The one-number diagnosis

Strip away the cosmetics and the live marigoldhome.co site has one fault that dwarfs the rest: a buyer with a card in hand cannot complete the purchase. Every path — the hero, the services grid, the pricing section — ends at the same place: call us, or fill a contact form and wait 24–48h callback — confirm. In a category where people decide at 11pm after a long day, that wait is where the sale dies. The buyer books whoever answers first, which is almost always a gig app.

So the diagnosis is one number, and the number is zero: zero ways to see a price and book online, today, on a site whose entire competitive advantage is being a real company the apps can't be. The trust assets Marigold actually pays for — a W-2 team, background checks, the 24-hour re-clean guarantee — are real and rare, and they are buried where the booking-ready visitor never reaches them.

Read Exhibit 1 top to bottom and the order is the work order: the two critical leaks are losing booking-ready demand this week; the two high ones are leaving Marigold's structural advantage and its best revenue on the table. Fix the conversion plumbing first, then aim it.

The positioning thesis, in one line

The home cleaning you'd trust with a key — a vetted, W-2 team that shows up the same way every time, booked in 60 seconds online. The position carried sitewide. The buyer's real anxiety isn't "will it be clean" — it's "who am I letting into my home, and will it be the same person next time."

This is not a slogan; it is a moat stated out loud. The category is run by marketplaces that dispatch a different 1099 contractor each visit. Three things Marigold already does are expensive enough that the gig model structurally cannot copy them — a W-2 team it trains and stands behind, the same small named team returning to learn your home, and a 24-hour re-clean guarantee a real company can honor. The whole site's job is to make those three facts impossible to miss, and to put them inside the booking path rather than on an About page.

The 20-minute first action

You do not need the rebuild to start. The cheapest, highest-trust win is also the fastest: turn off the stale "$25 off" popup today. It fires on load with a code (SPARKLE25) that expired last season and now errors at checkout — so it interrupts your readiest buyers and manufactures distrust at the worst possible moment. A premium, trust-led brand does not lead with a discount it can't honor.

Why everything points at one product

Fixing the booking flow matters most when it is pointed at the right thing. A one-time clean and a recurring plan are not equals: a recurring household is worth roughly twelve one-time cleans a year, every year it stays. The arithmetic is the reason the entire rebuild bends toward recurring.

$3,874 ÷ $329 ≈ 11.8 one-time cleans per year of recurring value pricing & cadence — confirm. With 94% staying past 6 months — confirm, a biweekly household clears roughly $7,700 over two years. That is where the rebuild — and any future ad spend — should go.

How to read the rest of this book

This book is a working tool, not a slide deck: write in it, print it, argue with it. It is print-complete — every interactive layer is an enhancement, and a printout shows 100% of the content. Four chapters follow, each ending where the next begins.

Three reading conventions, then you're set. First, every number is tagged — measured means observed on the site, benchmark means a category norm, and estimate — confirm means a plausible figure you should check against your own books before you act on it. Second, anything we haven't verified about Marigold is gated verify before publishing — don't put it on the live site until you've confirmed it. Third, terms of art carry a dotted underline and link to the glossary, where each gets one plain breath of explanation. Start at Exhibit 2; read the rest tonight.

How to read a defect log

What follows is the current site written up the way an engineer writes up a broken build: each defect gets an ID, a severity, the evidence that it is real, and the fix. There is nothing here about colors or fonts — the site looks fine. The faults are all in what the site does to demand: where a person who arrived ready to hand you money is sent instead to a dead end. A severity of Critical means the defect is losing booking-ready visitors today; High means it is leaving Marigold's structural advantage on the table; Medium means it actively erodes trust or compounds the rest.

Exhibit 4 is the whole log on one screen — the six defects, sorted by severity, each with its fix. The sections after it walk the two critical leaks and the two high ones in the detail they earn. Read the exhibit first; the prose only adds the evidence.

D1 · The buyer cannot buy Critical

This is the defect that dwarfs the others, and it is the simplest to state: a visitor with a card in hand cannot complete the purchase. Trace any path on the live site and it ends in the same place. The hero button reads "See your price & book" — and clicking it does not show a price; it opens a contact form. The services grid, the pricing band, the closing call-to-action all funnel to the same form or a phone number. The buyer is asked to describe their home, submit, and then wait for a callback.

