The voice agent explains any task, answers questions grounded in this plan, and records a brain dump. Voice feedback is read-only and cannot approve or execute work. Allow the microphone when your browser asks.
The explicit Sonnet rollback keeps the existing capsule-only protocol. A lost connection retries the same Sonnet engine twice and reloads the complete persisted conversation. It never changes models during a conversation.
Sonnet 5 · no fallbackVoice answers are review evidence. Approval, sending, spending, and production changes stay in the originating chat. Voice trouble in this page? Open the standalone voice page.
Plan: /Users/agency/Documents/Agty/Fusion/action-plans/2026-08-10-fusion-site-launch-and-ppc-plan.md
J/K, to open it.1 delegate, 2 drop, 3 idea, 4 ask first, 0 clear, J/K to move between tasks (and open the one you land on).! to make it a hard requirement it cannot ignore.Already shipped — 2 tasks, no decision needed
Rank the product categories by how ready each is to carry paid traffic, so budget goes where the catalogue can actually convert.
Spreading a small budget evenly across ten categories guarantees that no category gets enough data to learn from, and some categories are not commercially ready at all.
Budget fragments across categories whose pages or stock facts cannot support a sale, and the account never accumulates enough conversions in one place to optimize.
A defensible order of activation, so the first category launched is the one most likely to return evidence quickly.
Delivered. The readiness matrix ranks billiards first, ping-pong second and foosball third, with sixty products reconciled and two found outside the category structure. Explicitly preparation only, as no category was cleared for activation. That remains a client decision inside the launch approval.
flowchart LR A[Ten categories all treated equally] --> B[Readiness scored per category] B --> C[A ranked activation order with evidence]
flowchart TD
S1[Reconcile the product catalogue] --> S2[Score each category on readiness]
S2 --> S3[Flag products outside the category structure]
S3 --> D{Ranking supported by reconciled data}
D -- yes --> DONE[Matrix and verification published]
D -- no --> S1
Stand up and validate the eight-role advisory council as working infrastructure before asking it to produce strategy.
Strategy generated by an unvalidated multi-role system cannot be trusted, and a council that answers confidently without evidence is worse than no council at all.
Recommendations arrive with the appearance of authority and no grounding behind them, and real money is spent against them.
A validated advisory layer that abstains honestly when it lacks evidence, which is exactly the behaviour needed before pointing it at a real account.
Delivered. All eight roles plus the neutral orchestrator validated, every test passing, and every role returned an honest abstention rather than fabricating a recommendation. Infrastructure only. Producing actual strategy content is what the strategy pack asks of it.
flowchart LR A[Untested multi-role advisor] --> B[Validated with honest abstention behaviour] B --> C[Safe to point at a real account]
flowchart TD
S1[Define the eight roles and the neutral orchestrator] --> S2[Build the test suite]
S2 --> S3[Run every role against evidence-free prompts]
S3 --> D{Roles abstain rather than fabricate}
D -- yes --> DONE[Readiness receipt recorded]
D -- no --> S2
Dependencies encode this. Everything in stream 3 is deliberately unblocked, because the plan is to deliver the advertising work while the website review is still open.
Turn the client's comments on the 90-page review mirror into a triaged fix list, apply the fixes on staging, and produce a release candidate that is ready for a go-live decision.
The rebuilt site exists but nothing has been formally closed. Feedback arrives as free-form comments across ninety pages, and without a triage step those comments become an unbounded stream of small changes that never resolves into a launchable state.
Measurement, the SEO transition, the price experiment and the entire ad relaunch are all gated on go-live. The site sits finished but unlaunched while paid spend keeps running against the old production site, whose forms have already been shown to produce nothing.
One list, one pass, one release candidate. Every comment is answered item by item, and the team gets an unambiguous statement of what ships instead of an open-ended revision loop.
