Consumer Fusion · internal work plan · revised 31 August 2026
Stop shipping projects. Run a weekly line.
Rewritten for a different goal: traffic up week over week from organic and AI search. That is not the same objective as the previous version, and it changes both the primary lever and the operating model.
What changed and why
The earlier plan optimised for pipeline: fix the foundation, recover rankings, repair the funnel. Correct for revenue, wrong for a weekly number. Phased projects deliver in lumps, and three of the four phases produce nothing measurable for six weeks.
The lever also changes. Rankings are the slow way to more traffic here. Click through rate is the fast way, because the impressions already exist.
Segment of non branded impressions
Queries
Impressions
Clicks
CTR
Topically relevant, position 20 or better, 10+ impressions
562
39,419
31
0.079%
Topically relevant, all positions
2,994
66,799
38
0.057%
Off topic but real traffic
1,326
16,787
17
0.101%
Junk, site: operators and other companies
56
350
0
0%
39,419 impressions on relevant queries within reach, earning 31 clicks. Normal click through at those positions would be roughly 1 to 5 percent depending on rank. Even the bottom of that range is an order of magnitude more traffic with no new rankings, no new content and no backlinks. That is the engine, and it is the only lever on this account that can move a weekly number.
The levers, ranked by speed to effect
1Title and meta rewriting, as a production line3 to 14 days
Google recrawls and re-renders titles fast. This is the only change on the list where effect shows inside one reporting week.
Batch of 15 to 20 pages per week, worked in descending order of impressions. Roughly 100 pages carry meaningful impressions, so this is a six to eight week run, not a perpetual task.
Every rewrite is scored against the query actually driving the impressions, not the page's topic in the abstract.
Highest volume first: "can you remove google reviews" (1,308 impressions, position 13), "franchise review management" (839, position 13.4), "how to remove negative google reviews" (660, position 16.4).
2Bing Webmaster plus IndexNowImmediate
This is the real AI search lever, and it is the concrete one. ChatGPT search is grounded in Bing's index. Being properly indexed and fresh in Bing is measurable and free, unlike AI visibility, which we currently cannot measure at all.
Bing Webmaster Tools is not connected in the current setup. Connect it, submit the sitemap, verify.
Turn on IndexNow. It pushes changed URLs to Bing instantly rather than waiting for a crawl, which shortens the feedback loop on every single title rewrite in lever 1. The two compound.
Our own May audit found Bing organic produced 1,667 sessions and 52 tracked conversions over 12 months, a better per session rate than most channels. It is under worked.
Bonus: Bing's crawl stats will answer the AI crawler question the audit could not close.
3Internal linking2 to 4 weeks
All 85 articles sit at identical click depth, reachable only through the /resources hub. Zero contextual links between articles. Zero links from solutions pages to matching articles.
Nothing on the site is currently signalled as more important than anything else. This is the cheapest structural fix available and it lifts position, which multiplies lever 1.
Batch of 10 links per week. Solutions to article first, then article to article.
4Refresh existing articles, do not write new ones3 to 6 weeks
85 articles already carry impressions. A refresh inherits that history; a new article starts from nothing and takes months.
One article per week, chosen by impressions divided by position. Update the content, fix the heading structure, add the internal links from lever 3, resubmit through IndexNow.
Two known content defects to sweep while in there: several articles carry multiple H1 tags from a CMS templating bug, and the AEO article's jump navigation points at headings that do not exist on it.
5Backlinks8 to 12 weeks
Still the leading unresolved cause of the position decline, and still worth doing. But it will not move a weekly number, and it is blocked on a subscription.
Runs in the background. Do not let it gate anything else. That was the mistake in the previous version of this plan.
The weekly cadence
Monday
Pull last week against the week before: clicks, impressions, average position, page one CTR. Non branded only, de branded manually. One number leads the report: non branded clicks.
Mon to Wed
Ship the batch. 15 to 20 titles and metas, 10 internal links, one article refresh. Push every changed URL through IndexNow the same day.
Thursday
Check the batch from two weeks ago, not last week's. Title changes need seven to fourteen days to show. Anything that moved down gets reverted or rewritten.
Friday
Log what shipped and what moved. Pick next week's batch from the impressions list.
The discipline that matters: measure the batch, not the site. Sitewide numbers are noisy and still recovering from the replatform. Per batch measurement tells you whether the rewrites are working, which is the thing you actually control.
What I would cut
New content production. Slowest path to traffic on a site that already has 85 underperforming articles. Refresh instead.
The AI visibility page. Demand has never been measured. Building for it is a guess, and it competes with eight existing articles plus the gated guide.
The healthcare consolidation. Right call eventually, but it needs a decision from the client and produces nothing weekly.
Backlinks as a week one gate. Important, slow, and blocked. Background it.
Honest expectations
Weeks 1 to 3 may still decline. Position has fallen every week since the replatform, nine weeks running with no recovery week. That trend has momentum and the first job is arresting it, not reversing it. Anyone promising an up week by week two is guessing.
Growth will not be linear. Title rewrites land in steps as pages get recrawled. Expect flat weeks between moves. Judge the batch, not the week.
What should be true by week 6: the first non declining position week since July, and non branded clicks meaningfully above the current 55 per quarter baseline. The mechanism is entirely click through on impressions that already exist, so it does not depend on the backlink subscription, new content, or anything the client has to decide.
By week 10 to 12: the position recovery from levers 3 and 4 starts compounding with the click through work, which is where the curve should steepen rather than flatten.
Week one, specifically
Connect Bing Webmaster Tools and enable IndexNow. Free, immediate, and it accelerates everything else. Do this before the first batch ships.
Investigate the /lp/franchise-reputation collapse. Near total visibility loss 1 to 22 August, partial recovery from the 23rd, cause never established. If something broke, fixing it is instant traffic on the account's best commercial page.
Fix the many to one redirect mappings./franchise-reputation-management-4/, /franchise-b/ and /enterprise-franchise-solutions-b/ all fold into one page, and /review-calculator/ lands on the homepage.
Regenerate the sitemap with real lastmod dates. 81 of 85 articles claim they have not changed since before the move.
Ship batch one: the top 20 pages by impressions, titles and metas rewritten against the queries actually driving them.
Two traps
The data. Windsor holds 20 Search Console properties and blends all of them on an unfiltered pull, returning about 2.7 million impressions. Always pass accounts: ["sc-domain:consumerfusion.com"]. Its branded_vs_nonbranded field is also wrong here, filing "consumer fusion" as non branded because it matches the domain string without the space. De brand manually with consumer\s*fusion|consumerfusion.
The unexplained part. Some page one impressions convert at rates too low to be normal search behaviour, including queries at a flat position 1.0 that land on the homepage with no topical relevance. We could not explain it and the AI Overview theory did not survive checking. Do not assume every impression in the 39,419 is a real person who can be won. Work the batches, measure what actually moves, and let the data tell you which segments are real. Someone logged out still needs to manually search those position 1.0 queries; ours got CAPTCHA blocked and it is a ten minute job.