Notice the cruel detail: the site already shows a mock instant-price card in the hero — home size, frequency, a $149 every-2-weeks figure, a "Book this clean" button. It looks like the flow exists. It does not function; it is a static picture that, when clicked, drops the buyer into the same callback form. The site promises 60-second booking in its own copy (measured — the words "Book your first clean online in 60 seconds" are on the home page) and then cannot deliver it. That gap between promise and behavior is itself a trust cost.

The fix. A real 60-second self-serve booking flow with a live price on screen — realized in the rebuild as a dedicated booking page plus a working quote card carried on every page. No callback anywhere in the path. This is Phase 1 of the plan, and everything else in this book supports it.

D2 · "Call for a quote" pre-selects the wrong buyer Critical

D1 stops the purchase; D2 poisons who shows up to attempt it. Hidden pricing does not read as "premium" — it reads as "we'll size you up and charge what we think you'll pay." That sentence repels the exact buyer Marigold most wants: the busy professional who would happily pay $149 to never think about cleaning again, and who treats opacity as a red flag. What it attracts instead is the price-anxious bargain-hunter who calls three services to compare quotes — the lowest-value, highest-churn customer in the category.

The fix. Publish flat per-visit pricing for the common 2–3 bedroom home — $129 weekly / $149 every-2-weeks / $189 monthly — confirm against your books — with larger homes and first-time deep cleans resolving to an exact number inside the flow before any commitment. The good news: this work is already partly done in the rebuilt pricing band; the live site simply doesn't connect it to a booking.

D3 · The most expensive asset is the least visible High

Here is the defect that should sting most, because Marigold has already paid for the fix and then thrown it away. Running a W-2 payroll instead of dispatching 1099 contractors costs real money. Background-checking every cleaner costs real money. Carrying insurance and standing behind a 24-hour re-clean guarantee costs real money. These are exactly the facts a nervous "who am I letting into my house" buyer is desperate to hear — and on the live site they appear once, in a footer line, where no one mid-decision will ever scroll.

The principle the rebuild enforces: trust has to sit in the booking path, not on an About page. A reassurance the buyer reads after they've already decided is a reassurance that arrived too late. So the trust assets move to a persistent trust-bar directly under the hero — insured & bonded · W-2 team · background-checked · 24-hour re-clean guarantee — and get a dedicated "why you can book with confidence" section on the conversion page itself, where the card field is.

D4 · The site sells the cheap product the hardest High

The live services grid lists six things Marigold cleans, all the same size, all the same weight: recurring plans next to a one-time deep clean next to a move-out, as if they were equals. They are not equals. A recurring household is worth roughly twelve one-time cleans every year it stays — and at 94% staying past six months — confirm, it stays. Presenting them as peers trains demand toward the lowest-value, highest-churn product the business offers.

$3,874 ÷ $329 ≈ 11.8 one-time cleans of value per year pricing & cadence — confirm. Over two years a biweekly household clears roughly $7,700. Chapter 3 follows this arithmetic to its conclusion — fund recurring, stop funding the one-time chase — but the site-level consequence is already clear: recurring should be the spine, not one card in a grid.

The fix. A dedicated recurring landing page; the instant-price card defaulting to every-2-weeks rather than one-time; and the honest framing the reviews already support — "most Durham homes choose every 2 weeks." Make recurring the obvious default and the one-time clean the exception.

D5 and D6 · The avoidable wounds Medium

The last two defects are lower severity but worth naming because one of them is the cheapest fix in the entire engagement. D5 is a "$25 off your first clean" modal that fires on page load with a code (SPARKLE25) that expired last season and now errors at checkout. It interrupts the readiest buyers, and then — at the worst possible moment, right before the card field — the dead code manufactures distrust. A premium, trust-led brand does not lead with a discount it can't honor. Killing the popup removes harm rather than adding work, needs no developer, and is the recommended 20-minute first action from Chapter 0.

D6 is the generic "we clean homes" headline that claims no ground. On its own it is cosmetic; in context it is compounding — with no distinct promise, transparent pricing reads as a commodity price rather than a premium one, so D6 quietly makes D2 worse. The fix is the entire next chapter: stake the position only Marigold can take.

The takeaway, and what it sets up

Sort the log by severity and it reads as a work order. D1 and D2 are losing booking-ready demand this week — they are conversion plumbing, and plumbing comes first. D3 and D4 are leaving Marigold's structural advantage and its best revenue on the table — they are about aiming the fixed flow once it works. D5 and D6 are avoidable wounds — one fixed in 20 minutes, one answered by the positioning chapter. Fix the mechanics, then aim them.

