Actions

What to do next, ranked. The list is refreshed every Monday from the week's data; each item names its evidence, and the expected effect is a range or nothing.

Refreshed Mon 14 Sep, 04:40 · next Mon 21 Sep, 04:00

Open

Computed

14

Waiting on someone

High impact, open

Computed

4

Where the evidence is strongest

Done

Computed

4

Ticked here or found in the change log

Dismissed

Computed

0

Kept in the log with the reason

The list

Ranked by impact against effort. Tick an item when it is done; dismiss one that does not apply, with the reason kept.

Computed
  1. 1

    Rewrite titles and descriptions for eight queries under the click-through curve

    All eight sit in the top ten and click at less than 60% of what their position should earn.

    Contenteffort Shigh impactContent+130 to +230 clicks a period
  2. 2

    Fix LCP on the product template: preload the hero image

    Mobile LCP is 3.3 s against a 2.5 s target and product pages hold four of the six slowest URLs.

    Technicaleffort Mhigh impactDevLCP −0.6 to −0.9 s on product pages
    See in Site health
  3. 3

    Reduce INP on product pages: defer the reviews widget

    INP is 521 ms on mobile; the third-party reviews script blocks input for about 300 ms.

    Technicaleffort Shigh impactDevINP −200 to −300 ms
    See in Site health
  4. 4

    Write 'The best filter coffee brands' as a comparison page

    Position 11.2 for a commercial query with 2,900 searches a month, and no page targets it.

    Contenteffort Mhigh impactContent+84 to +140 clicks a period
  5. 5

    Refresh 'The benefits of coffee'

    The post fell from 72 to 12 clicks against the previous period; two cited studies are from 2019.

    Contenteffort Mmedium impactContent+40 to +70 clicks a period
    See in Blog
  6. 6

    Link the V60 guide from six older brewing posts

    The guide reached position 2.8 in its first week; links from the cluster hold it there.

    Contenteffort Smedium impactContent
    See in Blog
  7. 7

    Add Product and FAQ structured data to every product page

    Only 'product info' rich results show; review stars appear on one page and FAQ on three.

    Technicaleffort Smedium impactDev
    See in Google visibility
  8. 8

    Publish a comparison page: Harbor Lane against the alternatives

    The assistants answer four comparison questions without us; the one comparison that names us is answered at rank 1.

    GEOeffort Mmedium impactContent+2 to +4 mentions in twenty
    See in AI visibility
  9. 9

    Merge the two cold brew pages competing for one query

    The post and the pitcher page split 2,700 impressions and neither holds a position.

    Contenteffort Mmedium impactContent
  10. 10

    Fix the redirect chains on the old /yazilar/ URLs

    38 URLs pass through two hops since the August move.

    Technicaleffort Smedium impactDev
    See in Site health
  11. 11

    Connect Bing Webmaster Tools

    Bing feeds Copilot and ChatGPT web search, and there is no visibility data for it today.

    Trackingeffort Smedium impactMarketing ops
    See in Data sources
  12. 12

    State the monthly subscription price on the page

    Assistants answer price questions from marketplace listings; the subscription page gives no figure to quote.

    GEOeffort Slow impactContent
    See in AI visibility
  13. 13

    Sample ChatGPT and Perplexity alongside Gemini

    Only Gemini is measured; the other assistants send most of the AI-referred sessions.

    Trackingeffort Mmedium impactMarketing ops
    See in AI visibility
  14. 14

    Answer the trust questions on a brand FAQ page

    Brand questions are answered from marketplace reviews, not from the site.

    GEOeffort Slow impactContent
    See in AI visibility

Tests to run

Changes worth measuring before they are rolled out to every page. One hypothesis, one metric, one duration.

Computed
  • Add the year to the titles of five evergreen guides

    running

    A visible '2026' in the title lifts click-through on evergreen how-to queries.

    Metric: CTR of the five pages, Search Console · 4 weeks

  • Put the price in product meta descriptions

    proposed

    A price in the snippet raises clicks from buyers and lowers bounces from browsers.

    Metric: CTR and engagement rate on product pages · 4 weeks

  • Review snippets on the ten best-selling products

    proposed

    Stars in the result add clicks at the same position.

    Metric: Rich result clicks, Search Console · 6 weeks

  • Shorter blog introductions

    done

    The method above the fold keeps readers who came for the answer.

    Metric: Engagement rate, GA4 · 3 weeks

    Engagement rate 58% → 63% on eight posts; kept.

Done this period

What was closed, and what the numbers did afterwards.

Change log
  • Published the V60 brewing guide

    8 Sep 2026

    Position 2.8 in the first week, +204 clicks

  • Added the FAQ block to the product template

    2 Sep 2026

    FAQ rich results on three products so far

  • Moved the blog under /blog/ with 301 redirects

    19 Aug 2026

    Indexed pages held; 38 redirect chains remain (open)

  • Preloaded the home page hero image

    5 Aug 2026

    Home LCP 3.1 s → 2.4 s on mobile

Marked done from this list or found in the change log; the effect is read from the next weekly refresh.

Monday e-mail

The report as a PDF, on its way before the week starts.

E-mail
Schedule
Every Monday, 09:00
Recipients
owner@harborlanecoffee.comcontent@harborlanecoffee.comdev@harborlanecoffee.com
Provider
Resend
Last sent
Mon 14 Sep, 06:00delivered

Deliverability is watched: one of the last four sends landed in spam. A warm-up tool is being evaluated on a spare account first, never on the sending domain.

How the list is made

Three steps every Monday, one of them a person.

Computed
  1. 1

    The refresh reads the sources

    Search Console, GA4, PageSpeed, Semrush and the assistant sampler are read once into the store; nothing is fetched twice.

  2. 2

    The model ranks

    With the site's own priorities, the past decisions and the change log as context, so a rule the team already rejected is not proposed again.

  3. 3

    A person decides

    Marks done or dismisses; the log keeps both, and the effect is read from the next refresh.

The model proposes; the team changes the site. Nothing on this tab edits a page.