Every comment appears in a feedback receipt classifying it as a binding choice, a suggestion, a rejection or an uncertainty, then carries a disposition. Fixed items are verified live on staging and a release-candidate summary lists what changed and what is knowingly deferred. Nothing is silently dropped. 80/20. Triage every comment and fix only the launch blockers, deferring the rest with a written reason.
flowchart LR A[Ninety pages of open comments] --> B[Classified and triaged fix list] B --> C[Fixes applied on staging] C --> D[Release candidate ready for a go-live decision]
flowchart TD
S1[Fetch the comment feed and save the raw JSON] --> S2[Classify each comment and set a disposition]
S2 --> S3[Apply launch-blocking fixes on staging only]
S3 --> S4[Re-verify each fixed page live]
S4 --> D{Every comment triaged and blockers verified}
D -- yes --> DONE[Release candidate summary written]
D -- no --> S3
Design, implement on staging and prove end to end a measurement stack that records the two conversions agreed on the call, a form submission that lands in the CRM and a phone call above the agreed duration, and feeds them back to both platforms as the optimization signal.
There is no analytics, no tag manager and no pixel on the new site, and on the current one the lead event appears to fire when a form is started rather than submitted. Both platforms are being optimized toward an event that does not correspond to a real enquiry.
The algorithms keep learning from the wrong signal, so spend chases people who open a form and leave. Roughly a month of spend could not be attributed to any lead at all, and the trial is being judged on numbers that cannot answer whether the money worked.
Every click becomes traceable to a named enquiry or a real phone call, and both platforms start optimizing toward buyers instead of browsers. The end-of-August evaluation becomes a factual conversation.
The funnel ladder is defined from campaign click through to order, each stage carrying its business definition, source, verification test, owner and freshness boundary. Submission events fire only on confirmed success. Call tracking captures duration and is proven against a staging test number, never the live line. Every event passes a positive and a negative control with a platform read-back. 80/20. The submission event and the corrected lead event, each with both controls, proven by one traced test enquiry.
flowchart LR A[Optimizing on form starts] --> B[Events bound to real submissions and qualified calls] B --> C[Conversions fed back from the CRM] C --> D[Platforms learn from buyers not browsers]
flowchart TD
S1[Ground in current staging markup and the report findings] --> S2[Define the funnel ladder stage by stage]
S2 --> S3[Install the container and bind submission events]
S3 --> S4[Add call tracking on a staging test number]
S4 --> S5[Remap the lead event and configure CRM import]
S5 --> S6[Run positive and negative controls per event]
S6 --> D{One enquiry traced correctly at every stage}
D -- yes --> DONE[Stack proven and container exported]
D -- no --> S3
Prove that a submitted enquiry actually arrives, in the inbox and in the CRM, from every form and every page type on the new site, and repair every path that does not deliver.
In the last thirty days every real lead arrived through a single legacy enquiry form. The four forms on the current site produced no real submissions at all, which points at a delivery failure rather than a demand failure. The staging audit confirms that none of the seventeen category pages contains a form, and their only call to action is an anchor back to the homepage. Category pages are the natural destination for category-led advertising, so as built they send paid traffic on a detour before it can convert.
The site relaunches with the same silent failure and paid traffic converts into nothing. Every optimization decision downstream is then made on a conversion count that is missing its real submissions.
Confidence that a click which becomes an enquiry becomes a row somebody can call. Without this, no measurement work and no ad strategy has anything true underneath it.
Every form on the new site is submitted with a traceable test value and confirmed to arrive in both the inbox and the CRM. Deliverability is checked, the production failure mechanism is written down, and a recommendation is delivered on whether category pages need their own form. 80/20. Every new-site form proven to deliver, plus the written production diagnosis.
flowchart LR A[Forms that may silently fail] --> B[Every path submitted and traced] B --> C[Failures repaired and re-verified] C --> D[An enquiry reliably becomes a callable row]
flowchart TD
S1[Enumerate every form by page type] --> S2[Submit a uniquely traceable test value through each]
S2 --> S3[Verify arrival in the inbox and the CRM]
S3 --> S4[Diagnose the production failure read-only]
S4 --> S5[Repair failing paths on staging]
S5 --> D{Every path verified from click to CRM row}
D -- yes --> DONE[Delivery matrix and diagnosis published]
D -- no --> S5
Deliver the improved lead-register structure promised on the last call, with source and campaign columns so every lead can be attributed and status columns so real buyers can be fed back to the platforms.
