Skip to main content
How to write the monthly review doc, the thing you present at the review. It’s the companion to 3.8 Running the client review: 3.8 is the end-to-end procedure (pull the data, build the evidence, run the call), and this page is how you turn that data into commentary the client actually values.
🎥 Video: Prepare a good client review
Why the doc matters. A good review does three things at once:
  1. It establishes credibility: we did what we set out to do, shown as proof rather than adjectives.
  2. It extracts strategy: the client helps us focus the next cycle’s keywords and directions.
  3. It creates FOMO: the next-month plan is co-created, so the client feels what they’d lose by not continuing. See Part 1 §1.10 Attribution for the philosophy that grounds every number in this doc.
The review is written on Notion. Screenshots are uploaded to the current month’s page first (see 3.8); this SOP is about writing the commentary around them.

​
Inputs checklist (before you start)

  • Last month’s review doc on Notion (search "[Month] - [Client]"): the structure, cluster names, and tone to match
  • This month’s review page on Notion, with screenshots already uploaded (no commentary yet)
  • GSC screenshots: overview, per-category (/blog, /uses, features…), rank tracking
  • PostHog conversion-funnel screenshots (cluster-level)
  • AI-traffic data (Pages with AI Traffic insight; see Month-1 §PostHog SQL)
  • The cycle’s keyword plan and the What’s-Next roadmap (keywords targeted but not yet built)
  • The next-month output plan (what ships next cycle); this is the FOMO lever

​
Step 1 — Read last month’s review first

  • Open the previous month’s report and study the section layout, the client’s page clusters (Overall, Blog, Uses, feature areas…), and the tone of the insights
  • Match it. Some clients have fewer or extra sections, so always mirror what the previous month used rather than imposing a template
The structure is per-client. Don’t invent sections; inherit them.

​
Step 2 — Write the Overview (strategic, no numbers)

  • 3–5 bullets summarizing the most important takeaways across the whole report
  • Keep it strategic rather than a recitation of metrics: no numbers, no percentages, no before-and-after
  • Plain language. Ask: what would the client care about most?
Good: Overall increase in traffic · Bad: clicks up 5% to 5.09k

​
Step 3 — Write Key Metrics commentary (cluster by cluster)

One subsection per cluster the client tracks (Overall, Blog, Uses, feature areas…). Under each cluster’s screenshots, write commentary bullets. The rule that makes the whole doc work is to think in clusters, not pages. Every insight is about a category, theme, or pattern: a content type (listicles, comparison pages), an intent cluster (immigration docs, manga translation), or a product area, never an individual URL.
  • Group individual pages into clusters based on what they represent
  • Write cluster-level trends (which groups are growing / declining / emerging)
  • Note positioning signals, strategic implications, cluster-level anomalies, conversion patterns
  • Keep each bullet to one short sentence
The golden rule: no numbers. If a bullet contains a number, a percentage, or a before-and-after comparison, rewrite it. The screenshots already show the data; commentary exists to interpret the pattern. It’s the most important rule in this SOP.
Write thisNot this
"free" tool pages still doing well as expectedfree-manga-translation-tool surging: 102 clicks (+79)
the wedge is in format-specific use cases (PPT, scans, manga)document-translator page (chatgpt UTM) leading at 44 events
overall traffic looking good, still picking upclicks flat, impressions up ~5%, position 9.3 → 7
Also skip: channel-attribution statements ("ChatGPT is the primary AI driver", which the client already knows), pricing-page mentions (explicitly flagged as not useful), and “top page” lists with counts. Showing supporting URLs: when a cluster insight needs examples, indent 3–4 URLs under the insight, never as a standalone list: 
- Internal knowledge base pages are the breakout cluster
    - /uses/internal-hr-policies-chatbot
    - /uses/ai-search-bar
    - /uses/healthcare-provider-knowledge-base-chatbot
Tone: brief and plain, like a colleague giving a quick status update rather than an agency deliverable. No fluff, no performative warmth, no hedging, no marketing-speak.

​
Step 4 — AI Traffic & Conversion sections

AI Traffic / AI Referral:
  • Talk about which page clusters AI engines reference (e.g. uses pages with the 'free' modifier are the most-referenced cluster)
  • Note any new clusters appearing in AI referrals that weren’t there before; connect to strategy if relevant
  • Do not list individual pages with event counts; do not name the “primary” AI platform
Conversion (PostHog):
  • Report the funnel numbers as-is (visitors, conversion rate, avg time to convert); these are already cluster-level, so numbers are fine here
  • Compare to the previous month if available
  • Note the relative conversion strength of different clusters, and anything unusual in time-to-convert between clusters

​
Step 5 — Opportunities & Blockers (the strategic meat)

