Meta ads guide
How a creator membership with a free trial runs Meta ads on Facebook and Instagram in late 2026. Written for the first Sah Method campaigns on the 10-day trial.
What changed since you last ran Facebook ads
Creative is the targeting
Meta's Andromeda system picks a few thousand ads per person from tens of millions. It reads the ad to find the audience.
Most targeting is a suggestion
Only location, minimum age, custom audience exclusions and language hold. Interests, lookalikes, age and gender are hints.
One campaign flow
Sales campaigns open with Advantage+ on for budget, audience and placements. The existing-customer budget cap is gone.
Clicks mean link clicks
Since March 2026, likes, saves and shares count as engage-through. The 7-day and 28-day view windows are gone.
Placement exclusions are leaving
Since late August 2026, ad-set exclusions disappear in waves. Account-level controls and placement bid rules stay.
Wellness sites can lose Purchase
Meta puts websites into data categories. Health and wellness, or Religion, can block the Purchase event.
The four decisions
1. Advantage+ or manual
Pick: Advantage+
Sales objective, website conversion, campaign budget, Advantage+ audience and placements, one ad set. Set only the four controls and the exclusions.
A small account has few signals, and one ad set pools them. Against it: Wicked Reports saw manual win on new-customer cost in ecommerce in 2025.
Flip when there is no conversion event, a hard audience rule applies, or you A/B test two structures.
2. Broad or narrow
Pick: broad
US as the location control. No interests, no lookalike as a limit. Warm custom audiences as suggestions only.
Suggestions do not bind, and the creative finds the people. Jon Loomer touches only locations and exclusions in 2026.
Flip when the goal is a click, a message needs a warm reader, or a demographic spends with zero trials after 1,000 impressions.
3. Engagement audiences
Pick: build them, use three ways
Instagram engagers and message senders at 365 days, reel viewers, Page engagers, customer lists. Use them as suggestions. Exclude members and active trials. Add a small warm ad set only for a different message.
Meta already sends 20 to 40 percent to warm people by itself. Jon Loomer measured 35 percent. Ben Heath showed almost 40 percent of $80,000.
4. The event to optimize for
Pick: the trial start
An ad set needs about 50 results a week to leave learning. The first charge comes on day 10, after the 7-day click window. So Meta learns from the trial start.
Kajabi sends no trial-start event, only Lead and Purchase. Purchase fires on the first page after checkout, also for a $0 trial. Kajabi documents no value for it. Two support replies say this, likely written by an AI agent, so test it.
Flip when first payments reach about 50 a week, or the site's category blocks Purchase.
Budget
Restart after a two-month pause
Build new campaigns. Meta resets learning on any ad set paused for 7 days or longer, so an unpause saves nothing. Post IDs, audiences, customer lists and dataset history carry over.
The formulas
Cap per paid member = lifetime value ÷ 3
Stop line per trial = cap per paid × trial-to-paid rate
Start budget per day = 50 × stop line ÷ 7 ÷ 2
Weekly growth = (1 + raise per step)^(7 ÷ days between steps) − 1
The stop line is read in Kajabi. It is not a bid goal. Bid on highest volume.
How fast to raise
Meta gives no safe percentage. It only says a change counts by size. Each source's rule, per week:
| Source | Rule | Per week |
|---|---|---|
| Ben Heath, YouTube | 3% a day by rule, 7% at most | 23 to 61% |
| coreppc.com | 15 to 20% every 3 to 4 days | 28 to 53% |
| Chase Chappell, YouTube | 20% every 3 days | 53% |
| Reddit, most common | 20% every 48 hours | 89% |
| Nick Theriot, YouTube | 20% each day the target holds | up to 258% |
| This guide | 20%, twice a week | 44% |
Raise at or under the raise line. Hold between the raise line and the stop line. Stop above the stop line after day 7. Above about $150 a day, duplicate the campaign instead.
Free trials
Phase 1: the trial start
Optimize on the Purchase that Kajabi fires after the trial checkout while volume is small. Judge on cost per paid member, not cost per trial.
Cheap trials often bring people who never pay. Practitioners below about $1,000 a day still start here.
Phase 2: a qualified trial
Send the trial to Meta only if it is still active some hours later. RevenueCat waits 2 hours and drops at least 25 percent of trials.
