# Ben Heath: the full methodology

Every rule below comes from Ben Heath's own videos, mostly 2024–2026. Tags like
`[13s-G9Uj51A 2026-02]` are YouTube video IDs with their upload month; see `sources.md`
for the URL of each. **[caption?]** marks a number that may be an auto-caption error.
Where his advice changed over time, the newest date wins; see section 16.

---

## 0. Priority order (his end-to-end framework)

From [crId74rHjuA 2025-10], [kuSq-pmNfnM 2025-09], [3ReT1bzLFhU 2026-10] and [6TRy9eUAPVg 2026-08].

1. **Offer.** "By far the most important thing." A great offer rescues bad targeting, copy and
   creative; the reverse never happens. "A really good campaign is not going to make up for a
   bad offer." [8dJxOaZrK1M 2026-03]
2. **Creative.** "After your offer, creative is the most important part… we are in a creative
   platform."
3. **Structure.** Keep it simple. "If you have a nine-step sales funnel, you only need one
   step to not work well and the whole thing breaks."
4. **Targeting.** "Much more about don't mess it up as opposed to get it right."
5. **Tracking and data.** Pixel plus Conversions API at minimum; check it against the CRM.
6. **Budget and scaling.** Start with an amount you can afford to lose. Scale only once
   profitable, in steps.
7. **Bonus: LTV.** "He who can afford to pay more to acquire the customer wins."

"Go beyond the buttons": exceptional results always come from an advantage outside the ad
account, such as brand, product, offer, proof or creators [kuSq-pmNfnM 2025-09].

---

## 1. Know your numbers first

- Before scaling or judging anything, know your target cost per lead (CPL), cost per
  acquisition (CPA) or return on ad spend (ROAS). These come from margins, operating costs and
  customer value. He calls this "really important to know your numbers"
  [FyWyHJh_6Ng 2026-05].
- Set the **maximum you will pay to acquire a customer** from the customer's value, not from
  gut feel. His example: a customer worth $2,000, but the owner will "only pay $200". That
  caps CPL at $20 and blocks scaling, when $25+ per lead would have been very profitable
  [XLagiyzYYpE 2026-02].
- **Required ROAS depends on the business model.**
  - 1.5x can be great for software or info products with no cost of goods.
  - Others need far more.
  - 3.2x at scale is a success.
  - "Much better to generate a 4x… spending a million dollars a month than a 10x… spending
    $10,000 a month." [XLagiyzYYpE 2026-02]
- **Realistic expectations.** A "$30 cost per purchase" is realistic; anyone who promised $6
  was "misleading you". A CPM of $100 or more is normal in lucrative niches. "High CPM doesn't
  equal bad." [3ReT1bzLFhU 2026-10]
- **LTV is the lever.**
  - Starbucks: about $7 per visit vs "over $14,000" lifetime value, so it can pay $200–$500+
    per customer [6TRy9eUAPVg 2026-08].
  - Raise customer value "from $120 to $320" rather than chasing CPA "from $26 to $23".
  - Levers: subscriptions, upsells, repeat purchases, referrals, follow-up sequences.
- **Headroom before scaling.** "If you need to be at say a 3x ROAS… you want to be at like a
  4x maybe… close to a 5x at your current budget level before you start to scale."
  [crId74rHjuA 2025-10] Earlier version: "aim to run at 5x when 3x is acceptable"
  [846INGFormA 2024-11].
- **ROAS improvement compounds.** "Sometimes a 20% improvement in return on ad spend is all
  you need." At a 2.5x breakeven, it can allow 5–10x the spend [846INGFormA 2024-11].

---

## 2. Account foundations and tracking

- **Pixel plus Conversions API (CAPI).** "Not an either or." [kuSq-pmNfnM 2025-09]
  - "The pixel is no longer enough." Events Manager now has a one-click CAPI setup
    [6TRy9eUAPVg 2026-08].
  - On WordPress, use the official Meta plugin, tick "send website events using Conversions
    API" (often off by default), and turn on Automatic Advanced Matching.
  - On Shopify, it's a checkbox [7DzaSOj7o4c 2023-11].
- **Outsource the setup if needed.** Fiverr or Upwork, about $20. "If you have money for ads,
  you have money to pay someone." [crId74rHjuA 2025-10]
- **Check events fire correctly.** Use the Meta Pixel Helper extension. Only PageView should
  fire on landing-page load; purchase or lead events must not.
  > "You're actually much better off with a Meta pixel that doesn't know how to get you
  > results than one that thinks it is getting you results when it actually isn't."
  > [vaBxYgZ7MAU 2026-05]
- **Event match quality.** Check Events Manager suggestions regularly. Verify Meta's numbers
  against the CRM or Shopify [crId74rHjuA 2025-10].
- **Audience segments.** In Advertising settings → Audience segments, define the **engaged
  audience** (website visitors 180 days, page and Instagram engagers, video viewers, lead
  forms, email list) and **existing customers** (customer list, Shopify, purchase audience).
  - This mostly affects reporting: the breakdown splits results into new, engaged and
    existing.
  - If everything shows as "Unknown", the segments aren't defined
    [YZ_zNzXHOio 2026-03; 13s-G9Uj51A 2026-02].
- **Upload external data** (email and customer lists with a value column and labels).
  - This covers people outside Meta's lookback windows.
  - Purchase custom audiences now look back 730 days, up from 180 [X2Iww-AH7Zs 2026-10].
- **Attribution.**
  - Meta's maximum is a 7-day click window. Engaged-view and view-through windows are 1 day.
    The default setting is 7-day click, 1-day view, 1-day engaged-view
    [WRyAUXjBlJM 2025-04; 6TRy9eUAPVg 2026-08; bv8AsY0xkIk 2026-05].
  - Meta under-reports recurring and long-cycle revenue.
  - His example: a £97/month product at £23 CPA looks like 4x on the first payment but is
    "more like 20x" over the customer's life [mycqb92wJqk 2024-09].
- **Third-party attribution at reasonable budgets.** Hyros is his sponsor and he says he
  uses it. Treat the figures from its ad reads (e.g. "£96k tracked, £58k not reported by
  Meta") as sponsor content.
- **Incremental attribution** (ad-set attribution model).
  - For larger advertisers whose budget leaks to warm buyers who would have bought anyway.
  - Results will "look worse" in reporting.
  - Not for beginners or small customer bases [6TRy9eUAPVg 2026-08].
- **Never boost posts** from the apps. Use Ads Manager. Boosting from iOS adds a 30% "Apple
  tax" [vaBxYgZ7MAU 2026-05].
- **Declare special ad categories** (credit, employment, housing, social or political), or
  risk the account being disabled [-CsBl3tv-X0 2025-06].

