Growing client rankings

A programme Flux runs every month, not a tool purchase. Nine cards and six decisions. Tap Approve or Needs changes on each card, pick a chip on each decision, then Send my review at the bottom. Progress saves, so it is safe to close.

THE SHAPEHow a ranking actually moves

Six things in this order. Skipping step 1 is why most SEO retainers cannot prove anything.

1
Measure first, on three separate surfaces
Organic rank, local pack, and Google Maps disagree constantly. One blended "rank" number hides the disagreement and makes every later claim unprovable. Plus a geo-grid heatmap showing the radius where the client vanishes.
Vera · DataForSEO · monthly
▼
2
Raise the ceiling: fix the technical floor
Schema, the head layer, titles with keyword and geography, image weight, a real 404 handler. This does not create rankings by itself. It removes the things capping them, and it is one-time per site.
Vera and Mose · pull requests · one pass
▼
3
Build the page inventory the searches need
The biggest lever by a distance. One omnibus services page cannot rank for five service lines across a metro. Keyword research produces a taxonomy, the taxonomy produces pages, the pages produce rankings.
Wren researches · Iris writes · Mose builds
▼
4
Own the local surfaces
Category, services, hours, attributes, NAP consistency, real photos, review replies. Google says local ranking is relevance, distance and prominence, and that none of it can be bought. This is the unglamorous work that actually moves the map pack.
Iris drafts · in-house photography · monthly
▼
5
Publish on a cadence, in the words people type
Trade vocabulary and search vocabulary diverge. "For lease" can lose to "for rent" by six times in the same market. The fix is never to rewrite the brand voice, it is to carry the searched phrasing in titles, headings and the first paragraph.
Iris · Jacob approves before publish
▼
6
Prove it with the diff, not a snapshot
A crawl is stale the moment it is written. The client sees a dashboard that updates, showing what changed since last month: ranks gained, grid coverage widened, defects closed. That is what makes a retainer renew.
Mose builds the surface · Otto runs the cron
Why this order. Steps 2 and 3 are expensive to reverse. Taxonomy and URL structure are SEO decisions and get made before art direction; look and feel comes after. On an existing site the output of step 3 is a retitle list plus pages to add, not a critique of the build.

STARTING POINTWhat is already built

Checked live today, not recalled. This is why the programme is mostly wiring rather than a project.

PieceStateWhat that means here
DataForSEO feedlive, fundedVerified today: $48.30 balance on Jacob's own account, $51 lifetime spend. At $0.002 per grid pin that is roughly 24,000 heatmap points in hand. Every geo-grid product on the market resells this same feed.
Firecrawl crawlerself-hostedRunning in Docker on the Flux droplet, loopback only, no per-page fees. This is the site audit engine. No crawl budget to buy.
Vera (data and intelligence)liveSEO research, site capture and scoring is already her named domain. Heartbeat 2.8 hours ago, so she is awake.
Audit methodologywrittenEncoded from real launches: soft-404 hosts, preview hostnames baked into canonicals, JSON-LD that parses as dead, forms that return success and deliver nothing. More specific than any vendor checklist.
Reporting surfacepartialSignal dashboard exists and already calls DataForSEO for local competitor discovery. It has no per-client SEO view and no month-over-month diff. That is the build in phase 1.
Monthly automationpartialOtto owns cron on the droplet. Nothing is scheduled for SEO yet. It must be a droplet cron, not an agent heartbeat, for the reason in the Reliability card below.
Backlink indexdoes not existNo index of who links to a client. DataForSEO sells backlinks data as a separate line item. Until then, link and citation work is judged blind.
AI answer visibilitydoes not existNo tracking of whether clients get cited in ChatGPT, Gemini, Perplexity or AI Mode. Firecrawl has no model provider wired on purpose, to avoid per-token billing. This is the one gap worth paying a vendor for.
Coach (chief of staff)stale 138hCoach has not heartbeated in roughly 5.7 days, and the second Coach Desk agent is at 564 hours. Do not make Coach the owner of a monthly recurring programme until that is fixed. Everyone else is awake.

THE PROGRAMMENine cards to approve


YOUR CALLSix decisions

Each one has a recommendation. A skipped chip is read as not decided, not as agreement.

Nothing on this page is built or bought yet, and no client has been contacted. This page comes down as soon as your review lands.

0 of 0 reviewed