Parallax X Content Lab

Fictional demonstration from an actual local trial. This is not a real customer result. Local links and machine-specific paths are displayed as text.

Saved three complete single posts with compact alternatives, one connected five-entry thread, and one reply draft. The first four main items cover two weeks at your two-per-week capacity; Weeks 3–4 are clearly labeled conditional outlines. No posts or schedules were submitted.

The analytics script removed one exact duplicate and retained P104’s latest complete snapshot, leaving seven posts. Four comparable seven-day originals have 102 recorded link clicks / 3,200 impressions = 3.1875%. Destinations and inquiries are unknown, so this is not a sales or audit-conversion rate. The 20,000 quoted-parent impressions and account-period +17 followers were not counted as Avery’s own post reach or prospects.

The source map, analytics review, cleaned CSV, exclusions ledger, calculations, one exploratory experiment, four-week plan, and full drafts are in this folder. Counts are ordinary Unicode character counts, not proof of account-specific platform fit. No current algorithm tactics, client endorsements, software product, prices, or sales results were invented.

drafts.md

# Avery Ellis — complete first batch

The text inside each publishing-copy block is complete. No posts, replies, or schedules have been submitted. Account-specific limits were not supplied, so platform fit is unverified; character counts are ordinary Unicode counts recorded separately. The three singles plus one thread are four main items, covering the first two weeks at two main items a week. Reply R1 is optional and separate.

## S1 · Count the work, not just the emails

```text
Our clarification emails went from 18 to 21. That number alone hides the useful comparison.

**Before an intake pilot:** 18 emails across 12 jobs in two weeks.
**During it:** 21 across 30 jobs in six weeks.

That is 1.5, then 0.7 emails per job. We also had more repeat clients, so I cannot credit the form alone.

Which question sends your team back to the client most often?
```

Compact alternative, if the account requires shorter copy:

```text
In one intake pilot, clarification emails per job fell from 1.5 to 0.7. The pilot also had more repeat clients, so I cannot credit the form alone. Which question sends your team back to the client most often?
```

## S2 · Test the alternate's authority

```text
**Try this before the next handoff:** the approval owner is away. What, exactly, can the alternate approve?

A name in the brief is a starting point. Walk one real decision through the absence and record where work must wait. That is a more useful test than assuming the handoff is covered.
```

Compact alternative:

```text
Before the next handoff, try this: the approval owner is away. What can the alternate approve? Walk one decision through the absence and record where work must wait. A name in the brief is only the starting point.
```

## S3 · What a workflow audit actually includes

```text
My workflow audit has a deliberately small scope: one intake-to-handoff process, a conversation with the team, and a short change proposal.

If a job repeatedly gets stuck between “requested” and “ready,” that is a process we can examine. There is no software product to install and no promised efficiency percentage.
```

Compact alternative:

```text
My workflow audit covers one intake-to-handoff process, a team conversation, and a short change proposal. If work gets stuck between “requested” and “ready,” we can examine that process. No software product or promised efficiency percentage.
```

## T1 · Thread: when does a question belong in the form?

Each numbered block is one complete thread entry. The thread is one main publishing item.

```text
1/5
I reduced an intake form from eleven required questions to three. “Shorter is better” would be the wrong lesson. The useful question was which information the team needed before it could start.
```

```text
2/5
Two team members said customers were sending emails instead of filling in the first form. I did not count abandonment, so I cannot say the edit improved conversion. It told us the form needed another look.
```

```text
3/5
The three required fields became intended reader, delivery format, and who approves the work. Each makes a different decision visible: who the work is for, what must be handed over, and who can decide it is ready.
```

```text
4/5
The remaining questions moved into a follow-up guide. They were not declared useless. Some detail matters once a job's needs are clear; it does not all have to block the initial request.
```

```text
5/5
Try one field: name the decision its answer changes. If the answer will not change the next step, consider whether it belongs in follow-up. If it establishes readiness or responsibility, make that purpose clear to the person completing it.
```

## R1 · Reply to the supplied Dana post

```text
A kickoff call can help, but the decision still needs an owner. In one handoff example, the brief named a department instead of a person. Naming an approval owner and an alternate made that responsibility explicit. I would put that decision in the record so it survives the call.
```

The parent argues that a kickoff call and shared channel usually fix broken handoffs. This reply adds the authority/record distinction; it does not imply familiarity with Dana, endorsement, or a proven universal cause.

## Internal checks, outside publishing copy

S1 is based on the supplied six-week pilot and two-week baseline at a nine-person translation studio. Counts and periods are preserved; anonymous descriptions are used, without client names or testimonial quotations. S2 is a proposed diagnostic exercise derived from the ownership example, not a claim that Avery already ran that exact exercise. S3 describes the supplied current paid offer; it adds no price, software capability, customer result, or sales history. T1 preserves the unmeasured abandonment gap and gives the questions a decision role. R1 responds only to the supplied parent text.