---

## 3. Objectives and performance goals

- **The performance goal is the single most important setting.**
  > "You can completely undo that by selecting the wrong performance goal."
  > [WRyAUXjBlJM 2025-04; 846INGFormA 2024-11]
- **Use Sales or Leads** for almost everything. In his poll, "almost 85%" said Sales or Leads
  performs best. "Please don't try and be extra clever and game the system" with traffic
  campaigns [-6okjIwMmmU 2025-03].
- **"Warming up the pixel"** with Traffic or Engagement is "all absolute nonsense"
  [vaBxYgZ7MAU 2026-05].
- **Hidden traps.** Inside a Sales or Leads campaign:
  - A link-click or landing-page-view goal turns it into a traffic campaign.
  - A reach or impressions goal turns it into an awareness campaign.
- **Meta data he cites:** optimising for purchases instead of link clicks gives "74% more
  sales with the same budget" [846INGFormA 2024-11].
- **Pick the event.** Optimise for "the action that is furthest along your sales funnel that
  you can accurately track", within the 7-day click window [-6okjIwMmmU 2025-03].
  - "More than 20-25 purchases a week, absolutely go with purchase." Otherwise move up a step
    (booked call, webinar attendee, registrant) [6TRy9eUAPVg 2026-08].
  - Example from his Learning Limited video: 7 purchases a week but 35 add-to-carts, so test
    add-to-cart for about a month. At around 20 purchases a week with room to scale, stay on
    purchase [dlVx7HgKL_I 2026-03].
- **Value vs count** (his advice moved toward value over time):
  - 2025: count is fine by default; value is better once set up [WRyAUXjBlJM 2025-04].
  - 2026-08: for e-commerce, optimise for value "provided you have enough conversions", not
    with "a handful a week".
    > "If it cost me $10 to get a $20 sale, I would much rather spend $30 to get a $200
    > sale." [6TRy9eUAPVg]
  - 2026-09 (newest): "most of them should change this" to **Maximize value of conversions**,
    even when Meta doesn't recommend it [S-F3yxya2WE].
  - A single-product business can stay on count.
- **Qualified-lead optimisation.** Only qualified leads should fire the lead event. See
  section 11.
- **When other objectives are acceptable:**
  - **Traffic:** a third-party site with no pixel (e.g. Amazon), a blog post, or an Instagram
    profile. Use landing page views over link clicks when there's a pixel.
  - **Engagement:** page likes, groups, messaging (Maximize conversations), or ThruPlay views.
  - **Awareness:** big brands, or his omnipresent content strategy.
  - **App:** optimise for app events over installs.
  [-6okjIwMmmU 2025-03; WRyAUXjBlJM 2025-04]
- **You can't change an existing campaign's objective.** Recreate the campaign instead.
- **"If in doubt, go with the default option"** once the objective is right.

---

## 4. Campaign structure

### Default (2025–2026)
- **Consolidate.** "Run as few campaigns and as few ad sets as possible. Often just using the
  one campaign and the one ad set." [13s-G9Uj51A 2026-02]
  - Why: concentrated conversion data. 70 conversions a week in one ad set beats 7 in each of
    10. It also avoids **auction overlap**.
  - Auction overlap is not the same as audience overlap. Meta plans impression frequency per
    person, e.g. "four or five impressions over 24–48 hours". It can't plan that across
    competing ad sets or campaigns.
- **Don't split cold and warm.** Custom audiences now sit in "suggest an audience", so Meta
  reaches both in either ad set anyway. Separate cold and warm campaigns are "a little bit
  silly" now [13s-G9Uj51A 2026-02; YZ_zNzXHOio 2026-03].
- **Don't test audiences in separate ad sets** (interest A vs interest B vs lookalike). Meta
  treats them as suggestions and ignores them. **Exception:** location, which is a hard
  control. Test different countries or cities as separate ad sets or campaigns
  [13s-G9Uj51A 2026-02].
- **One campaign per product range or service.** Hats vs shoes, not red hats vs blue hats.
  Different offers need different optimisation, and you can switch off a range that's out of
  stock or at capacity [13s-G9Uj51A 2026-02].
  - A general dynamic catalog campaign can span all ranges. At small budgets, it mostly
    retargets.
- **Funnel stages live inside one ad set.** Create top, middle and bottom-of-funnel ads
  (founder story, demo, testimonial) in the same ad set and let Meta sequence them.
  - For higher-priced or more involved offers, some ads get spend but few conversions because
    Meta uses them as top-of-funnel ads.
  - "You don't want to make the mistake of turning off the ad that is being used in a
    top-of-funnel capacity." [13s-G9Uj51A 2026-02]
- **Advantage+.**
  - "The majority of advertisers should be using Advantage Plus campaigns."
    [kuSq-pmNfnM 2025-09]
  - Advantage+ Shopping became **Advantage+ Sales**: one setup with campaign, ad set and ad
    levels, plus Advantage+ Leads. Campaign budget is the default [6avwp4gBZsQ 2025-03].
  - "Campaign score" of 100 means every recommended setting is on. A higher score usually,
    but not always, means better results.
- **Ads per ad set.**
  - Old Meta rule: no more than 6. 2024: "use 6 or fewer"; at $10–20 a day, use 1–3 ads
    [cqNsMJPoM7c 2024-05].
  - Now: "the recommendation typically is 20 plus and we will often run ad sets with 20 plus.
    I don't worry about an upper limit anymore. The constraining factor is usually how many
    good quality ads can be created." [13s-G9Uj51A 2026-02]
  - Hormozi's teams run "20, 30, 50 more" [oTWfDI_91Fs 2026-09].
  - Small budgets: "don't necessarily run them all at once" [a6-l6BD3dDg 2025-11].
- **Copy variations go inside one ad, not separate ads.** Use up to 5 primary texts, 5
  headlines and 5 descriptions per ad. Separate ads are for different creative
  [13s-G9Uj51A 2026-02].
- **Budget type.** Use daily budgets rather than lifetime budgets (more flexible, easier to
  scale) [FyWyHJh_6Ng 2026-05]. Lifetime budgets are only needed for ad scheduling or ad
  sequencing. With one ad set, CBO vs ABO "doesn't make any difference"
  [mycqb92wJqk 2024-09].