The register records leads but not where they came from or what happened to them, so spend cannot be attributed and the client's own judgement about which leads were serious never reaches the ad platforms.
Attribution stays impossible, the platforms keep optimizing on raw volume including the price-testers the client explicitly wants filtered out, and the evaluation has no cost per qualified lead to evaluate.
A register that answers which channel produced buyers rather than how many people filled something in, and that doubles as the conversion feed for import back into the platforms.
The upgraded sheet carries source, campaign, landing page, status and qualification columns, new submissions populate source and campaign automatically, the client has a one-page guide to the status values, and the shape supports offline conversion import. 80/20. Source and status columns live and auto-populating on new submissions.
flowchart LR A[A list of names with no origin] --> B[Source and campaign captured automatically] B --> C[Client marks qualification] C --> D[Qualified buyers imported back as conversions]
flowchart TD
S1[Read the current register structure] --> S2[Design the column set]
S2 --> S3[Automate source and campaign capture on submission]
S3 --> S4[Write the status guide for the client]
S4 --> D{A test submission lands fully attributed}
D -- yes --> DONE[Register upgraded and documented]
D -- no --> S3
Deliver the strategy the client is waiting on, covering campaign architecture per category, the media split, the agreed geography and audience, creative direction, and the sequence in which categories get activated.
This is the headline deliverable the previous plan never produced, because it was gated behind two tasks that were never authorized. The council exists and the category ranking exists, and nothing has drawn them together into something the client can approve.
The site launches and the ad relaunch is improvised, repeating the current pattern where a single undifferentiated campaign absorbs the entire budget with no category-level learning.
The client gets a plan she can say yes or no to, with a stated first category, a stated split and a stated reason for both, which is exactly the early clarity she asked for.
Evidence-labelled buyer-funnel hypotheses are written first and the ones worth building are chosen from that set. Each chosen buyer owns a complete unit of ad promise, landing page, primary action, form and follow-up. On top of that sit the architecture, the split, the geography and audience, creative direction, landing mapping and the activation sequence. Every performance claim carries its evidence and every abstention is reported as one. 80/20. The hypothesis set plus architecture, split and activation sequence for the top-ranked category.
flowchart LR A[One campaign absorbing the whole budget] --> B[Buyer hypotheses tested against evidence] B --> C[Per-category architecture and split] C --> D[A sequenced activation the client can approve]
flowchart TD
S1[Ground in the readiness matrix and the report] --> S2[Write evidence-labelled buyer-funnel hypotheses]
S2 --> S3[Convene the council and record abstentions honestly]
S3 --> S4[Specify the whole unit per chosen buyer]
S4 --> S5[Draft architecture, split, geography, creative and landing mapping]
S5 --> D{Every claim carries evidence and a first category is named}
D -- yes --> DONE[Strategy pack complete]
D -- no --> S3
Fix the price-list button that already exists but points at the wrong document, make it measurable, then run the three price-presentation variants one after another at roughly two hundred visitors each.
Price is the first question every prospect asks and the client will not maintain per-product prices. The staging audit found the price-list button already built on all sixty product pages, but every one links to a supplier-hosted English, euro-denominated price list dated April 2025, which does not cover several of the products showing it. It is hosted off-site, so the click cannot be tracked, which means variant one of the agreed experiment cannot be measured as it stands.
The first thing a price-sensitive visitor clicks hands them a foreign-language price list in the wrong currency, sometimes for products it does not cover. The experiment the client agreed to cannot produce a result, because its key interaction is invisible to measurement.
A correct, self-hosted, trackable price list, and then a decided answer on price presentation backed by real visitor behaviour rather than opinion.