Start from the buyer's real fear

Most cleaning sites answer a question the buyer isn't actually asking. They promise sparkle, spotless, five-star results — they argue about the cleaning. But a clean home is table stakes; nobody hires a service expecting a dirty one. The thing that keeps a busy household from booking is not doubt about the result. It is the older, quieter fear: a stranger with a key to my house, and a different stranger every time.

That fear is the whole game, and it is the one fear the dominant model in this category creates rather than relieves. Hold that thought — it is the hinge of the entire positioning. The site that names the real fear and answers it out loud wins the buyer before the price is ever discussed.

The thesis: trusted with a key

State it in one line and carry it everywhere:

The home cleaning you'd trust with a key — a vetted, W-2 team that shows up the same way every time, booked in 60 seconds online. The single position, carried on the hero, the trust-bar, and the service promise. Every headline on the site is a restatement of this sentence.

Read it as three promises stacked in order of the buyer's worry. Trust answers "who is coming." The same team every time answers "will it be the same person." Booked in 60 seconds answers "and can I do this right now, tonight, without a phone call" — the conversion fix from Chapter 1, folded into the promise so trust and ease arrive together. This is not a slogan painted over a generic service; it is a moat stated out loud, and the next section is the proof that Marigold can actually stand behind every word of it.

The honest moat — three facts the apps can't copy

The category is dominated by gig marketplaces that dispatch a different 1099 contractor to your door each visit. That model is brilliant at one thing — matching idle labor to demand cheaply — and structurally bad at exactly the three things Marigold's buyer is anxious about. Each of these costs Marigold real money, and that cost is the point: it is what makes the moat both hard to copy and easy to believe.

Exhibit 5 is the argument in full: the middle column is what the buyer wants to feel, the right column is why the competition cannot deliver that feeling without abandoning its own economics. Notice that these are not marketing claims to be polished — they are operational facts Marigold already pays for. The reviews already say it back in the buyer's own words: "Same two people every time. They know where everything goes, they know my dog. I stopped dreading the 'who's coming today' text" review voice — Trinity Park, every 2 weeks.

Where you sit — the 2×2 the buyer can't see today

Two axes decide this category. The horizontal axis is price transparency: does the buyer see a real price and book themselves, or do they call and wait. The vertical axis is trust and consistency: is it a real, vetted, same-every-time team, or interchangeable rotating labor. Plotting the field on those two axes shows the open quadrant — and shows that the live site puts Marigold in the wrong one.

Read Exhibit 6 as a story about an empty room. The gig apps own the bottom-right — easy booking, no trust — and they cannot climb up, because climbing means carrying W-2 cost and assigning fixed teams, which breaks their model. The old-school local cleaners own the top-left — real trust, but "call for a quote" — and most cannot move right, because building a 60-second booking flow with transparent pricing is work they have never done. The top-right quadrant is therefore not just open; it is structurally defended on both sides. The dashed arrow is the entire engagement: Marigold already has the trust (it is sitting in the top-left with the local cleaners) and only needs to move right into self-serve booking to land alone in the quadrant the buyer actually wants.

Hero candidates — verbatim, pick one and A/B the rest

The position has to land in the first words a visitor reads. Exhibit 7 gives three hero candidates that carry the "trusted with a key" thesis; all three are finished copy, not templates. Candidate 1 is the recommended default — it states the thesis in seven words and is the one already carried on the live home page hero (measured).

The three are not interchangeable; they lead with different proof. Candidate 1 leads with the metaphor — the key — which is the fastest way to make the whole anxiety legible in a glance. Candidate 2 leads with the consistency proof and names the guarantee outright; it is the strongest challenger for a buyer who has been burned by a rotating crew before. Candidate 3 leads with provenance — Renata Cole's boutique-hotel housekeeping background — and is the right test for an audience that responds to craft and standard. A/B in that order; the metaphor is the safest first bet, the consistency proof the most likely to beat it.