### When to deviate (do it deliberately and ignore Meta's warnings)
- **Location or market tests.** Separate countries or cities.
- **Different product ranges or offers.** Separate campaigns.
- **Omnipresent content.** 14 ad sets; see section 12.
- **Testing.** A separate testing campaign, only at enough volume; see section 7.
- **Large budgets** (around $100k a month). Multiple campaigns by product, offer or funnel
  stage (e.g. a lead magnet campaign plus a retargeting campaign to book calls), with 10–30%
  of budget in testing campaigns [du6rDgRf06g 2026-08].
- **True retargeting-only offers** (warm-only offers, upsells from a small purchase to high
  ticket). See section 6.
- **A messy account reset.** "Turn everything off and then create one campaign with one ad
  set, use Advantage Plus audience… best offer… best ad… give it a week, don't mess around
  with it." [846INGFormA 2024-11]

---

## 5. Targeting

### Controls vs suggestions [-CsBl3tv-X0 2025-06; 6avwp4gBZsQ 2025-03; crId74rHjuA 2025-10]
- **Controls are hard limits.**
  - Location (where you can actually serve). A city gets a 25-mile radius by default,
    adjustable from 10 to 50 miles. "Don't overthink your targeting."
  - Minimum age, only for legal or regulatory reasons (selectable from 18 up to 25).
  - Language, only where it really matters.
  - Custom-audience exclusions.
- **Suggestions are soft.**
  - Age range (e.g. 25–55 is fine).
  - Gender, only if the skew is about 90%. "If it's like a 60/40, probably just leave all."
  - Detailed targeting: one to five interests. "You can't really get this wrong." It helps
    most in new accounts.
  - Custom audiences (put them all in).
  - Lookalikes.
- **"Further limit the reach of your ads"** (switching to manual, hard targeting) is
  "likely to cause more harm than good". "You need a really good reason."
  - Detailed targeting can never be hard.
  - If you go hard, warm and cold must be separate ad sets, and lookalikes count as cold.
  - Valid reasons: leads aged 55+ never convert; women's jewellery delivered mostly to men.
- **Targeting importance.** It "used to be incredibly important" and "just doesn't matter
  anywhere near as much as it used to". "As long as you don't make a big mistake like…
  only targeting a really small audience you're fine." [846INGFormA 2024-11]
- **Open targeting** (location only): "one of the best performing targeting options… for a
  number of years". It works better for sales than for leads. It needs conversion data, and
  new accounts struggle with it [-CsBl3tv-X0 2025-06].

### Custom audiences he always builds [-CsBl3tv-X0 2025-06; hHuxr_ancXQ 2025-11]
- All website visitors, **180 days** (the maximum; the default is 30).
- Customer list and email list, uploaded separately.
  - At least 100 people. Email is the main identifier; add more identifiers to improve
    matching.
  - Add a **customer value** column and audience labels (qualified, disqualified, existing,
    high or low value).
- Video viewers, **3 seconds, 365 days**.
- Instagram engagers and Facebook page engagers, "everyone who engaged", **365 days**.
- Lead form openers, **90 days**, if you use instant forms.
- Always choose the broadest event and the longest retention. Meta prioritises the
  highest-intent people itself.
- Uploaded lists don't update unless synced (e.g. via Mailchimp). Refresh them manually.

### Exclusions
- He is not a fan by default. Don't exclude past purchasers: they rebuy, leave good comments
  and refer friends.
- Exclude only when a group must not see something, e.g. a promo price existing customers
  would resent [-CsBl3tv-X0 2025-06].
- Excluding custom audiences "really seems to disrupt delivery". His example: a $200/day ad
  set suddenly spending $50/day [X2Iww-AH7Zs 2026-10].
- To focus on new customers, use the ad-set **Customer Lifecycle Strategy → "Acquire new
  customers"** instead.
  - "By default, no." Use it mainly for big-share brands or one-time purchases.
  - Don't also exclude the engaged audience: "they're your best prospects."
  - Customer-list match rates can be about 50% [X2Iww-AH7Zs 2026-10].

### Lookalikes
- "Now much less valuable", because Meta does lookalike modelling automatically. They are
  still useful when Meta optimises badly, which happens more in lead gen.
- Sizes: US 1% (about 2.8M), UK 2%, Ireland 8–10% (about 400–500k).
- The source audience can be in another country.
- They are cold audiences, so never put them in warm ad sets [-CsBl3tv-X0 2025-06].