No draft requires an invented factual answer before use. Length limits and actual platform acceptance still need checking in the account. Where fuller copy does not fit, the supplied compact alternatives preserve the essential factual boundary; a longer-post entitlement has not been assumed.

analytics-review.md

# Descriptive analytics review

## Scope and cleaning

The supplied CSV has nine rows of cumulative post metrics. The export was assembled at 2026-09-18 12:00 UTC. Each row is a post at a snapshot, not an independent post or a common account-period total. The script `analyze.py` parsed the CSV, preserved quoted commas and blank fields, and retained source row numbers. The raw fixture hash was unchanged after execution.

Exact duplicate P102 at physical row 4 was excluded in favor of row 3. P104's one-day snapshot at row 6 was superseded by its complete seven-day snapshot at row 7; metrics were not selected independently by maximum. Seven distinct posts remain in `cleaned-analytics.csv`. `exclusions.csv` records both excluded observations.

## Comparable seven-day original posts

| Post | Impressions | Recorded link clicks | Clicks / impressions | Recorded attributed follows |
|---|---:|---:|---:|---:|
| P101 | 1,000 | 40 | 4.0% | 3 |
| P102 | 800 | 20 | 2.5% | 2 |
| P104 | 1,400 | 42 | 3.0% | 4 |
| P107 | 0 | 0 | Undefined, zero denominator | 0 |
| Pooled | 3,200 | 102 | 3.1875% | 9 |


The pooled click rate is the sum of recorded link clicks divided by the sum of impressions across these four original posts at age seven days. The zero-impression row remains in coverage and contributes zero counts, but it has no defined individual rate. This is not an average of individual percentages. Link destinations were not supplied, so these clicks cannot be called audit-page visits, inquiries, or purchases.

The median impressions across all four rows is 900; the range is 0–1,400. P104 supplies 43.75% of this cohort's impressions. There are only two intake posts, one approval post, and one documentation post. This is too small and uncontrolled to establish a durable topic preference, audience segment, or causal format effect.

## Keep the other observations separate

- P103 is a seven-day reply, with 500 impressions, six recorded replies, and one attributed follow. It is not pooled into main-post ranking.
- P105 is only three hours old. Its 1,600 impressions are not evidence that automation is the winning topic against seven-day posts. Its attributed-follows field is unknown, not zero.
- P106 is a five-day quote post. Its own impressions are 300. The quoted parent's 20,000 impressions are not Avery's reach. Its link clicks and attributed follows are unavailable.
- P107's zero impressions make its rate undefined; the row is not silently deleted to improve averages.
- Across all seven retained posts, own-post impressions total 5,600, but the ages and types differ. This is an audit sum, not a common-period performance comparison. Do not add the quoted parent's impressions.
- The notebook's net account follower change of +17 over September 1–17 is a different population/window from post-attributed follows. It is not 17 audit prospects and is not added to any post count.

There is no defined total-engagement column. Likes, replies, bookmarks, and link clicks are not summed and labeled as the platform's engagement total. Reply counts also do not show whether the comments were relevant, distinct people, or inquiries.

## What this can support

The useful descriptive observation is that the seven-day intake posts attracted recorded clicks and replies, while the export cannot tell us whether those interactions came from intended buyers or led to audits. Treat intake and approval questions as plausible writing material, not validated market positioning. Use the next batch to observe relevant problems in people's own words.

The separate professional pilot arithmetic is 18/12 = 1.5 clarification emails per job before, and 21/30 = 0.7 during. The baseline is two weeks and the pilot six weeks; more repeat clients and no randomization prevent attributing that change to the form. No staff-time, profit, satisfaction, or abandonment measure was supplied.

`calculations.json` contains executed numbers. No account access, publication, scheduling, attribution setup, revenue calculation, or audience-demographic inference was performed.

experiments.md

# One exploratory conversation experiment

**Audience question:** Which intake/handoff problem is concrete enough for an intended reader to describe their own example?

**Evidence or hypothesis:** Existing seven-day posts contain recorded replies and clicks, but their relevance and commercial meaning are unknown. Hypothesis: a specific diagnostic question at the end of a useful single post invites more relevant problem descriptions than ending with a self-serve action alone.

**Controllable change:** Ending type. Keep the same audience, single-post format, comparable length, and practical workflow level. Use distinct but comparable examples instead of near-identical reposts. Drafts O5 and O6 are Week 3 outlines to finish after the first two weeks provide actual reader questions. O5 ends with a specific question; O6 ends with a practical self-check. Topic and timing will still differ, so the comparison is exploratory rather than randomized causal evidence.

**Primary observation:** Number of distinct external people who describe a relevant intake/handoff problem in response to each post within seven days. Denominator: one eligible main post per condition, two total; also record impressions at the same seven-day age if available. Exclude Avery's own replies, spam, and repeat messages from the same person within a post. Do not infer relevance from a raw reply count.

