eXp Luxury and LuxVT, member experience and revenue audit

Where eXp Luxury loses members

Eighteen leaks, found by walking the member product and reading the billing ledger, and confirmed against the behavioural data the client already collects. Original count the member product on two accounts and reading the full billing ledger of 5,309 subscriptions. One of them accounts for 50 per cent of every dollar this business has ever lost. Internal draft, 28 September 2026.

1. What this looked at, and what a member has to believe

The goal is more members, kept longer. We walked the member product on both platforms as a member does, using two accounts the client provided, and read the billing ledger directly. Every figure here is sourced to a live test on a stated date or to a full ledger read on 3 or 28 September 2026.

We covered the eXp Luxury and LuxVT membership core. The billing organisation also carries Sports and Entertainment, Land and Ranch and New Homes, 57 plans in total. Those three are named as out of scope rather than covered thinly, and they sit in the risk register in chapter 11.

Two products, one platform, one ledger

LuxVT is the marketing engine. eXp Luxury is the membership programme its parent company wraps around that engine, and LuxVT is also sold on its own. They share one codebase and one billing system, so every finding below is tagged by where it lives, and the money findings belong to both because the ledger does not separate them.

Features differ between the two and some are visibly under construction, for example a Microsite Creator marked Beta and MLS listings being fed into activation. Where a finding sits on something clearly still being built, we say so rather than scoring it.

2. The shape of the loss

More annual contract value has been cancelled in the lifetime of this business than is currently live, and the book is shedding close to half its own value every year.

1,045 subscriptions were cancelled in the twelve months to September 2026, carrying USD 1,879,838 of annual contract value between them. That is 48.7 per cent of the entire live book, gone in a year, at an average of USD 156,653 a month.

Source: the client's own Zoho Billing subscription records, read 28 September 2026. Live and cancelled contract value are the sum of contract amount over live and cancelled subscriptions; the 70 day figure is the combined value of the 988 cancellations falling between day 330 and 400. Method and date range for each are itemised in chapter 3.

Date range: all time (every cancellation ever recorded), read from the full ledger on 28 September 2026.

When cancellations happen, measured from the day each member joined. 2,052 cancellations carrying both dates, out of 2,222 total. Billing ledger, all time, read 28 September 2026.

This is not a gradual decline. It is two spikes. 270 members leave inside the first month and 988 leave at the first renewal. Between them they are 61.3 per cent of everyone who has ever left. The middle of the first year is comparatively quiet, which tells us the product is not annoying people out of the door. It is failing to give them a reason to stay in.

Net subscription movement by month, signups minus cancellations. Billing ledger, 28 September 2026.

The figures above are read directly from the client's own Zoho Billing system on 28 September 2026. Chapter 3 reproduces every source screen and states the method and date range behind each number.

3. How to read these numbers: proof and dates

Every figure in this report is one of two kinds. It is either shown on a screen in the client's own systems, in which case the screen is here, or it is counted from the client's own records, in which case the exact method and date range are stated. Nothing is estimated, and nothing is an industry benchmark. Where a number is a projection rather than a measurement, it is labelled as one and kept separate.

2A. Figures shown directly on the platform

These are read straight off the client's Zoho Billing screens, captured on 28 September 2026. The screens are reproduced in this report so any figure can be checked against its source.

2B. Figures counted from the client's records

The renewal cliff, the survival curve and the revenue-lost totals are not single tiles on a screen. They are counted from the client's own subscription records, all 5,309 of them, read through the billing system's own data interface on 28 September 2026. The method and window for each is stated here and repeated on each finding. The platform's native Revenue Churn report above corroborates the monthly shape of these figures.

2C. Why the date ranges differ, and how to line them up

Three windows appear in this report, each chosen for a reason. When two figures look like they should match but do not, this is almost always why.

4. Who this happens to

5. What the briefs said, and what we found

Two documents told us what was wrong before we looked: the July 2026 UX framework with its signal scan, and the Diagnosis and Engagement Plan. We tested every assertion in both. They are directionally right more often than not, two of their findings have already been fixed since they were written, and the largest one is confirmed and considerably bigger than stated.

6. The leaks

2

There are two ways to get a listing live, and the manual one dead-ends

Shared platformP0Product team

4 of 4manual listings stuck at Draft

Corrected since the first draft. An earlier version said there was no way to activate a listing at all. That was too broad. There are two paths. Listings fed from the MLS appear on the Activate Listings screen with a working Activate this listing link, and that path works. The manual path, for a listing an agent builds by hand, is the one that dead-ends: it saves as Draft with no route forward. Both audit accounts held only manual drafts, which is why the Activate screen showed nothing to activate.

✗ The manual path

eXp LUXURY

Create Listing

Submit

→

eXp LUXURY

0

Doubt

→

eXp LUXURY

→

eXp LUXURY

!