A correct price list is hosted on the client's own domain and covers every product showing the button, or the button is scoped to the products it covers. The click is tracked as an event. The three variants then run sequentially with only that variable changing, roughly two hundred visitors per variant, and a written result names the winner. 80/20. The correct price list hosted and its click tracked, so variant one is measurable.
flowchart LR A[Wrong price list, untrackable click] --> B[Correct list hosted and the click measured] B --> C[Three variants run one at a time] C --> D[A decided price presentation]
flowchart TD
S1[Establish what the current supplier list actually covers] --> S2[Host a correct list or scope the button to what it covers]
S2 --> S3[Track the price-list click as an event]
S3 --> S4[Run variant one to roughly two hundred visitors]
S4 --> S5[Run variant two, then variant three]
S5 --> D{Only the price variable changed between periods}
D -- yes --> DONE[Winner named with measured rates]
D -- no --> S4
Produce and verify the transition package, covering a complete old-to-new redirect map, retention of the article that currently ranks and produces organic leads, a search console property, and a verification pass after cutover.
The old site is being replaced rather than migrated page by page. Without a redirect map the existing rankings and the organic leads they produce disappear at cutover, and there is currently no search console property to even observe the damage. The staging audit adds a second gap, since the new site serves no XML sitemap at all and its robots file does not declare one.
Organic enquiries stop arriving, the loss is invisible because nobody is measuring organic, and it gets misread as a paid-performance problem during the end-of-August evaluation.
Rankings and organic leads survive the launch, and the team can prove they did rather than hope so.
Every old URL is mapped to a destination or explicitly dropped, the ranking buying-guide article is present on the new site, a working XML sitemap is served and declared in the robots file, the search console property is verified and collecting, noindex comes off only at cutover, and a post-cutover crawl confirms no unintended dead ends or redirect chains. 80/20. The redirect map verified against a full crawl of the old site, plus the sitemap and the property.
flowchart LR A[Old site replaced wholesale] --> B[Every URL mapped or explicitly dropped] B --> C[Rankings and organic leads carried over] C --> D[Verified after cutover, not assumed]
flowchart TD
S1[Crawl production and list every indexable URL] --> S2[Map each to a destination or record a drop decision]
S2 --> S3[Confirm the ranking article exists on the new site]
S3 --> S4[Restore the XML sitemap and declare it in robots]
S4 --> S5[Verify the search console property]
S5 --> S6[Stage the noindex removal as a cutover step]
S6 --> D{Post-cutover crawl finds no unintended dead ends}
D -- yes --> DONE[Transition package verified]
D -- no --> S2
Package the per-page keyword research already reflected in the site copy into a readable table the client can review, veto and add to, as promised on the last call.
The client raised search visibility herself, was told the research exists, and has not seen it. She has product and market knowledge the research cannot have, and her own brainstorm was agreed as an input to it.
A promise made on a recorded call goes unkept, and the site launches on keyword choices the person who knows the buyers best never had a chance to correct.
Her category knowledge gets folded into the copy before launch rather than after, and a small promise is visibly kept during the month she is being asked to judge the work on.
A per-page table lists the target keyword, supporting terms, volume where known and where each is used on the site, in a format she can comment on directly, with an empty column for her own additions. 80/20. The table for the main pages and the top-ranked category.
flowchart LR A[Research the client has never seen] --> B[A readable per-page table] B --> C[Her additions and vetoes folded into the copy]
flowchart TD
S1[Collect the existing per-page research] --> S2[Build the table of page, keyword, terms and placement]
S2 --> S3[Add an empty column for her additions]
S3 --> D{Every main page and category covered}
D -- yes --> DONE[Comment-ready table packaged]
D -- no --> S2
Determine whether a carousel showing different product families can deep-link each card to its own product page, and if it cannot, build the fallback the client already accepted, a dedicated collection landing page per carousel ad.