The service promise — verbatim

One line above the hero states the position; one block lower in the page makes the promise concrete enough to act on. This is the service promise, finished and ready to set wherever the buyer is deciding — most usefully in the trust-bar region and on the booking page, beside the card field:

Same team, same standard, every time — or we re-clean it within 24 hours, free. A real, vetted, W-2 team that learns your home, a flat price you see before you book, and no contract to walk away from. The service promise, carried into the booking path. Every clause is a moat fact restated as a guarantee to the buyer.

Take the promise clause by clause and it is the moat written as commitments. "Same team, same standard, every time" is the same-team fact. "Or we re-clean it within 24 hours, free" is the guarantee. "A real, vetted, W-2 team" is the W-2 fact. "A flat price you see before you book" is the transparency that moves Marigold into the open quadrant. "No contract to walk away from" is the final reassurance that removes the last reason to hesitate — and, not by accident, the thing that makes committing to recurring feel safe rather than binding, which is exactly where Chapter 3 takes the argument next.

The takeaway, and what it sets up

One position, said the same way everywhere. The hero states it, the trust-bar proves it, the service promise commits to it, and the moat — W-2, same team, the guarantee — makes every word survive a skeptic. Lock "trusted with a key" sitewide and the brand stops competing on a clean home (which everyone claims) and starts competing on trust delivered to a door (which almost no one can). With the position fixed, the only question left is who this promise is aimed at — which buyer earns the spend, and which one comes first. That is Chapter 3.

Why choosing is the point

A service that says yes to everyone aims its site at no one. The home page headline, the default in the price card, the photos, the order of the menu — each is a small bet on who walks in. Spread those bets evenly across five buyers and the site speaks to an average buyer who does not exist. So the work of this chapter is not to rank customers by who is nicest to serve; it is to rank them by what one of them is worth over the life of the relationship, and then to let that number, not sentiment, decide where the site and the spend point.

The unit we will use throughout is the value of a single acquired customer over the time they stay — what a marketer would call lifetime value. It is the honest way to compare a buyer who pays once with a buyer who pays every two weeks for years, and it is the lens that turns a flat list of five services into a clear order of four-plus-one. Exhibit 8 is that order on one screen; the sections after it walk each group and show the arithmetic.

Read Exhibit 8 down the "worth" column and the strategy is already visible. The recurring household and the two business channels that feed it are worth thousands a year each; the one-time clean is worth three hundred dollars and then nothing. The rest of the chapter puts each group on its own line, with an envelope — who they are in one breath — and a worked-out figure that re-multiplies so you can check the number rather than trust it.

Group 1 · Recurring busy households Primary

Envelope. Dual-income professional households in and around Durham — Trinity Park, Hope Valley, Watts-Hillandale — who will gladly pay to never think about cleaning again, and who want one reliable team rather than a different stranger every visit. This is the buyer the whole positioning was built for in Chapter 2, and the buyer every other group ultimately feeds.

The reason this group is the spine of the business is not that there are more of them. It is that one of them, once acquired, pays for years. The arithmetic is the argument:

And it does not stop at one year. With 94% of recurring households staying past six months — confirm, a biweekly home clears roughly $3,874 × 2 ≈ $7,748 — call it $7,700 over two years. Set that beside a one-time deep clean at $329 and the same acquired customer is worth $3,874 ÷ $329 ≈ 11.8 one-time cleans every year they stay — about 23 over two years. This is where the spend goes.

Group 2 · Realtors — the channel that refers Channel

Envelope. Agents who need a documented, on-the-day deep clean that makes a listing show-ready or returns a seller's deposit — and who, once they trust a vendor, reuse the same one across deal after deal. You acquire the agent once; the agent brings the jobs.

$329 × 24 ≈ $7,896, with no acquisition cost after the first job, so one loyal agent is worth roughly two recurring biweekly households ($7,896 ÷ $3,874 ≈ 2.0) deal volume — confirm per agent. The second, larger payoff is indirect: a realtor stands at the exact moment a buyer is setting up a new home — the peak "we need a cleaner" moment — and can refer that household straight into Group 1.

That dual return is why realtors are a channel, not just a customer: the direct revenue is good, the referral flow into the primary group is better, and both come from acquiring one relationship rather than one job.

Group 3 · Short-term-rental hosts — recurring in disguise Channel

Envelope. Airbnb and short-let hosts who need a reliable same-day turnover — linen reset, a photo checklist, timed to checkout — so a missed clean never costs them a review. Their whole business depends on the turnover happening, which makes them both high-frequency and extremely reluctant to switch.