No listing found for activation

Lost here

Why members drop out here

The listing page never says what Draft means or what is missing.

The Activate screen contradicts itself: a quota is in use, yet nothing is available to activate.

Returning to finish always restarts at tab one. Later tabs cannot be linked to directly.

Nothing names Tour Submission as the step that completes a listing.

The harm: this is the mechanism behind leak 1. A member who never gets a listing live arrives at renewal with nothing to show for the year, and 48.1 per cent of all cancellations happen in that window.

Money spentNonethese are members already paying, USD 3.86M of live contract value

Money lostNot countedwe cannot yet count listings that never reached the portals

How we worked this out

Every listing on both audited accounts sits at Draft, so neither member has received the syndication the membership sells. We will not extrapolate two accounts to the whole base. One query prices this exactly: the number of live subscriptions whose listings have never reached Active Campaign, and their combined annual value. If that share resembles what these two accounts show, this is the largest priced leak in the report.

What to change

Make the manual path reach the same place the MLS path does. A Draft listing should name the unfinished tab and link to it, and offer the activate step once complete. The Activate screen should never report that nothing was found while a quota is in use; it should list the Draft listings and the single thing each one needs. A visible status ladder would let a member see where a listing stopped.

For the product team

The five statuses already exist in the Manage Listings filter: DRAFT, IN PROGRESS, ACTION NEEDED, ACTIVE CAMPAIGN, INACTIVE CAMPAIGN. The ladder is modelled in the data and never surfaced on the listing or the activation screen.

✓ After the fix

eXp LUXURY

0

→

eXp LUXURY

Finish Tour Submission

Send

→

eXp LUXURY

✓

Thank you

Success

What we could not test, and why

Completing Tour Submission writes to live client data, and the client instructed us to make no changes. We stopped at that boundary. We can evidence that every listing sits at Draft and that no screen names the missing step. We cannot evidence what Tour Submission itself requires. A sandbox account settles it.

9

The dashboard sells add-ons above the one action the membership exists for

eXp LuxuryP1Product team

Below the foldand collapsed, on every visit

What the order of this screen says

The largest element is a webinar replay. The second is a Concierge Services advertisement with its own call to action.

Concierge Services is a paid add on. The member is sold to before they have received what they already bought.

Launch Your Listing, Win The Listing and Master Your Branding are collapsed headings at the bottom, visually identical, none marked as the place to start.

The Onboarding card is genuinely good. It is one of three equal cards, competing with Most Popular for attention.

The harm: the information hierarchy inverts the business economics. The action that prevents USD 2,229,013 of renewal churn is collapsed at the bottom; the upsell is at the top. A member in their first week clicks the largest thing, and the largest thing is not the product.

Money spentNonethe traffic is members who already pay

Money lostNot countedno dashboard interaction is tracked

How we worked this out

The dashboard is the first screen on every visit, so its order is the strongest signal the product sends about what matters. We cannot price this because no dashboard interaction is instrumented. What would price it: clicks on each dashboard element, and the share of members who reach Create Listing within their first 30 days.

What to change

Give the dashboard one primary action that changes with the member state. A member with no listing sees Create your first listing. A member with a Draft sees Finish your listing, and which step. A member with a live campaign sees its results. Move Concierge Services below that, because selling an add on to a member who has not yet received the core service is the wrong order commercially as well as experientially.

12

The form that creates every listing cannot be completed reliably

Shared platformP1Product team

38fields, three required in name only

What we measured on the live form

Street Address, City and Price carry an asterisk in their label. None carries a required attribute and none carries aria-required. Required status is painted on, not programmatic.

Two fields have no label a screen reader can reach: the square footage box and its unit selector.

Twenty navigation targets measure 20 pixels tall or less, against a 24 pixel minimum.

The address fields declare an autocomplete value that is not valid, so browsers will not offer to fill them. Agents retype the same addresses by hand.

The harm: a member using a screen reader is not told which fields are mandatory and two fields are announced with no name at all. For everyone else, address autofill silently does not work on the form that stands between a member and the only thing the membership delivers.

Money spentNoneno advertising is involved

Money lostNot countedabandonment on this form is not tracked

How we worked this out

This form is the only route to the product core value, so anything that makes it harder to finish is a revenue leak as much as an accessibility one. Pricing it needs form start and form completion events, which the platform does not emit. Those events are named in the instrumentation brief in chapter 7.

What to change

Add the required attribute to the three fields that already claim it visually, and label the square footage box and its unit selector. Raise navigation targets to at least 24 pixels. Correct the autocomplete values so browsers can fill the address fields. Then remove from the member view the fields that belong to operations, listed below.