The client's concept for the relaunch depends on one carousel mixing billiard, shuffleboard, outdoor ping-pong, foosball and dining tables, each card leading to its own page. Whether the platform supports per-card destinations in that configuration was left explicitly unverified on the call.
The relaunch creative is designed around an assumption. If it turns out to be wrong at build time the launch either slips or ships a carousel where every card lands on the same page, which wastes the format's main advantage.
A settled answer before creative production starts, and a working pattern either way, since a purpose-built collection page is arguably a better landing experience than a generic category page.
The behaviour is verified against current official documentation and confirmed in a draft ad that is never published, with a written recommendation. If per-card destinations are unavailable, a mixed-product collection landing page template exists on staging. 80/20. The documented answer plus the recommendation, without the template.
flowchart LR A[Creative built on an unverified assumption] --> B[Behaviour checked in a draft ad] B --> C[Per-card links, or a collection landing page] C --> D[Relaunch creative on a settled pattern]
flowchart TD
S1[Check current official documentation] --> S2[Build a draft carousel without publishing it]
S2 --> D{Per-card destinations work in this configuration}
D -- yes --> R1[Recommend per-card deep links]
D -- no --> R2[Build the collection landing page template]
R2 --> DONE[Recommendation written with evidence]
R1 --> DONE
Check every product with colour options that its swatches carry the manufacturer's exact codes and that clicking a swatch shows the product in that colour, then repair the ones that do not.
The colour-switching mechanism is built but unverified across the catalogue, and the staging audit found three products whose swatches carry empty image references, so clicking them does nothing. A customer has already been unhappy with a delivered shade, which makes colour accuracy a complaint risk rather than a cosmetic detail.
A customer orders from a swatch that does not match what arrives. On a premium-priced item that is a return, a refund and a reputational problem, and it is the specific failure the client named on the call.
The swatch becomes trustworthy, which removes the highest-consequence content defect on the site before it carries paid traffic.
Every colour-bearing product is audited, each swatch confirmed to carry a working image reference and the manufacturer's exact code, the known empty references repaired, and an audit table published listing every product checked with its result. 80/20. The known broken products fixed and the top-ranked category fully audited.
flowchart LR A[Swatches built but unverified] --> B[Every colour-bearing product audited] B --> C[Broken references repaired] C --> D[What the customer clicks is what arrives]
flowchart TD
S1[Enumerate products with colour options] --> S2[Extract each swatch code and image reference]
S2 --> S3[Flag empty references and mismatched codes]
S3 --> S4[Repair on staging and re-verify the click behaviour]
S4 --> D{Every colour-bearing product resolves correctly}
D -- yes --> DONE[Audit table published]
D -- no --> S4
Fix the structural gaps the staging audit found and finish applying the content decisions that are not yet uniform: the empty billiard category, the two other empty categories, the two orphaned products, the leftover sample post, and surviving Title Case headings.
The audit found the billiard category listing zero products. That is the category ranked most ready for paid traffic, so the single most important landing destination on the site is currently empty. Two further categories are empty, two products sit outside the category structure entirely, the default sample post is still published, and Title Case headings remain even though sentence case was decided. The merged navigation and the interior category imagery are already in place, so they are not part of this task.
Paid traffic gets sent to a category page that shows nothing, which is worse than not advertising that category at all. Orphaned products cannot be reached by browsing or internal linking, the sample post is a visible sign of an unfinished build, and the client re-reports a decision she already made.
The site's most valuable landing page actually lists the products it is meant to sell, and the release candidate reflects what was agreed rather than resurfacing settled items.