### Value rules: "the new version of targeting" [bv8AsY0xkIk 2026-05; 4eTmzmDLkjA 2026-11; 1Or1B4L_bdY 2025-12]
- **Path:** Advertising settings → Value rules → Create a rule set.
- **Criteria:** age, gender, location, conversion location (website vs instant form),
  device, mobile OS and placement.
- **Limits:** up to 10 rules, and two criteria can be combined in one rule.
- **Apply per ad set.** The rule does nothing until it is applied.
- **Bid up for segments worth more.**
  - Example: women +80%.
  - Meta's default is +50%. Set it to your real measured uplift: if a segment is 30% more
    valuable, use +30%.
  - > "If our overall cost per purchase goes up by 10%, but women on average are worth 2.4
    > times what men are, that's a great trade."
- **Base rules on data Meta can't see:** lifetime value beyond the attribution window,
  refund rates, lead-to-client rates. Don't base them on where purchases already come from,
  which Meta already sees [6TRy9eUAPVg 2026-08].
- **Do the homework first.** Spend "90 minutes, 2 hours" with the CRM and ad data before
  setting percentages.
- **Gender hard targeting was removed (Nov 2026).** Gender exists only as a suggestion. The
  workaround is a value rule with **decrease bid by 90%** (the maximum) on the unwanted
  gender, applied per ad set [4eTmzmDLkjA 2026-11].
- **Placement value rules.** For example, Audience Network −90%. Use these now that
  ad-set-level placement and platform controls are being removed [qkXMTQfLgvE 2026-09].
- **Don't use value rules in omnipresent content campaigns** [hHuxr_ancXQ 2025-11].

### Placements [-2Uevcin8eg 2026-01; qkXMTQfLgvE 2026-09; 1Or1B4L_bdY 2025-12]
- **True Sales or Leads campaigns** (conversion goal, real purchase or lead event): use
  **Advantage+ placements**. Don't trim placements based on breakdown reports.
- **Traffic, engagement, awareness, reach or link-click goals:** Meta defaults to the
  cheapest inventory.
  - Restrict to Facebook and Instagram feeds, stories and reels.
  - Remove Audience Network, Messenger, Threads, in-stream and search.
  - Untick "Allow limited spend to excluded placements" (otherwise "up to 5%" of budget goes
    to each excluded placement).
- **Since Sept 2026,** ad-set placement, platform and device controls are being removed.
  - Switch to Sales or Leads where possible.
  - Use account-level placement controls (Advertising settings → Account controls), or a
    placement value rule (Audience Network −90%).
- Customise creative for the 3 main placement shapes (feed, vertical, landscape). Put
  captions in different positions for reels and stories [846INGFormA 2024-11].

---

## 6. Retargeting

- **The default is a hybrid ad set.**
  - Put all custom audiences in as suggestions. Meta then spends dynamically on new, engaged
    and existing people.
  - His example of a "cold" ad set: about 40% of spend went to engaged plus existing
    customers, and more than half of its 942 purchases came from them, at lower CPA
    [YZ_zNzXHOio 2026-03].
- **Make creative work for both warm and cold** audiences.
- **No fixed retargeting budget share is given in any video.** Meta allocates it.
- **Keep building custom audiences and the pixel.** Calling the pixel unnecessary "is
  absolutely false". Custom audiences define the audience segments and still enable true
  retargeting.
- **True retargeting-only** (warm-only offers, ascension to high ticket, special warm
  discounts):
  - "Further limit the reach" → add custom audiences → untick **"Use as suggestion"**.
  - "Targeting is never 100% accurate." It is not his default.
  - Earlier method (2025-03): a warm ad set using the original audience options with
    "Advantage custom audience" deselected, plus a cold ad set that excludes the warm
    audiences [raSBSNu1fYE].
- **Instant-form openers who didn't submit** are a useful retargeting pool. The pool grows
  once SMS verification is on [Cz8Bi6_rkwM 2025-05].
- **High-ticket offers** (over about $1,000) should run omnipresent content alongside the
  conversion campaign [raSBSNu1fYE 2025-03].
- **Ad sequencing.** Meta sequences ads automatically in a single ad set, more often when
  the ask is bigger. A formal sequencing feature exists but requires lifetime budgets, and he
  doesn't recommend it for most advertisers [hHuxr_ancXQ 2025-11; v_u9qcbW5hY 2026-01].

---

## 7. Learning phase, change cadence and testing

### Learning phase [dlVx7HgKL_I 2026-03; XLagiyzYYpE 2026-02; vaBxYgZ7MAU 2026-05]
- It typically lasts **24–48 hours** and depends on conversion volume.
- It **re-enters** when you launch an ad set, make significant changes, or add new ads.
- **Meta's threshold** is "50 results per week" of the optimised event.
  - In practice, ad sets sometimes exit at 20–40.
  - He has never seen Learning Limited above 50 a week.
- **Fixing Learning Limited:**
  1. Consolidate ad sets. Five ad sets at 15 purchases each become one ad set at 75.
  2. Improve performance.
  3. Keep things stable.
  4. Optimise higher up the funnel only if 50 a week is unrealistic.
  - Don't just raise the budget on an unprofitable campaign. "If you're operating at a 0.6x
    return on ad spend, and you need to be a 3x… that's not the fix."
  - Leaving Learning Limited gives maybe a 10–20% lift, and only when results are already
    decent.
- **Change cadence.**
  - "I'm not going to make any changes to this ad set more than once every 7 days." Add new
    creatives in batches, e.g. five at the end of the week [dlVx7HgKL_I].
  - Small budgets: set an optimisation schedule of once every 7, 10 or 14 days
    [XLagiyzYYpE].
  - New accounts: "give it 5 days… 7 days… 10 days, then I'm going to assess."
    [vaBxYgZ7MAU]
- **Patience is measured in conversions.**
  > "A campaign that has generated 10 conversions in 20 minutes is just as old as a campaign
  > that has generated 10 conversions in 3 months." [3ReT1bzLFhU 2026-10]
  - At three conversions a week, leave the campaign alone "for best part of a month".
  - At thousands a day, you might know in 3 hours [XLagiyzYYpE].
- **Don't "helicopter parent" campaigns.** Between scheduled changes you can still analyse,
  produce creative and plan [XLagiyzYYpE 2026-02].