**Commercial observation:** Log separately whether a person directly asks about the workflow audit or asks to discuss a relevant process. Those are the supplied inquiry definitions. An optional, voluntary “What prompted you to ask?” question may give self-reported source context. This instrument is proposed; no inquiry log or attribution is currently configured. Likes, bookmarks, follows, and link clicks are not inquiries.

**Comparison:** Read both conversations after each has a full seven-day window. Record the examples that readers describe and show counts per condition. Avoid a winner claim from one pair. Friday's 20-minute review can log partial observations, but label each immature result and wait for its seven-day check.

**Proposed decision rule:** If the question-ending post yields two or more distinct relevant problem descriptions, use that type of question once more with a fresh example. If the action-ending post also produces relevant descriptions, keep both approaches and compare the content of the problems rather than declaring a format winner. If neither produces any relevant description, revise the example's specificity or reconsider audience fit before repeating. Two is a practical trigger for another small trial, not a historical benchmark or statistical threshold.

**Stop/revise condition:** Stop an angle if readers repeatedly misunderstand the implied result or the available source does not support a requested claim. Correct the claim before continuing. If relevant conversations appear but no audit questions do, do not reclassify the conversations as inquiries or forecast sales. Examine whether the offer and problem fit before increasing promotion.

Suggested manual log columns: post ID, published time/timezone, seven-day observation time, draft/ending type, distinct relevant respondents, brief problem summaries, direct audit inquiries, self-reported source if volunteered, impressions if available, exclusions, and next decision. No personal details beyond what is necessary for the user's own record are required here.

four-week-plan.md

# Four-week rolling plan

Two main items per week, with optional occasional replies. A thread counts as one main item. No posting hour, timezone, or calendar start date has been established; Slot A and Slot B are chosen by Avery. The plan is not an active schedule or automation. Each Friday, use 20 minutes to read responses and record maturity of the observation window.

| Week / slot | ID and format | Audience question / purpose | Source | Draft status and intended action | Observation |
|---|---|---|---|---|---|
| 1 A | S1 single | What does a clarification count mean per job? | Intake pilot counts and caveats | Complete, plus compact alternative. Invite a concrete recurring question. | Seven days after publication; Friday partial check labeled by age. |
| 1 B | S2 single | Can the alternate actually make the decision? | Handoff incident plus proposed absence test | Complete, plus compact alternative. Encourage one practical decision walkthrough. | Seven days after publication; record relevant examples, not only totals. |
| 2 A | T1 five-entry thread | When should information be required? | Eleven-to-three false start and follow-up guide | Complete. Explain decisions and invite a one-field self-check. | Seven days after first entry; evaluate the thread as one main item and inspect where replies occur. |
| 2 B | S3 single | What does the audit include? | Current offer | Complete, plus compact alternative. Clarify paid-service scope without a result promise. | Seven days after publication; log direct audit/process-discussion inquiries separately. |
| 3 A | O5 single | Which missing intake detail blocks the next action? | A relevant question observed in Weeks 1–2, or a clearly labeled hypothetical if none | Outline only. Write a fresh concrete example and end with a diagnostic question. Experiment condition A. | Seven days after publication using the experiment's relevance rule. |
| 3 B | O6 single | Which handoff detail would you check before work leaves? | A distinct relevant question from Weeks 1–2, or a labeled hypothetical | Outline only. Same approximate length/format as O5; finish with a self-serve check. Experiment condition B. | Seven days after publication; do not compare an immature snapshot. |
| 4 A | O7 single | What did readers find unclear? | Actual responses to earlier posts | Conditional outline. Correct one recurring misunderstanding; if no responses, show a more specific worked example without inventing feedback. | Seven-day check may fall in Week 5. |
| 4 B | O8 single or short thread | What useful next step follows a real workflow question? | Actual relevant questions or a fresh author-supplied case | Conditional outline. Continue a promising question once, or retire the angle if sources/interest are thin. No automatic promotion. | Seven-day check in following week; append results instead of rewriting the original hypothesis. |


R1 is a finished optional reply to the supplied parent text. Review its relevance at the moment Avery chooses to use it; no parent account was accessed and no reply was sent.

## Weekly revision points

- After Week 1: retain S1/S2 only as working hypotheses. If respondents confuse observation with causation, sharpen the caveat before building further examples.
- After Week 2: choose O5/O6 from actual questions if any exist. If none do, label the examples hypothetical and avoid writing “people keep asking.” Keep cadence at two main items rather than increasing volume to manufacture signal.
- After Week 3: compare only mature observations. Use the experiment's declared practical trigger; preserve the original plan and raw response notes. No permanent algorithm rule follows from one pair.
- After Week 4: record what was actually learned and what remains unknown. A direct inquiry is worth examining separately from engagement, and no inquiries is a result to retain. Decide the next four-week scope after that review.