The billiard category lists its products, no category is empty or is documented as deliberately empty, both orphaned products appear in their listing, the sample post is removed, no Title Case headings remain, and a before-and-after list documents each change verified live. 80/20. The billiard category populated and both orphaned products placed.
flowchart LR A[Top category page listing nothing] --> B[Products assigned and listings repaired] B --> C[Headings and leftovers swept] C --> D[Every landing page ready to receive traffic]
flowchart TD
S1[Diagnose why the billiard category lists zero products] --> S2[Repair the assignment or the listing query]
S2 --> S3[Resolve or document the other empty categories]
S3 --> S4[Assign the two orphaned products and verify they list]
S4 --> S5[Remove the leftover sample post and sweep Title Case headings]
S5 --> D{Every change verified live on staging}
D -- yes --> DONE[Before-and-after list published]
D -- no --> S2
Take the ergonomic chair pages that already exist on staging from present to sellable: a price reference that applies to them, positioning that answers the comparison-shopping objection, and the supplied marketing materials actually used.
The staging audit confirms the chairs are on the site, so the content-creation part is done. What is missing is the commercial layer. The pages carry the same supplier price-list button as the game tables, which does not cover them at all, and the buyer and objection profile for a desk chair is completely different from a billiard table, so the line cannot inherit the existing positioning. The client supplied a full marketing pack for exactly this purpose.
A line the client identified as an opportunity, and personally supplied materials for, sits on the site in a state where a serious buyer cannot get a price and is given no reason to buy here rather than from a cheaper comparison-shopping result.
A second revenue line with a different buyer becomes genuinely sellable at launch, which spreads the account's dependence on a small number of high-consideration game-table sales.
Both chair pages carry a price reference that applies to them, the supplied imagery and video are used, positioning copy answers the comparison-shopping objection using delivery, trial and local support, and the line is complete and reachable from its category. 80/20. The wrong price reference resolved and the positioning copy written.
Three readings below. The sources leave the intent open, so each chart shows a different reading of what this task would actually be. Pick one in your comment.
flowchart LR A1[If the goal was finishing the pages] --> B1[Correct price reference and supplied media added] B1 --> C1[Line complete and quotable at launch]
flowchart LR A2[If the goal was answering the price comparison] --> B2[Positioning built on delivery, trial and local support] B2 --> C2[A reason to buy here rather than cheaper elsewhere]
flowchart LR A3[If the goal was a separate revenue line] --> B3[Pages plus its own audience and campaigns] B3 --> C3[A second funnel independent of the game tables]
Turn the Diagonal billiard brand from a single standalone page into purchasable products in the catalogue.
The staging audit found Diagonal on the site as a brand page only, with no products behind it. A visitor who arrives interested in Diagonal has nothing to look at and nothing to request a quote for, which is the same as the brand not being there.
A newly permitted range that already has a page pointing at it stays unbuyable, in the category ranked most ready for paid traffic, at exactly the moment the catalogue is being rebuilt anyway.
A wider billiard range in the highest-priority category, and a brand page that leads somewhere instead of dead-ending.
Not expanded — the sources confirm only that permission was granted and that a brand page was built. Which products, from what data source, with what imagery rights, and whether a supply arrangement exists are all unstated, so any sequence drawn here would be invented rather than sourced.
Bring the release candidate, the proven measurement stack, the verified lead paths, the strategy pack and the transition package into one approval document, with a build-ready campaign specification behind it.
Launch involves a production cutover and new spend at the same moment. Approving those separately, out of five different artifacts, is how one of them gets skipped.
The launch happens piecemeal. Something goes live before its dependency is proven, and the failure later surfaces as unexplained performance rather than as a known gap somebody accepted.
One document, one approval, one clean go-live, and a campaign specification detailed enough that the build is mechanical once the client says yes.
The approval pack lists every precondition with its evidence and verified state, a change manifest names everything that will change at cutover with a preview, the runbook includes the rollback step, a post-cutover read-back checklist verifies live status, destinations and events rather than assuming them, and there is one explicit place where approval is recorded before anything executes. 80/20. The approval pack with precondition evidence, the change manifest and the runbook.
flowchart LR A[Five separate artifacts to approve] --> B[One pack with evidence per precondition] B --> C[A single approval point] C --> D[Clean cutover with a rollback path]
flowchart TD
S1[Collect the evidence artifact from each precondition] --> S2[Write the change manifest with previews]
S2 --> S3[Write the cutover runbook including rollback]
S3 --> S4[Write the post-cutover read-back checklist]
S4 --> S5[Assemble the pack with one approval point]
S5 --> D{Every precondition carries verified evidence}
D -- yes --> DONE[Pack ready, nothing executed]
D -- no --> S1
Stand up the shared dashboard and the written update cadence agreed at onboarding, and prepare the end-of-August evaluation of the trial against the conversion definitions now in place.