### Testing (see creative.md for creative-specific tests)
- **What to test, in priority order** [fV6P9MF7FqU 2025-04; 8dJxOaZrK1M 2026-03]:
  1. Product
  2. Offer
  3. Angle
  4. Format
  5. Style
  6. Hook
  7. Headline
  8. Primary text
  9. CTA button
  10. Description
- **Test design.**
  - Isolate one variable and start from a hypothesis.
  - To test a style, use a batch of ads (one UGC ad failing doesn't prove UGC fails).
  - Small variables "might be a 10% difference at most". Big swings "could get you a
    doubling".
- **Mechanisms by volume** [v_u9qcbW5hY 2026-01; 9mUwsvSK0NY 2026-10; onFwSud9C2Y 2026-02]:
  1. **Do nothing.** For one-off or time-limited sales (BFCM), huge slow-fatiguing
     audiences, or very small budgets.
  2. **Add clearly different ads to the same ad set.** This is Meta's recommendation and his
     choice below about 50 conversions a week. Background-colour tweaks won't get spend.
  3. **Push delivery to this ad** (ad level, new in 2026-10). Defaults to 20% of budget for
     7 days. It is his preferred method for low-volume accounts (1–2 conversions a day): one
     ad at a time, no fragmentation, and it protects the winners.
  4. **Creative Testing tool** (ad level).
     - Tests 2–7 ads (it was 2–5 until 2026). You choose the budget share and duration
       (default 7 days).
     - Switch the comparison metric from cost per post engagement to CPA, CPL or ROAS.
     - It splits the audience, so you get clean results with no auction overlap.
     - For accounts with dozens to hundreds of conversions a week.
     - Push delivery and the testing tool can't run at the same time.
  5. **Separate testing campaign** with ad-set budgets to force spend.
     - Only above 50 optimised conversions a week ("100, 200 a week or more" is fine).
     - Winners graduate to the scaling or evergreen campaign.
     - At small budgets: "I really don't like that." [du6rDgRf06g 2026-08]
- **Testing budget.**
  - "We normally sit somewhere in the 10 to 50, maybe even 60% range." 100% if nothing
    converts; 10–20% if the account is very profitable [9mUwsvSK0NY 2026-10].
  - The agency itself puts 30–40% into testing new ads.
  - Other splits he gives: 80/20 scaling/testing [crId74rHjuA 2025-10], and "20/80 or maybe
    50/50" [3ReT1bzLFhU 2026-10].
- **How long to test.**
  - "About 7 to 10 days" by default; adjust by conversion volume, from hours up to a month
    or more [fV6P9MF7FqU 2025-04].
  - Use a statistical-significance calculator and act at **85–90% confidence**.
  - Three conversions is not enough; rerun longer.
  - Kill early only when far off target. With a $30 target CPA: $35 means wait; $60
    ("double where you need to be") means kill early.
  - "If in doubt, be patient." A winner can run for years.
- **Low volume.** "How do I assess performance earlier? You don't. You can't." Judge on the
  north-star metric [9mUwsvSK0NY 2026-10].
- **Short-term vs long-term trade.**
  - "I'm willing to take a 10, 20% hit in the short run in order to potentially find an ad
    that performs 50% better… spend three times as much." [v_u9qcbW5hY 2026-01]
  - "The best short-term results almost certainly come from doing nothing."
    [3ReT1bzLFhU 2026-10]
  - Think like a venture capitalist: most tests fail, and one winner pays for all of them
    [8dJxOaZrK1M 2026-03].
- **Don't "split test your way to the perfect ad."** After Andromeda, an ad with slightly
  lower average ROAS may bring incremental conversions from other people. Keep diverse ads
  running [9mUwsvSK0NY 2026-10].
- **Never turn off the best ad to force spend onto new ads.** "Performance tanks."
- **Turn off ads Meta isn't spending on.** If Meta keeps spending on a lower-ROAS ad, leave
  it on [kuSq-pmNfnM 2025-09].

---

## 8. Budgets

### Small and tiny budgets [XLagiyzYYpE 2026-02; du6rDgRf06g 2026-08]
- **Small** means under $3,000 a month (about $100 a day). **Tiny** means $600 a month or
  less ($20 a day or less).
- **On a small budget:**
  - One offer only, your best seller or most profitable, with one campaign and one ad set.
  - No awareness, engagement or traffic campaigns: direct Leads or Sales only. The exception
    is omnipresent content, and even that "less keen" for small businesses.
  - Go niche: target a hyper-specific ideal customer with tailored messaging. "Most
    businesses… are far too broad as opposed to too niche."
  - Copy what already works: in the Ad Library, competitor ads running 6+ months are almost
    certainly profitable.
  - Use organic Instagram content as free creative testing. Take the reels that do well, add
    a 5–10 second CTA and run them as ads.
  - The goal is **proof of concept**, then profitability to fund reinvestment.
  - Be willing to pay more per conversion than you'd like.
- **$100 a month** [du6rDgRf06g 2026-08]:
  - One campaign, one ad set, one offer, broad / Advantage+, no hard constraints, no separate
    testing campaign.
  - "One meaningful test per month." Keep bidding simple (most conversions).
  - It "might literally take 3 months" to know. Spend "an hour or two probably a month" in
    the account.
- **$100k a month:** multiple campaigns, testing campaigns at 10–30% of budget, value
  optimisation, CRM integration, value rules, a "creative factory", incrementality testing,
  about 5% of budget on expertise, and partnership ads at about 10% of budget.
- **Starting budget:** the "sweet spot" you could afford to lose "but it… would sting a
  little bit" [crId74rHjuA 2025-10].
- **Seasonality:** spend more in good periods and less in bad ones; don't fight it. Compare
  results year on year for seasonal businesses [846INGFormA 2024-11]. Example: sunglasses
  sales fell during five rainy days in the UK; ads create an opportunity, not a guarantee
  [Robert Plank podcast 2025-05].

