Resonance means a defined audience found a post useful enough to take the action you predicted. It is not a large view count or a national conclusion.

In this lesson, you will turn one narrow audience belief into a seven-day Instagram test: three comparable content variants, one native metric set, and a learning log that tells you what to test next.

// concept

Write One Audience Hypothesis You Can Disprove

Start with a segment, not a country. Food audience is too broad. Karachi-based working women aged 25–35 who want packable weekday lunches states a location, life situation, and need. Age is a test boundary, not a claim about everyone in it.

Use this hypothesis formula:

// prompt — copy me2 lines

For [specific audience], a [consistent format] that promises [specific utility]
will produce [expected native signal], because [evidence or observation].

Example:

// prompt — copy me4 lines

For Karachi-based working women aged 25–35 who meal-prep,
a 30-second captioned Reel showing one heat-safe lunch method
will earn saves relative to accounts reached,
because three recent first-party questions asked how to pack lunch for long commutes.

Those questions justify testing, not demand. Log their source and remove personal details.

// concept

Hold the Promise Still and Vary the Opening

Create three Reels around the same lunch method. Keep their format, length, cover style, language, CTA, posting window, and 24-hour review window comparable. Change only the opening:

Variant	Opening frame	First line	What it tests
A	Direct utility	`Pack one office lunch that stays practical on a long commute.`	Does a plain benefit earn saves?
B	Common mistake	`Your lunch may be getting soggy before noon for one fixable reason.`	Does problem recognition earn saves?
C	Constraint-led	`No microwave at work? Build this lunch around one insulated container.`	Does a familiar constraint earn saves?

Keep the method, quality, and promise fixed. Publish on days 2, 4, and 6 in comparable windows; read each after 24 hours.

// concept

Use One Native Metric Set and a Learning Log

Use Instagram's post-level Accounts reached and Saves. Insights require a professional account. Open the post and tap View insights; check official help if the menu has moved.

Normalize for unequal exposure:

// prompt — copy me1 line

save rate = saves ÷ accounts reached × 100

Do not switch metrics between variants. Comments may suggest another test but do not replace save rate.

// template — copy me2 lines

| Variant | Post URL | Published | Read after | Reach | Saves | Save rate | Confound | Learning |
|---|---|---|---|---:|---:|---:|---|---|

Log a repost, outage, cricket match, or delayed upload as a confound. Rerun a heavily confounded variant.

// worked_example

Worked Example

This hypothetical example is not a real result. A Karachi meal-prep creator prompts an AI assistant:

// prompt — copy me8 lines

Audience: Karachi-based working women aged 25–35 who meal-prep.
Hypothesis: a captioned Reel about one heat-safe packed-lunch method may earn saves.

Create three 30-second Reel openings for the SAME method and promise.
A = direct utility, B = common mistake, C = no-microwave constraint.
Keep the CTA exactly: "Save this for your next office-lunch prep."
Use clear English; do not invent nutrition or safety facts.
Return a table with opening line, first shot, on-screen text, and voiceover.

The returned outline is:

// prompt — copy me3 lines

A: packed container | "Pack one office lunch..." | direct benefit
B: condensation on lid | "Getting soggy before noon?" | common mistake
C: office desk | "No microwave at work?" | practical constraint

Draft one wrongly changes B to a different recipe. The creator fixes B so every variant teaches the same method and uses the same CTA.

This is a fabricated sample log for calculation practice:

Variant	Accounts reached	Saves	Save rate	Confound	Learning
A	240	12	5.0%	None logged	Baseline.
B	310	10	3.2%	Reposted	Rerun before comparing.
C	200	14	7.0%	None logged	Retest this frame.

Bad conclusion: Karachi women prefer no-microwave videos, so this is my niche. One account proves little and B was confounded. Fix: C led the two clean variants; repeat A versus C with another lunch method.

// failure_cases

Failure Cases to Diagnose

6 cases to diagnose

The segment is `Pakistanis aged 18–34`.
Add region, life situation, need, and platform context.
Variants teach different ideas.
Rebuild them around one method and promise; vary only the opening.
Review windows differ.
Capture every variant after the same 24 hours.
The denominator is missing.
Compare saves relative to accounts reached, not raw saves.
A repost is treated as clean.
Mark it as a confound and rerun the variant.
The conclusion becomes national.
Limit it to this account, segment hypothesis, topic, week, and metric.

// pakistan_angle

Pakistan Angle

Karachi, Lahore, Peshawar, Quetta, and smaller cities do not share one language mix or commute pattern. A Roman-Urdu opening for a Karachi audience may sound forced to a Pashto-first audience in Khyber Pakhtunkhwa. Keep one defined audience, use a fluent language reviewer, and never use an accent as a joke.

For Pakistani connectivity, draft offline, compress exports, add burned-in captions, and upload when power and mobile data are stable. Log load-shedding or a delayed upload. Do not expose a WhatsApp number, customer message, CNIC, workplace badge, vehicle plate, or family member to make a Reel feel local.

// hands_on

Hands-On Exercise

7 steps

Build the seven-day resonance test now:

Write one segment statement with region, life situation, need, and Instagram context.
Save one first-party observation; redact personal data.
Complete the hypothesis with one format, promise, and save-rate signal.
Draft A, B, and C around the same method, changing only the opening frame.
Schedule days 2, 4, and 6; set the same 24-hour review reminder.
Copy each post's reach and saves into the log, calculate save rate, and record confounds.
On day 7, write one narrow conclusion and follow-up test. Done means you have a completed plan, three variant drafts, and a result table ready for real entries or clearly labelled practice data.

// completion_rubric

Completion Rubric

6 checks — tick as you verify

0/6

// sources

Sources

3 official sources — check every claim yourself

// check_yourself

Check yourself

4 questions · answers and options are taken word-for-word from this course

0/4

1 / 4 · diagnose
Your work shows this failure mode: “The segment is `Pakistanis aged 18–34`.” What does the lesson tell you to do about it?