For the web team

Measured on the live form at app.expluxury.com/user/ID/tour/add, 28 September 2026. Required fields: 0 with the required attribute, 0 with aria-required, 3 with an asterisk in the label. Unlabelled controls: input[name=squareFeet], select[name=unit_of_measurement]. Autocomplete tokens read "property address-line1", "property address-level2", "property postal-code"; "property" is not a valid section token so the whole value is discarded. Operations fields visible to members: orderOwner, productionNotes, tourSiteBcc, tourReseller, feedsListingId, nextHomeListingId. The last is a competitor brokerage identifier. Criteria touched: 1.3.1, 1.3.5, 2.5.8, 3.3.2, 4.1.2.

RBI rulesThis is a screening pass on the conversion journey, not a conformance audit. We report what we measured, on which pages, on which date.

SEBI rulesSafe. SEBI covers investor information, not loan pages.

Dark patternsNo compliance level is claimed for the client.

13

The brochure that is meant to win listings ships with blank fields

Shared platformP0Product and marketing

Blankbeds, baths and MLS on a shareable brochure

The whole proposition is that an agent walks into a listing appointment and impresses a seller. We opened the seller facing email brochure the platform generates. The template is good. What it fills in is not.

What we found on the generated brochure

The specification block renders with Bedrooms, Baths, Half Baths and MLS number all empty. A seller receiving this sees a luxury brochure that does not say how many bedrooms the house has.

The agent identity block carried placeholder details: a stand-in name, the phone number plus one two three four five six seven eight nine zero, and a nonsense web address. Nothing stopped that brochure being previewed or shared.

There is no validation. The system will preview and let an agent share a brochure with no specs and fake contact details.

The public sample presentation at elite.luxvt.com/sample-presentation, and the member Onboarding Guide, are both Canva embeds, so neither can be tracked, versioned or counted as delivery.

And on a phone

On a 390 pixel screen the member greeting sits on top of the page heading on every authenticated page we tested.

The dead end from leak 2 is identical on mobile, with the only escape route a wrapped text link rather than a button.

The harm: the seller facing output is the product, and the product can go out broken. An agent who shares this brochure at a listing appointment shows a seller a luxury document with no bedroom count and a fake phone number. That is worse than sending nothing, and it is the exact moment the whole membership is meant to prove itself.

Money spentNoneno advertising is involved

Money lostNot countedbrochure open and share rates are not tracked

How we worked this out

We previewed the generated brochure on a test listing, so the blank fields may partly reflect an unfinished draft. But the finding stands: nothing in the system prevents a brochure with empty specs and placeholder identity from being previewed and shared. A validation gate and a check on how many live brochures currently carry blank fields would size it. The listing website and activity report on a fully activated listing remain unseen, because both audit accounts held only drafts.

What to change

Add validation: a brochure with empty required fields or placeholder contact details should not be previewable or shareable. Bring the sample presentation and onboarding guide inside the product from Canva so they can be versioned and counted. Instrument the brochure so an open or a share is recorded against the member, because a brochure the agent actually used is the strongest thing to show at renewal. Then fix the header overlap on mobile, which affects every authenticated page.

14

The scoreboard reports growth while the business shrinks

Shared billingP0Leadership

25xswing in the same tile in 25 days

The billing dashboard reports active subscriptions up 15.7 per cent year on year. That figure is accurate and it is trailing. Monthly net movement has been at or below zero since May 2026.

Month	Net	Signups minus cancellations
Mar 2026	+88
Apr 2026	+26
May 2026	-9
Jun 2026	+1
Jul 2026	-4
Aug 2026	-27
Sep 2026	-6

Net subscription movement by month, signups minus cancellations. Billing ledger, 28 September 2026. Bars are scaled to the March peak of 88.

2,371live subscriptions, down 52 in 25 days

USD 4.46Mof contract value cancelled to date

USD 3.86Mof contract value live today

Three default measures that mislead on this business

Lifetime value divides by a month to date churn rate, so it is meaningless early in a month.

The month to date churn rate is read as though it were a monthly rate.

A trailing twelve month subscriber count hides a turn that happened five months ago.

The harm: more contract value has now been cancelled in the lifetime of this business than is currently live, USD 4,458,325 against USD 3,861,237, and the scoreboard has been reporting growth throughout.

Money spentNoneno advertising is involved

Money lostNot countedwe will not price decisions that have already been taken

How we worked this out

On 3 September 2026 the lifetime value tile read USD 180,363. On 28 September 2026 the same tile on the same business read USD 7,153. Nothing about the business changed. The tile divides average revenue per user by a month to date churn rate, and that rate matured from 0.08 per cent on day three to 2.00 per cent on day twenty eight. USD 143.05 divided by 0.02 is USD 7,152.50. The figure is arithmetic, not measurement, and it reads differently every day of the month.

What to change

Make monthly net movement the headline, because it is the number that turned. Retire the lifetime value tile or recompute it on trailing twelve month churn and per plan family. Build a scorecard on four measures: net monthly movement, share of members with a live listing, collection health, and renewal rate at the cliff. Then ask what was decided while the old number was on screen.