---

## 9. Scaling [FyWyHJh_6Ng 2026-05; nL0L-9dXM9k 2024-11; 1Or1B4L_bdY 2025-12; du6rDgRf06g 2026-08]

- **Choose the method by personality.**
  - If you can't help tinkering, or you react emotionally to daily results: Option 1.
  - If you're level-headed: either option.
- **Option 1: automated rules.**
  - **Increase rule ("Scaling 3%"):**
    - Increase the daily budget by **3%**, as a percentage rather than a fixed amount,
      **once daily**, scheduled daily between midnight and 1am.
    - Condition: cost per result below target (e.g. under £50), or ROAS above target (e.g.
      over 4) for varying order values.
    - Keep Meta's auto-added "lifetime impressions > 8,000".
    - Optional maximum daily budget cap (e.g. £500).
  - **Time range** (never the 37-month maximum):
    - Last **3 days** with high volume. This was his default in 2024.
    - Last **7 days** as the middle of the road. This is his default in 2025–26.
    - **14–28 days** with low volume.
  - **Decrease rule ("Budget decrease 3%"):**
    - Decrease 3% once daily when CPR is above a higher threshold (e.g. £75), with a minimum
      floor (e.g. £20/day).
    - Leave a buffer band (£50–75) where the budget holds.
  - **Limits:** you can use 5–7% instead of 3%, but "I wouldn't go… much above that if
    you're going to be scaling daily". "Don't put that at 30% once daily."
  - **Track record:** he scaled from $100 a day to over $1,400 a day profitably with this
    method [nL0L-9dXM9k].
- **Option 2: manual stepped scaling.**
  - Larger steps, less often. Wait 5–10 days per step, never more often than every 5 days,
    because a budget step re-enters learning for 1–2 days and then needs a few days to judge.
  - Make the percentage smaller as spend rises. His ladder:
    $50→100 (+100%) →150 (+50%) →200→300→400→500→650→800→1,000 (+20%).
  - Don't double at $800 a day.
  - Don't panic over a 1–2 day wobble; judge on the roughly 7-day average after the step.
  - Earlier manual guidance: "about 20% once a week", never daily [nL0L-9dXM9k 2024-11].
  - Newest nuance: "I don't like to give fixed numbers or fixed percentages."
    - $3→$10 a day ("more than a tripling") is fine at tiny budgets.
    - At $100k a month, he wouldn't even go to $120k in one go [du6rDgRf06g 2026-08].
- **Never duplicate campaigns to scale horizontally.** It creates auction overlap, slower
  learning and "11 identical campaigns" chaos.
- **Never make a massive single jump** (10x, or $10 to $500 a day).
- **Scale existing campaigns** rather than creating new ones [kuSq-pmNfnM 2025-09].
- **Find the scaling ceiling.** Example: at $50 → 6x, $100 → 5.2, $150 → 4.9, $200 → 4.6,
  $300 → 4.3, $400 → 4.1. If you need 4x, the ceiling is about $400 a day. To go higher,
  improve the ads and offer.
- **ROAS falls as you scale.** Meta finds the hyper-responsive buyers first.
  - If results collapse at the first step, your ads only convert warm or hyper-responsive
    people. "The solution is better ads and better offers."
  - Scale creative along with budget. One ad won't survive a 10x budget increase
    [kuSq-pmNfnM].

---

## 10. Metrics, KPIs and analysis [mycqb92wJqk 2024-09; 846INGFormA 2024-11; vaBxYgZ7MAU 2026-05]

- **"Only two Facebook ad metrics matter": ROAS and cost per conversion.** Pick one as the
  north star.
  - Never "turn off an ad that has a good return on ad spend because it has a low
    click-through rate."
  - Never optimise for CPM, CPC, CTR or hook rate.
- **His column preset, in order:**
  1. Name, delivery, Actions (opportunity score), budget, amount spent
  2. Results, cost per result, conversion value, ROAS
  3. Reach, impressions, frequency
  4. CPM
  5. Hook rate (custom metric: 3-second video plays ÷ impressions), average video play time
  6. Link clicks (not all clicks), landing page views
  7. CPC (link), CTR (link)
- **Secondary benchmarks:**
  - **Frequency:** "To cold audiences I don't like to see a frequency above 2.5." Warm: under
    roughly 6–8 for most. His own warm audience runs 10–16.
    - Rising frequency plus falling results means fatigue: add new creative.
    - Example automated-rule ideas: frequency > 3 lowers the budget; > 7 turns the ad off
      [1Or1B4L_bdY 2025-12].
  - **Hook rate:** "Above a 10 is really quite good, anything below a five is quite poor."
    B2B runs lower.
  - **Link CTR:** "Anything above 1% is considered really quite good." Over 2% means very
    well. Below about 0.65% is poor **[caption?** the caption reads "8.86" and "7.65"; almost
    certainly 0.86% and 0.65%**]**.
  - **Link clicks to landing page views:** a gap of about 20% is fine; over 50% means a slow
    page or accidental clicks.
  - **E-commerce click-to-purchase rate:** under 2% is low.
  - **CPM:** secondary. A high CPM can mean a higher-quality audience. B2B costs more than
    health and fitness.
- **Diagnostic pairs:**
  - High hook rate + worse CPA: combine that hook with the better-converting body.
  - High hook rate + low watch time: the hook is misleading or the body is weak.
  - Hook rate > 10% + link CTR < 0.5%: the offer is weak or there isn't enough proof.
  - Good CTR + poor conversion: a landing-page problem (congruence, surprise shipping fees,
    no reviews). Run it through "a common sense filter" in case the ad is engagement bait.
  - Image ads often get a higher CTR but can convert worse than video.
- **Breakdowns:**
  - **Audience segments:** new, engaged, existing.
  - **Placement and platform.**
  - **Charts:** demographics, platform, and the video retention curve (cut where viewers drop
    off, do more of what holds them).
  - On a demographic skew (e.g. 83% male), change the **messaging, not the targeting**.
- **Five-step diagnostic** [vaBxYgZ7MAU 2026-05]:

  | What you see | Problem |
  |---|---|
  | No conversions | Offer |
  | Some conversions, low hook rate and CTR | Creative (the most common case) |
  | Some conversions, high CTR and hook rate | Landing page / post-click |
  | Many conversions, unprofitable | Business model |
  | Many conversions, profitable | Scale |

- **Time horizon:** compare month on month (e.g. "a 15% improvement in cost per purchase"),
  or year on year for seasonal businesses. Never day to day [846INGFormA 2024-11].

