# Every · NPS & reader perceptions

## Open the dashboard

Open **index.html** in a browser. It is a self-contained, offline dashboard: no installation, server, CDN, analytics or network calls. It includes a score distribution, theme comparisons, evidence-backed findings, and a searchable/filterable explorer of all 195 comments. Its CSV export downloads the currently filtered responses; its brief button downloads the full written analysis. JavaScript is required for charts and filters.

## The score

**NPS: +45.1**, calculated as `(107 promoters − 19 detractors) / 195 × 100`.

| Measure | Result |
|---|---:|
| Responses / nonblank descriptions | 195 / 195 |
| Promoters (9–10) | 107 · 54.9% |
| Passives (7–8) | 69 · 35.4% |
| Detractors (0–6) | 19 · 9.7% |
| Mean rating / median | 8.38 / 9 |
| Score of 8 | 44 · 22.6% |

Promoters outnumber detractors, but there is no historical comparison or external benchmark in the supplied data. Do not interpret +45.1 as an improvement or an industry-relative rating.

## What the comments suggest

### 1. Practical learning is a promising message to test

55 descriptions explicitly mention learning, applying AI, workflows or concrete work benefits. This group has NPS **+65.5**, compared with **+37.1** among the other 140 responses: a **28.3-point descriptive difference**. There are 39 promoters, 13 passives and 3 detractors in this theme. This is an association in how people describe Every, not proof that practical content causes advocacy; themes overlap and coding is interpretive.

> “The best resource for having any hope of actually using AI in your daily work” — response #002, score 10

> “A playbook for how to work AI-first” — response #142, score 10

**Action:** Test “learn how to put AI to work” messaging and short, outcome-led guides. Compare activation and subsequent recommendation scores in a planned test before calling this an NPS driver.

### 2. The writing–software combination is recognized, but mention is not adoption

141 descriptions mention content or analysis (72.3%), 101 mention products or building (51.8%), and **81 mention both (41.5%)**. The four mutually exclusive positioning groups are: both 81, content without products 60, products without content 20, neither 34. “Content without products” means no explicit product reference, not that the person uses only content. These are descriptive mentions, not inferred preferences.

Three comments explicitly say the bundled apps are unused or not fully used: **#041 (10), #056 (8), #061 (2)**. This includes one security concern. Separately, #052 (5) says they lack time to learn more, but does not specifically attribute that to the apps. High satisfaction and incomplete app adoption can coexist.

> “Premium newsletter bundle with products I should be using.” — response #056, score 8

> “Subscription would be worth it for the writing alone, incredible benefit for writing + Cora + Monologue (particularly Monologue, which I absolutely adore)” — response #184, score 10

**Action:** Protect editorial quality while testing a one-job, one-app onboarding path. Measure first meaningful use rather than assuming bundle awareness equals activation.

### 3. Platform expectations are a small but concrete access problem

Only **3 of 195** descriptions explicitly raise platform availability: **#030 (8), #051 (7), #120 (0)**. Two are passives and one is a detractor. #120 says they discovered Mac limitations after subscribing and states an intention to cancel. That comment also discusses price, content and the survey experience, so its low score cannot be assigned solely to platform coverage. The data establish neither actual cancellations nor the overall prevalence of this issue.

**Action:** Make per-product OS availability prominent before purchase, and provide an immediate useful path for non-Mac subscribers. Validate demand with platform/usage data before committing to a port. Separately investigate the explicit security concern in #061 and the product-confusion/customer-service complaint in #009.

### 4. The AI pivot attracts some readers and alienates others

Two detractors explicitly contrast Every with its earlier editorial identity: **#008 (3), #137 (3)**. In contrast, promoter **#126 (10)** positively describes the shift to an AI product-development house. This is a qualitative positioning tension, not evidence of a measured decline over time. Passive #114 (7) says “WAS a great resource,” but offers too little context to classify it as an explicit pivot objection.

> “I would about 2 years ago, but not anymore. I find Every less interesting after it switched over to AI and building products” — response #137, score 3

**Action:** Explain the current editorial promise and how products support it. Follow up with legacy readers to learn whether they miss topic breadth, writing quality, or a different offer; do not assume which from these descriptions alone.

### 5. Price and value are not experienced uniformly

Five responses explicitly praise value or justify the subscription price: **#040 (9), #043 (9), #102 (8), #135 (10), #184 (10)**. Two raise monetization or pricing concerns: **#034 (0), #120 (0)**. These selected, overlapping signals do not form an exhaustive value-sentiment scale. #034 alleges a breach of trust; the dataset cannot verify that allegation.