A 30-second cancel looks like a long-term member to Meta. For Sah this needs a small relay outside Kajabi.
The first charge
Send the day-10 payment to Meta as Subscribe from the server. Kajabi's Payment Succeeded webhook is the likely trigger. It carries name and email. Kajabi documents no Meta event at that charge and no field that marks the first charge after a trial.
Day 10 is outside Meta's 7-day click window. It feeds reports and audiences, not delivery.
Not yet: predicted value
Meta's predicted-lifetime-value bidding needs 100 attributed conversions a week for 4 weeks. Sah is far below that.
It also needs a site that Meta has not limited as health or wellness.
| Trial-to-paid benchmark | Source | Rate |
|---|---|---|
| Card-required trials | Kyle Poyar, Growth Unhinged, Jul 2026 | 30% |
| Health and fitness apps, median | RevenueCat, 2025 | 39.9% |
| Trials of 17 to 32 days | RevenueCat, 2025 | 45.7% |
| Sah, measured | 2 of 13, mostly invited testers | 15% |
Sah's 15 percent is half the outside benchmarks. The morning check reads when trials cancel, hour by hour.
Sah worked example
Lifetime value $150 to $200, so the cap per paid member is $50 to $67. Trial-to-paid: 15 percent measured, from mostly invited testers (2 of 13). The call planned 30 to 50 percent.
| Trial-to-paid row | Stop line | Raise line | Full learning budget |
|---|---|---|---|
| 15%, measured, weak | $10.05 | $7.50 | $72 a day |
| 30%, plan | $20.10 | $15.00 | $144 a day |
| 50%, plan | $33.50 | $25.00 | $239 a day |
The 15 percent row governs week 1. Start: 50 × $10.05 ÷ 7 ÷ 2 = $36 a day.
| Days | Daily budget | Spend |
|---|---|---|
| 1 to 7 | $36 | $252 |
| 8 to 14 | $43, then $52 | $337 |
| 15 to 21 | $62, then $75 | $486 |
| 22 to 28 | $90, then $108 | $702 |
Four weeks cost $1,777 if every gate passes. When the launch-week cohort charges (3 to 7 October), switch to its row.
A cold trial through the home page costs about $12 ($0.60 click ÷ 5% visit-to-trial). That fails the 15 percent stop line. A warm visitor sent to the checkout costs about $2.88. At 15 percent, only warm traffic to checkout pays back. This is an estimate.
Creative and set-up
- 3 to 6 concepts that differ in angle, one angle per ad, 3 to 5 assets inside each ad. Start with 3. Three edits of one video count as one concept.
- 9:16 video, sound on, captions burned in, key elements inside the safe zone.
- Sell the state the person is in. No health claims. The same words on the website feed Meta's data category.
- Turn off every creative enhancement Sah has not approved. Check again after each duplicate, edit and preview.
- Attribution for a trial: 1-day click, view-through off.
- Warm people to the checkout, cold people to the sales page. A UTM on every link. Fix the ManyChat buttons first.
- Judge nothing before day 7. Measure trials and paid members in Kajabi.
- No AI agent gets write access to the ad account.
What the morning analysis checks first
- The data category of a new dataset on the Sah sites, in Events Manager.
- The Purchase value on a test trial checkout. Kajabi documents none.
- Whether Payment Succeeded fires on the day-10 charge. Kajabi's docs do not say. Kajabi did not answer whether the webhook carries fbp and fbc.
- The hour of each cancel inside the trial. It sets the qualified-trial wait.
- The pause date, objective and event of each old campaign.
- The size of each warm audience at 365 days.
- The launch-week trial-to-paid rate. It picks the budget row.
What practitioners say that Meta does not
YouTube, 104 talks
- Meta dropped its old "no more than 6 ads" advice with Andromeda. Jon Loomer starts with a couple of ads.
- Small budgets run sales or leads only and change things every 7 to 10 days (Ben Heath).
- Interests, lookalikes and the prospecting and retargeting split matter much less now (Dara Denney).
- No talk covered free trials or restarting a paused account.
Reddit, 167 threads
- At $100 to $200 a day, 40 creatives dilute the signal.
- Traffic campaigns run "to build data" teach the pixel to find clickers.
- Most buyers past a ceiling duplicate campaigns of $100 to $300 a day.
- The subreddit is loud with complaints since October 2025. Weight it low.