$119 × 72 ≈ $8,568 per property; a two-property host is $8,568 × 2 ≈ $17,136/yr turnover frequency & price — confirm. The switching risk is the moat here: a host who changes cleaners gambles their ratings, so retention is structurally high. Functionally this is a recurring plan wearing a different label — the same compounding value as Group 1, sold to a business buyer.

Group 4 · Post-renovation cleanup — the on-ramp On-ramp

Envelope. Homeowners and contractors who need construction dust, paint flecks, and film cleared so a finished project finally looks finished. It is project-driven, not recurring — but it puts Marigold in a freshly renovated home at the precise moment the owner is most likely to want it kept that way.

The job itself is a larger-scope deep clean at $449. Its real value is the on-ramp: roughly 15–20% convert to a standing plan — confirm once they see a finished-clean home, so each job is worth $449 now plus an expected 0.18 × $3,874 ≈ $697 of recurring — about $1,146 blended. Worth accepting and serving well; not worth a dedicated acquisition push, because the dollar still converts harder aimed straight at Group 1.

Group 5 · One-time deep cleans Not pursued

Envelope. People who want a single seasonal or pre-event deep clean and have no intention of returning. Useful cash, lowest value, highest churn — and, on the old site, the product the menu sold the hardest.

$329 once, with no second visit assumed, is worth roughly 4% of a two-year biweekly household ($329 ÷ $7,748 ≈ 4.2%) conversion negligible — confirm. The booking path stays open — it pays for itself and seeds a few conversions into Group 1 — but it earns zero proactive acquisition spend.

The takeaway, and what it sets up

Put the five groups in one sentence: fund recurring households, feed them with realtors and short-term-rental hosts, accept post-renovation as an on-ramp, and stop funding the one-time chase. The arithmetic forced that order — an acquired recurring household is worth about 11.8 one-time cleans a year, the channels each refer or repeat into that same pool, and the one-time clean returns roughly 4% of a two-year recurring relationship. Nothing here is a guess about who is pleasant to serve; it is a ranking by what a customer is worth. Chapter 4 takes this ranking and the position from Chapter 2 and builds the site that aims at them — the rebuilt pages and the 90-day plan that ships them.

What the rebuilt site does

The fastest way to understand the plan is to see the thing it ships. The rebuilt site is live as a working preview — open it, click through it, and you are reading the same five pages a Durham buyer would. It is not a mockup of a site; it is the site, annotated. Exhibit 9 is the map of what each page is for, and every claim below is something you can confirm by clicking.

The whole rebuild turns on one inversion. The old site described a cleaning service and asked the buyer to call. The rebuilt site lets the buyer book, and uses every other page to remove the reasons a careful person would hesitate before they do. The hero on every page carries the same instant-price card, defaulting to the every-2-weeks plan — so the high-value product is the one the buyer is one tap from buying, on whatever page they land.

Three things are worth pointing out before you click, because they are the spine of the rebuild and easy to walk past as cosmetics.

Second, the trust story now lives in the booking path, not on an About page nobody reaches. The trust-bar — insured & bonded, W-2 team, background- checked, 24-hour re-clean guarantee — sits under the hero on every page, and the Book page repeats it as a "why you can book with confidence" section directly above the card field. This is deliberate: the buyer's hesitation peaks at the moment of commitment, so the reassurance has to be there, not three clicks away. Third, the stale "$25 off" popup is gone — no on-load interrupt, no dead SPARKLE25 code manufacturing distrust at the card field. A premium, trust-led brand does not open with a discount it can't honor.

Open the working preview to walk all five pages with the persuasion reasoning annotated alongside each section: the rebuilt site preview → Every annotation links back to the chapter that argued for it.

The two gates the plan is built around

The 90-day plan is not a list of improvements done in order of how nice they'd be. It is sequenced behind two gates, and a phase only opens once the gate before it is passed — because a fix delivered out of order quietly wastes the fix that follows it. Get the gates right and the order of the four phases is no longer a matter of taste; it is forced.

The first is the access gate: can a buyer with a card in hand actually complete a booking, today, without waiting for a callback. Until that is true, nothing else matters — a more trustworthy site that still can't take a booking just loses better-qualified demand. The second is the credibility gate: does the trust story sit inside the path the buyer takes to that booking, so the moat does the closing. Exhibit 10 shows how the two gates order the work.