**Action:** Clarify which apps, events and courses are included versus paid separately. Test expectation-setting before considering broad price changes.

### 6. Score-8 respondents are a useful follow-up audience

**44 people rated 8**, the largest group below the promoter threshold and 63.8% of all 69 passives. Their descriptions include praise as well as access and adoption concerns. Ask them what would make the recommendation a 9 or 10; this survey did not ask that.

For scale only: moving 10 existing passives to promoters would add **5.1 NPS points**, taking the fixed 195-response sample to **+50.3**. This is arithmetic, not a forecast or a claim that the proposed actions will cause that change.

## Theme audit

Themes are multi-label, so counts and percentages do not sum to 195 or 100%. “Outside theme” means the theme was not explicitly coded in that description, not a matched control group. The community/live-learning theme is small (n=9); do not rank it as a proven driver.

| Theme | Responses | Share of 195 | Within-theme NPS | Outside-theme NPS |
|---|---:|---:|---:|---:|
| Content & analysis | 141 | 72.3% | +45.4 | +44.4 |
| Products & building | 101 | 51.8% | +42.6 | +47.9 |
| Practical learning | 55 | 28.2% | +65.5 | +37.1 |
| Frontier & expertise | 39 | 20.0% | +64.1 | +40.4 |
| Community & live learning | 9 | 4.6% | +55.6 | +44.6 |

The complete coding definitions and response assignments are in **comment_coding.json**; raw comments and readable labels are in **coded_responses.csv**. Exact calculations, cohorts, metadata and all records are in **analysis.json**. Response IDs are one-based data-row numbers, not customer IDs; CSV physical lines are one higher because of the header.

## Suggested next steps

| Priority | Next step | Evidence / validation |
|---|---|---|
| Now | Clarify platform coverage and included benefits before purchase | 3 platform comments; 2 price/trust concerns. Audit the purchase journey. |
| Next | Follow up with score-8 respondents and explicit friction cases, if a source system can link them | 44 score-8 responses. The supplied file has no identities or contact details. |
| Next | Test a practical, outcome-led first-use experience | Practical-learning association; 3 explicit unused-app descriptions. Measure activation and follow-up NPS. |
| Explore | Check legacy-reader needs and appetite for other platforms | 2 pivot objections; 3 platform comments. Use targeted research before roadmap changes. |

## Assumptions and limitations

- Source: **nps_and_descriptions_only.csv**, containing only a score and a description. All 195 scores are valid integers from 0 to 10 and all descriptions are nonblank. No records were dropped or imputed.
- All rows are treated as separate responses. There are no respondent identifiers for deduplication; similar descriptions are not grounds to remove a row.
- The score field is assumed to use the standard 0–10 recommendation scale. Promoters = 9–10; passives = 7–8; detractors = 0–6. NPS is not the mean rating or a percentage.
- No survey collection date, invitation count, response rate, sampling frame, customer segment, usage history, prior period or benchmark was provided. Nothing is inferred from the file date or an individual comment's year reference.
- The comment prompt was **“How would you describe Every?”**, not “Why did you give that score?” Findings are perceptions and hypotheses, not established reasons for scores.
- Themes and narrow qualitative signals were analyst-reviewed across all rows. Coding is subjective, not independently validated, and does not use automated sentiment scores. A label does not imply a positive or negative view unless its definition explicitly says so.
- Absence of a theme or complaint does not mean the respondent does not hold that view. Small complaint counts are leads for follow-up, not estimates of incidence across all customers. No statistical-significance claim is made.
- All computations use full precision; display values are rounded to one decimal, except the mean. NPS differences are calculated before rounding.

## Verification

See **VALIDATION.md** for checks and limitations. Source fidelity, score and theme calculations, embedded data, CSV contents, and JavaScript interaction logic were checked. Browser startup failed in this environment, so visual layout and actual browser download behavior remain unverified.

## Reproduce

Run `python3 build_dashboard.py` from this folder (Python standard library only). It reads the source CSV, **comment_coding.json**, and **dashboard.template.html**, then rebuilds **index.html**, **analysis.json**, **coded_responses.csv**, and **ANSWER.md**. The HTML embeds both its data and the written brief, so it can be shared as one file.