The client agreed to a trial period with an evaluation at the end of August, and asked directly to be kept informed rather than left without news between updates. Right now there is no dashboard she can open and no fixed rhythm of written updates.
The trial reaches its evaluation with no agreed scoreboard, so the conversation becomes an argument about which numbers count rather than a decision based on them, and the reassurance she explicitly asked for never gets delivered.
She can see the account herself between updates, and the evaluation happens against metrics both sides agreed to in advance.
A shared dashboard covers spend, real submitted leads, qualified calls, cost per qualified lead and organic traffic. A written update template exists on the agreed cadence. The evaluation document is drafted with its metric definitions stated up front and every figure carrying its source. 80/20. The dashboard live with the agreed metrics, and the evaluation definitions written down.
flowchart LR A[No shared scoreboard] --> B[Dashboard on the agreed metrics] B --> C[Written updates on a fixed rhythm] C --> D[An evaluation both sides can read the same way]
flowchart TD
S1[Build the dashboard from the measurement stack and the register] --> S2[Include spend, leads, qualified calls and organic]
S2 --> S3[Write the recurring update template]
S3 --> S4[Draft the evaluation with stated metric definitions]
S4 --> D{Every figure traceable to a source}
D -- yes --> DONE[Reporting rhythm in place]
D -- no --> S1
Use blog content as the top of a funnel that remarkets readers and closes them with a product carousel.
Every current campaign asks for a high-consideration purchase from a cold visitor in a single step. A content-first entry point would let interest build before the enquiry is asked for.
The account stays entirely dependent on cold traffic converting immediately, which is the hardest possible ask for a premium considered purchase.
A cheaper entry point into the funnel and a warm audience to close, built on content that already exists on the site.
Not expanded — the client proposed this once at onboarding and it has not been revisited on any call since. There is no agreed content inventory, no remarketing audience size, and no sign it survived into the current scope.
Turn the strategy pack into one reviewable Hungarian presentation the client can read and comment on while she is still reviewing the website, so the work lands as a finished deliverable rather than as an internal document.
The strategy pack is written for the team. A dense internal document is the wrong artifact for someone who has asked to be kept clearly informed during a trial period. There is also a deliberate scheduling reason, since her website review and this delivery run in parallel and this must stand on its own without waiting for her comments.
The most substantial work of the engagement stays invisible to the person paying for it during exactly the month agreed as the trial. Good work that cannot be seen reads as no work at all.
She sees one connected story covering what is happening now, what is proposed, what it should produce and how it will be measured, and can react in writing on the same page. That turns the end-of-August evaluation into a continuation rather than a surprise.
A published Hungarian presentation tells one connected story rather than listing internal tasks, showing the current measured position with its evidence, the proposed category-led approach, the expected outcome and the measurement definitions. Every figure carries its source and the page carries the inline comment layer so she can respond in place. 80/20. The current position and the proposed approach for the first activated category, published and commentable.
flowchart LR A[Strategy written for the team] --> B[One connected story in the client's language] B --> C[Published and commentable in place] C --> D[Her reaction captured before the evaluation]
flowchart TD
S1[Read the strategy pack and the evidence behind each figure] --> S2[Write the narrative of position, proposal, outcome and measurement]
S2 --> S3[Build the page with the inline comment layer and publish it]
S3 --> S4[Verify live at desktop, tablet and mobile widths]
S4 --> D{Every figure traceable and no restricted detail on the page}
D -- yes --> DONE[Delivered and commentable]
D -- no --> S2