---

## 11. Lead generation specifics

- **Instant forms vs website** [Cz8Bi6_rkwM 2025-05; 0Ji-AhLkR8w 2025-08]:
  - Forms are easier, cheaper (CPL "often significantly less") and higher volume, with
    cheaper CPMs. The trade-off is lower quality and contact rates.
  - You can now test **"Website or instant form"** in one ad set and let Meta choose per
    user. "Definitely something to test."
- **Lead quality settings.** These became his default in 2026 [S-F3yxya2WE 2026-09;
  oTWfDI_91Fs 2026-09]:
  - **Form type: Higher intent**, with **"require leads to verify their phone number with a
    one-time passcode"** (SMS OTP). "By default, I'm now using this."
  - **Conditional logic.** Add a multiple-choice qualifier (budget, timeline, location,
    industry, life stage). Unqualified answers go to a "close the form / end page for
    non-leads", so they never count as leads. Meta then hunts only for qualified people.
  - **Autofill off** for email and phone. Prefilled details can be 15+ years old. It is less
    drastic than SMS.
  - **Phone number mandatory**, except for lead magnets.
  - Keep forms short. Leave flexible form delivery on, except in regulated industries.
  - **Website leads:** use qualifying questions, then route qualified people to a thank-you
    page that fires a distinct event (e.g. "Submit Application"). Unqualified people go to a
    page with no lead event. Optimise for the qualified event and measure cost per qualified
    lead.
  - Impact is "minimal" on day 1 and "absolutely monstrous" after 30–90 days
    [oTWfDI_91Fs].
  - **His old advice** was "Go with more volume. Don't have any conditional logic. Reduce
    friction…". It is now reversed [S-F3yxya2WE].
- **Cost trade-off.** "If… more than half the leads you generate, you would classify as low
  quality… maybe it's worth paying 50% or 80% more per lead, maybe even more."
  [Cz8Bi6_rkwM 2025-05]
  - Be careful with older, less tech-savvy audiences and countries where SMS codes are
    uncommon.
- **CRM integration.** Optimising for **conversion leads** or qualified leads is "the much
  better way" [qkXMTQfLgvE 2026-09].
  - Most advertisers find a CRM import "tricky"; value rules are the easier proxy for
    lead-to-client rate [bv8AsY0xkIk 2026-05].
- **Speed to lead.** "If you can call a lead within a few minutes of becoming a lead, you
  are far more likely to make contact."
  - Use ad scheduling for B2B leads (e.g. 9–5, Mon–Fri). It needs a lifetime budget and can
    "double or more your return on ad spend" for some lead businesses
    [1Or1B4L_bdY 2025-12].
- **Follow-up features** [0Ji-AhLkR8w 2025-08]:
  - Messenger and Instagram follow-up chat after a form ("potentially a really, really big
    deal"; needs staff or an AI chatbot).
  - **WhatsApp chat** as an additional action, for markets where "the whole country runs on
    WhatsApp".
  - Lead-email delivery and promo-code redemption.
- **"Some leads are good some leads are bad, get over it."** You can't remove every bad lead
  [846INGFormA 2024-11].
- **Lead quality is a value-rule use case.** If lead-to-client rates differ by age,
  location or device, bid accordingly.

---

## 12. Omnipresent content strategy (high-ticket and expertise businesses only) [hHuxr_ancXQ 2025-11; oEPequOQTHI 2024-06]

- **Who it's for:** customer value of at least about **$1,000** over the lifetime.
  - Expertise sellers (coaches, consultants, agencies, professional services, courses),
    bespoke or high-end products, and purchases that depend on timing.
  - Not for $20 impulse e-commerce. "Probably more than half of all businesses don't meet
    that criteria."
- **Campaign:**
  - Objective: Awareness for broad markets, or Engagement for niche or small local markets
    (Engagement now supports a frequency cap). Buying type: Auction.
  - **Advantage campaign budget OFF** ("if you turn that on, it's going to mess up
    everything").
- **Ad set:**
  - Goal: **Maximize reach**, with a frequency **cap of 1 impression per 7 days**.
  - No value rules, dynamic creative, scheduling or end date.
  - The same daily budget in each ad set: "as little as a pound or a dollar per day per ad
    set". His example: £5 × 14 = £70 a day.
- **Targeting:**
  - Advantage+ audience with all custom audiences as suggestions. Location is where
    prospects are.
  - No detailed targeting, no exclusions, no lookalikes in the warm set.
  - Don't mix countries with very different CPMs.
  - **Older version (2024):** warm only if 20–30k+ people; otherwise cold kept under
    250,000 using interest AND interest (narrow audience).
- **Placements:** manual. Remove Audience Network; keep feeds, stories, reels, in-stream and
  search.
- **Structure:** **14 identical ad sets, one ad each**. The frequency cap applies per ad
  set, so each person sees about 2 different ads a day.
  - Rule: cap days = number of ads ÷ 2 (10 ads → 1 per 5 days).
- **Creative mix (example):** 4 value, 4 demonstration (process, what's included, how it's
  made), 4 testimonials, 2 CTA (book a call, limited slots). Mostly video.
  - Refresh every 3–6 months.
  - Stagger launches: value ad sets first, CTA ones later.
- **Run it alongside** a direct conversion campaign.
- **Measure qualitatively:** positive comments, prospects saying "I see you everywhere", the
  conversion campaign improving, follower growth.
- It compounds over 6–24 months. **Ignore Meta's fragmentation warnings and Meta reps**
  here; Andromeda consolidation advice applies to direct-to-offer campaigns only.