The logic of Exhibit 10 is the whole argument for the order. Spending on demand (Phase 4) before the access gate is open is the most expensive mistake available — it pays to drive buyers to a site that asks them to call and wait. Surfacing trust (Phase 2) before booking works (Phase 1) reassures a buyer who then can't act on the reassurance. So the plumbing comes first, then the trust that sits inside it, then the aim at recurring, and only then the demand — each phase passing the gate that makes the next one worth doing.

The four phases, in order

Each phase is roughly three weeks, owner-runnable with a developer for the booking build, and measured by one number that tells you the gate has actually passed — not "the page shipped" but "the page changed the behavior it was meant to." Exhibit 11 is the four phases on one screen; the full plan, with task-level owners, minutes, and CMS click-paths, is the companion document.

Phase 1 — Booking + instant price. This is the access gate, and it is the only phase that is non-negotiably first. Ship the 60-second flow with a real price on screen and the flat pricing tables, and make the primary call to action everywhere "See your price & book." Killing the stale promo popup belongs in this same pass, not as a separate task — it actively harms the flow it sits in front of. The phase is done when a buyer who has never spoken to Marigold can reach a price and confirm a clean without a phone call.

Phase 2 — Surface the trust. Once a buyer can book, the credibility gate decides whether they do. Move the trust-bar under the hero, put a "why you can book with confidence" block inside the booking path, tell the Renata Cole hotel-grade story on the team page, and land the "trusted with a key" headline sitewide. The point is mechanical, not decorative: with trust sitting at the moment of commitment, the flow stops competing on price — the one ground where a $10-cheaper app wins.

Phase 3 — Emphasize recurring. With a trusted flow working, the aim gate points it at the product that compounds. The dedicated recurring page, the price card defaulting to every-2-weeks, the "most Durham homes choose every 2 weeks" framing, and recurring reviews surfaced first all make the standing plan the obvious default and the one-time clean the exception. The number to watch is the share of new bookings that are recurring — because at roughly eleven one-time cleans of value per year, that share is what actually grows the business.

Phase 4 — Demand generation. Only now, with a flow that takes a buyer, reassures them, and points them at recurring, is it safe to pour demand in. Local SEO for "recurring house cleaning Durham," a referral mechanic for the 94%- retention recurring base, and paced outreach to realtors and short-term-rental hosts open the two channels that feed the primary group. One-time deep cleans receive no spend — the door stays open, the dollar goes elsewhere.

The capacity-honest first moves

There is one thing this plan refuses to pretend, and it changes how Phase 4 is run. Marigold is a real company with roughly 14 W-2 cleaners — confirm, matched into small named teams. That is the moat — and it is also a ceiling. Unlike a gig marketplace that conjures supply on demand, Marigold cannot absorb a flood of new recurring households overnight without breaking the very same-team consistency it sells. A plan that floored the gas on demand would, if it worked, manufacture the one failure the brand cannot survive: a new client matched to a rushed, rotating, or apologetic team.

So the first moves are paced to the team, not to the funnel. The honest sequence is to fix conversion (Phases 1–3) on the existing demand first — that lifts revenue per visitor with zero added load on the cleaners — and to open Phase 4 demand in a metered way, sized to the recurring slots the team can actually staff well. Exhibit 12 lays out what capacity-honest looks like in practice.

Where to start

The plan is a working tool, not a slide. The full version lists every task with an owner, an estimate in minutes, a target day, and the exact CMS click-path to do it — including a capacity budget so a busy owner can see what fits this week. Two things to do now: walk the preview so you can see what Phase 1 builds, then open the plan and start the first task.

That is the whole engagement in one arc: Chapter 1 found six named leaks, Chapter 2 staked the one position the apps can't take, Chapter 3 ranked the customers and funded the recurring household, and this chapter built the site and the order to ship it — a flow gated on access, then credibility, then aimed at recurring, with demand paced honestly to a real team. Approve the four phases, start Phase 1, and the site stops describing a cleaning service and starts booking one.

How to use this list

Type in the box to filter. Search "trust" and you'll see every entry that touches it; clear the box to see them all again. Each entry is anchored, so the dotted links inside the chapters jump you straight to the right place and back. Terms are ordered the way they first matter in the argument — buyer behaviour, then the things Marigold sells, then the mechanics of the rebuilt site — not alphabetically, because the book is read as a story, not looked up like a dictionary.