign

This section speaks most directly to a paid media measurement overhaul.

• A brand weighing a large spend increase can ask Rego for a scenario model showing the cash required, when the spend pays back, and what each spend level does to profit before committing.
• They flag the method as most critical for brands with an average order value above $150 or long consideration periods.

Examples: Source 1

Run a geo holdout — split regions into two matched groups, turn Meta off in one, leave it on in the other, run three to six weeks (longer at low order volume), and compare each group against its own pre-test baseline.

• The gap between groups is incremental revenue.
• Divide that by what Meta reported for the same regions and weeks to get a correction factor; in the worked account, $127,500 incremental against $231,875 reported gives 0.55x.
• Multiply every Meta ROAS the brand tracks — campaign tables, dashboards, forecasts — by that factor, so a campaign reported at 4.0x is treated as 2.2x and budget decisions follow the corrected number.
• Rego recommends re-testing twice a year because the factor moves with channel mix, creative and Meta's own attribution changes, and checking each result against the post-purchase survey.

Matched-region geo holdout over three to six weeks, an incremental-to-reported ratio applied across reporting, and twice-yearly re-testing against post-purchase survey responses.

Rego notes that when a factor comes in above 1.0x — more common for high average order value products with longer journeys — Meta may be under-attributing. In those accounts, Google is usually over-counted.

For brands planning a campaign

For a brand whose media decisions rely on platform-reported ROAS, this offers a concrete deliverable: a measured correction factor applied wherever the number is used, with a defined re-test cadence. It also gives a prospective partner a clear test: will they propose a holdout and corrected number, or accept platform attribution as given?

Examples: Source 1

Each opens with a blunt claim ("Meta overcounts revenue," "How $100M+ brands lower CPAs 15%+"), then walks through a numbered sequence of four to six steps in plain language, with arrows and short lines rather than paragraphs.

• The steps are framed as something the reader can do this week: build a graph, add a survey question, tag ads in UTMs, run a holdout.
• Two of the four close by handing the decision back to the reader — "Which of these is YOUR bottleneck right now?" and "If you have never tested, you do not know where you sit on this scale."
• Two close with an offer to help directly, including one that points to a client example.
• Arguments recur across posts rather than contradicting each other: incrementality over reported attribution, the current dollar over the average, and named bottlenecks over general advice.

For brands planning a campaign

A brand evaluating Rego can expect a diagnostic framing — a defined problem, a numbered method and a specific number to act on — rather than a broad growth retainer pitch. Ask how the work is staffed and what it costs; the material does not discuss team structure or pricing.

Examples: Source 1 · Source 2 · Source 3 · Source 4

They do not say whether they own FORGE, work there, or contract with it; ask what their relationship is to the business.

View example

Shakoor takes a more reflective, pipeline-oriented approach, with less emphasis on measurement design. Rego adds diagnostic charts, geo-holdout correction and cash modelling to guide spend decisions.

Read original · Original 2 · Original 3 · Original 4

Nyadu focuses on creative velocity and aggressive budget scaling. Rego focuses on the measurement behind a scale decision, insisting on marginal ROAS, payback and corrected platform numbers.

Read original · Original 2 · Original 3 · Original 4 · Original 5

Tesselaar's examples centre on Google Ads and Performance Max, with structures for branded traffic and margin-split asset groups. Rego's are Meta-specific, built around geo holdouts, creative grids and cash-flow modelling.

Read original · Original 2 · Original 3 · Original 4

Shakoor argues that relying on a single winning Meta ad is risky because the account depends on one hook, angle and message.

• The operator, they say, should always be testing the next hook, angle, customer problem and format.
• For a scaling ecommerce brand whose corrected numbers show spend concentrated in a few aging winners, that is a more direct answer than another attribution model.

The tradeoff is scope.

• Delamon Rego's posts do not describe a geo holdout, marginal ROAS, payback curve, cash waterfall or attribution correction factor.
• A brand that needs to know whether Meta-reported revenue is real or how much cash a spend increase will consume before payback would still need their measurement work or another specialist.
• Shakoor's posts make a creative-testing argument, not a measurement overhaul.

Examples: Source 1

The posts argue that brands scaling past $500k a month launch 40–60 new creative assets monthly, kill losers in 72 hours and scale winners until fatigue.

• A second post says top performers doubled budgets overnight at signal thresholds and scaled from $50k to $150k in 18 days.
• For a brand whose bottleneck is creative velocity or timid budget increases, that is closer to the operating question than a holdout design.

The tradeoff is that these posts do not explain how to validate the revenue numbers those decisions depend on.

• They include no geo holdout, marginal ROAS calculation, payback curve or cash waterfall, and present performance claims as analysis of 31 brands without disclosing the dataset.
• The advice to double budgets overnight also sits uneasily with Delamon Rego's cash-trough warning for brands with high average order value or long consideration periods.
• A brand that needs both aggressive creative volume and a measurement guardrail would need to combine the two approaches, not substitute one for the other.

Examples: Source 1 · Source 2

The posts argue that ROAS is not return because it ignores cost of goods, shipping, returns and fulfillment.

• They recommend POAS (gross profit divided by ad spend) and break-even ROAS at 1 divided by gross margin, and describe how to protect branded traffic, split PMax asset groups by margin, feed CRM and lifetime-value data, and track conversion value rather than volume.
• For a brand whose paid search or PMax spend is the main profit leak, that is a more direct fit than a Meta-centric correction.

The tradeoff is channel and method.

• These posts focus on Google Ads and PMax profitability; they provide no Meta geo-holdout factor, post-purchase payback curve or monthly cash waterfall.
• If the need is specifically to correct Meta-attributed revenue and model the cash cost of a Meta spend increase, Delamon Rego's posts address that task more directly.
• Tesselaar is the stronger alternative when the buyer's task shifts to Google Ads margin and new-customer measurement.

Examples: Source 1 · Source 2