---

## 13. Offer and LTV [crId74rHjuA 2025-10; 8dJxOaZrK1M 2026-03]

- **Hormozi's value equation:** (Dream Outcome × Perceived Likelihood of Achievement) ÷
  (Time Delay × Effort & Sacrifice). Raise the top, lower the bottom.
- **Levers:**
  - **Differentiation:** a better outcome, more likely to work, faster, or less effort.
  - **Guarantees** ("massively underutilized"): results or money back, keep working for
    free, durability, delivery time (Domino's 30 minutes, 72 hours), speed of outcome.
  - **Scarcity:** limited slots or quantity.
  - **Urgency:** a deadline discount, BOGO, Black Friday.
  - **Proof:** reviews, Trustpilot, accreditations, big-name clients.
  - **Finance options:** borrowed from adjacent industries [fV6P9MF7FqU].
- **Book:** Alex Hormozi's *$100M Offers*, "a foundational educational resource"
  [8dJxOaZrK1M]. The caption mis-hears it as "100 Billion Dollar Offers".
- **"What's truly special about you?"** Without a differentiator you're asking for an
  "illogical decision". Copying a competitor's ads fails if they have a differentiator you
  don't [3ReT1bzLFhU 2026-10].
- **Do the hard things.** Test price points over about 6 months (example: a "30% more
  profitable" price), rework the offer, and recruit creators (contact 20, two reply).
- **Tailor offers and landing pages per avatar.** HelloFresh and Huel examples: one offer and
  one landing page for everyone leaves results on the table [XtoJVbu4Bys 2026-07].

---

## 14. Settings to check [1Or1B4L_bdY 2025-12; pxPx6YGA1wk 2026-08; S-F3yxya2WE 2026-09]

- **ON:**
  - Combine social proof ("allow Meta to combine likes and reactions for similar ads").
  - Standard enhancements and "optimize text per person", most of the time. Not in heavily
    regulated industries.
  - CAPI with Automatic Advanced Matching.
  - Maximize value of conversions for sales.
- **Check:**
  - **Creative enhancements** (AI text overlays, image variations, music, cropping). Meta
    re-enables them whenever you duplicate, edit, or even preview an ad. "If you touch it in
    any way, they can be turned on."
    - His team turns off off-brand overlays and AI image variations: a small ROAS bump
      isn't worth brand damage.
    - Compliance-heavy industries face regulatory risk.
  - **Personalized destinations** (Meta sends clicks to whichever site page it predicts
    converts). "It's one I'd turn off", except possibly pure e-commerce product pages, and
    even then he is unsure.
  - **Ad scheduling** for B2B lead gen (needs a lifetime budget).
  - "Allow limited spend to excluded placements". Untick it when restricting placements.

---

## 15. Mindset and common mistakes (consolidated)

**Mindset:**
- Patience is measured in conversions.
- Control your emotions; failed tests "are not a waste of money".
- Focus on high-leverage work: partnership ads ("as close to a silver bullet as you can
  get"), value rules and optimisation events ("might triple your return on ad spend"),
  rather than "six or eight ads in an ad set" [3ReT1bzLFhU 2026-10].
- Apply advice to your own business model (e-commerce vs lead gen).
- Stay current: best practices "age out".
- Meta reps' advice is generally poor: "awful and should mostly be ignored"
  [qnJZ1mmxGps 2026-09].

**Mistakes he lists repeatedly:**
- Blaming targeting.
- Too many campaigns or ad sets.
- Splitting cold and warm.
- One interest per ad set.
- Duplicating campaigns to scale.
- 10x budget jumps.
- Constant tinkering.
- Killing tests too early.
- Testing only small variables.
- Turning off the winner, or turning off top-of-funnel ads that Meta still spends on.
- Engagement or traffic campaigns to "warm the pixel".
- Optimising for clicks or reach inside Sales.
- Boosting posts.
- Wrong performance goal.
- Lifetime budgets by default.
- Unrealistic CPA or ROAS expectations.
- Weak offer.
- Feature-stuffed, multi-angle ads.
- Too broad an ideal customer.
- Hashtags in ads.
- No clear CTA.
- Generic landing page.
- Ad and landing page mismatch.
- Repurposed TV or billboard creative.
- Leaving creative enhancements or personalized destinations on unchecked.
- DIY tracking.
- Pixel without CAPI.
- Pixel firing lead or purchase events on page load.

---

## 16. What changed: timeline of his advice

| Topic | Before | Now (newest source) |
|---|---|---|
| Cold vs warm | Separate cold and warm campaigns, one ad set each (2024-05) | Combined in one ad set; split only for distinct messaging (2026-02, 2026-03) |
| Audience testing | One interest per ad set, tested against each other | Never; only location is tested (2025-06, 2026-02) |
| Ads per ad set | ≤ 6 (Meta), he used ≤ 5; 1–3 at $10–20/day (2024-05) | 20+, no upper limit; quality is the limit (2025-11, 2026-02, 2026-09) |
| Copy tests | Separate ads per headline | Text variations inside one ad (2026-02) |
| Funnel | TOF / MOF / BOF campaigns | Funnel-stage ads inside one ad set; Meta sequences them (2026-02) |
| Testing method | Separate testing campaign with ABO (2025-04) | Creative Testing tool (2–7 ads) or push delivery; testing campaign only at 50+ conversions/week (2026-01, 2026-10) |
| Format | Video wins ~60% (poll 2025-02) | Statics win 54% vs video 38% (poll 2026-10); run both |
| UGC | "UGC style video is king" (2025-09) | "All UGC is not diversification"; diversify styles (2026-08) |
| Lead forms | More volume, no conditional logic, low friction | Higher intent + SMS OTP + conditional logic + qualified-only lead event (2026-09) |
| Value vs count | Count fine by default (2025) | Maximize value for most sales accounts (2026-09) |
| Gender targeting | Hard gender via "further limit reach" (2025-06) | Removed; value rule −90% (2026-11) |
| Placements | Manual ad-set placement selection for non-conversion goals | Ad-set controls removed (2026-09); account-level controls or placement value rules |
| Automated rule window | Last 3 days (2024-11) | Last 7 days default (2025-12, 2026-05) |
| Scaling % | 3%/day automated or 20%/week manual (2024-11) | 3%/day automated, or stepped manual with bigger steps at low spend and smaller at high; "no fixed percentages" (2026-05, 2026-08) |
| Lookalikes | A "game changer" | "Much less valuable"; occasional use for lead gen (2025-06) |
| Omnipresent targeting | Warm only, or narrow cold < 250k with interests (2024-06) | Advantage+ suggestions with all custom audiences, no detailed targeting (2025-11) |