Two subsections: What’s Working and What’s Next. What’s Working: keyword rankings that prove the strategy is paying off. Group by cluster with a bold header; each line is just the keyword plus position. No narrative, the rankings speak for themselves. 
**Document translation (core cluster)**
- MCP server document translation - #1
- document translation API for AI agents - #1
- dialect specific translation service - #1
What’s Next, the keyword roadmap (the section that drives retention):
  • Numbered clusters, each with a bold header and a one-sentence strategic framing
  • Keyword lists under each cluster, grouped into sub-verticals where relevant, indented
  • A Recommended Split table at the end: how to allocate targets across clusters by % and count
  • Every target keyword ideally gets both a landing page (/uses) and a blog; blogs sometimes outrank LPs for competitive terms (e.g. language pairs), so note both content types
FOMO comes from volume, not from narrative urgency. Don’t write “this is a land grab” or “window closing.” List 50–80+ specific keywords across 4–6 clusters. The client sees the opportunity is massive and this cycle’s output is finite, and that gap is the pitch.
  • Show the pace / ETA math for big backlogs: last month we shipped 7 of 100 language pairs — at this pace, ~14 months just for this cluster. (Raymond’s framing: “last month we tackled 20, the list is 100, it’ll take 5 months just for this.”) This makes the capacity constraint tangible.
  • Prove capability before scale. What’s Working shows the keywords we targeted are winning; What’s Next shows the massive remaining list. The flow is: we targeted these, we won, there are 100 more, and at this pace it’s X months.
## What's Next
1. **MCP / AI agents**
    - Not yet ranking for "Translation for AI agents", so target the adjacent keywords
        - keyword 1
        - keyword 2
2. **API-led growth** — developer-vertical deepening (Legaltech, Insurtech, eCommerce…)
        **Insurtech expansion:**
        - keyword 1
        - keyword 2
3. **File type expansion** — 22 formats supported
    - keyword 1

### Recommended Split
| Cluster | % of Keywords | Count |
| --- | --- | --- |
| API Vertical Deepening | 30% | 10 |
| AI Agents & MCP | 20% | 7 |

​
Step 6 — Output plan + Action Items

  • Output plan: state this cycle’s committed output right after What’s Next: the number of pages shipping this month, split across clusters (landing pages and blogs). Keeping it simple next to the huge What’s Next roadmap is what makes the constraint visible. ❓ [needs Raymond / [DECIDE]: standard-tier output per cycle — pricing rework in progress] ❓ [DECIDE: whether to present the cycle’s output as a named customer-facing package]
  • Action Items: concrete, assignable next steps that came out of the review:
    • Debug sessions (e.g. file-type pages not ranking — debug session next week)
    • Technical fixes (e.g. fix language-pair display on uses pages)
    • Indexing tasks (e.g. manually index new pages — don't wait for the sitemap crawl)
    • Content priorities for the next cycle
  • Leave placeholder sections (Billing, Others) empty if no data is provided

​
Real prompts

Align the write-up to the last review, so the doc lands on the concerns already agreed: 
based on everything we've discussed in this chat, summarize your findings and make sure
they're aligned to what was discussed in the [month] review — in terms of the plans and
concerns raised, especially the ones [stakeholder] brought up.
Pi produces a review-ready summary that ties back to the named stakeholder’s concerns from the prior call. (See prompt library §5.6.) Feed the review transcript to set the next cycle (used at the end of the review): 
[paste review meeting transcript]  this is the [month] review with [client]. based on this
and the product bible (which should be the most updated), come up with keywords and select.
[client] also flagged: [paste any client-suggested keywords / news angles].
Pi inherits what was agreed into the next cycle’s keyword strategy. Hand off to 3.4 Organise keywords.

​
Acceptance checks

  • Overview has 3–5 strategic bullets with zero numbers
  • Every Key Metrics bullet passes the golden rule (no number, %, or before-and-after)
  • All commentary is at the cluster level, not per-URL; supporting URLs are indented
  • AI Traffic section names no individual pages and no “primary” AI platform
  • What’s Working = keyword + rank only; What’s Next has numbered clusters + a Recommended Split table
  • The pace/ETA math appears for at least one large backlog cluster
  • Output plan states the cycle’s committed page count; Action Items are concrete and assignable
  • Section structure matches last month’s doc; all uploaded screenshots preserved

​
Edge cases — quick lookup

ScenarioWhat to do
Client questioned content qualityLead with the highest-ranking posts as counter-evidence (Step 5, What’s Working)
A discontinued service still shows in the dataSkip it entirely, even if it’s performing (e.g. Bluente stopped certified translation; never mention it)
A “known-working” strategy (e.g. “free” modifier)Mention it briefly (still doing well as expected) but don’t dwell; spotlight the emerging clusters instead
No data for a sectionLeave the section as an empty placeholder; don’t invent commentary
Editing an existing Notion pageUse targeted replacements to add structure/commentary around existing screenshots; preserve all images

​
See also