17

The behavioural data confirms it: a third of sessions dead-click, a third bounce straight back

Shared platformP0Product team

29.73%of sessions have dead clicks

The client runs Microsoft Clarity on the product. It records what members actually do: where they click, where they give up, where the interface does not respond. We were given access to it, and it turns several of the findings above from expert judgement into measured behaviour. The window is short, the last three days, about 333 sessions, so read these as direction confirmed rather than a full baseline.

What the members are actually doing

Dead clicks, 29.73% (99 sessions). Nearly one session in three clicks something that does not respond. This is the measured version of the Draft dead end and the unresponsive controls in leaks 2 and 12.

Quick backs, 36.64% (122 sessions). More than a third of sessions hit a page and immediately go back. People are landing somewhere wrong and retreating, which is the measured version of the navigation and hierarchy problems in leaks 9 and 5.

JavaScript errors on 25.66% of sessions (247 errors). Clarity names them: a React render error, "target container is not a dom element" (32%), and "cannot read properties of null reading classList" on a dashboard-alert element (26% and 4%). These are concrete and reproducible, not vague slowness.

Outbound clicks, 59 sessions. This is leak 11 measured: members are clicking off the platform to the six other domains it links to, in nearly one session in five.

7.1% of sessions are on mobile Safari. The mobile header overlap in leak 12 is hitting real iPhone users, not a hypothetical.

Rage clicks and excessive scrolling are both low, 0.30%. So the problem is not members hammering in frustration. It is quieter than that: they click, nothing happens, and they leave.

Where they spend their time, and where it breaks

The single most-visited page on the product is My Listings, on all three brands. That is where members live, and it is where leaks 2 and 12 sit.

The eXp listing manager was hit in an error state 26 times in three days. That is the Activate Listings screen from leak 2, failing, measured.

The harm: the behavioural data and the audit agree. The places members spend the most time are the places that fail, and they fail quietly, by not responding, so a third of sessions dead-click and a third retreat. None of this shows up in a revenue report until a renewal is refused.

Money spentNoneno advertising is involved

Money lostNot countedClarity measures behaviour, not money; the loss is priced in leaks 1 and 6

How we worked this out

Source: Microsoft Clarity, LUXVT project, last 3 days to 29 September 2026, 333 sessions, 31 bot sessions excluded. Dead clicks 29.73%, quick backs 36.64%, JavaScript errors 25.53%. This is a short window on a recently installed tool, so it confirms direction rather than establishing a baseline. Extending Clarity retention, and adding funnels for the register-to-activate path, would turn this into a measured funnel.

For the web team: the named errors and event counts

JavaScript errors, share of the 247 recorded: 32.39% target container is not a dom element; 28.34% script error; 25.91% cannot read properties of null (reading classList); 4.45% document.getElementById(dashboard-alert).classList is null; 1.62% cannot read properties of null (reading settings). Smart events by session: Submit form 73, Outbound click 59, Upload 33, Log-in 28, Download 9, Subscribe 6, Search 5, Contact us 4. Devices: Chrome 73.96%, Safari 15.38%, MobileSafari 7.10%, Edge 2%. Web Vitals show LCP and INP needing improvement, CLS good.

What to change

Two things, both cheap. First, act on what Clarity already shows: find the dead-click targets on My Listings and the listing manager and make them respond, and fix the error state that is being hit 26 times in three days. Second, get more from the tool: extend its data retention, and build a Clarity funnel for register to first listing to activation, which is the one measured funnel this business could have tomorrow without any new instrumentation. It is the cheapest way to size leaks 2 and 15.

7. Why nobody could see this

Five measurement failures explain why a business that turned negative in May is still reported as growing.

• No cancellation reason exists on the record. Not empty, absent. 2,222 members have left and the business holds no record of why any of them did.
• The lifetime value tile is arithmetic, not measurement. It read USD 180,363 on 3 September and USD 7,153 on 28 September, on the same business, because its denominator matured.
• The headline subscriber count is trailing. A twelve month window cannot show a turn that happened five months ago.
• The product emits no activation events. No form start, no form completion, no listing status change, so the single most important journey in the product is invisible.
• Nobody had looked at time to cancel. The renewal cliff has been in the data the whole time. It takes one query to see it, and it reframes the entire retention problem.
• The acquisition funnel collects no data. The two sites where members sign up carry no analytics at all, so the 86 per cent signup collapse happened on surfaces that record nothing. See leak 15.

8. The four questions, answered with evidence

The engagement asks four things: why members leave, how to bring more in, how to keep them, and how to grow what they pay. Here is what the evidence says about each, and the single move that matters most. Every claim below traces to a finding above and a source: the billing ledger, or Microsoft Clarity, both read this week.

9. What to do, and what it adds up to

Every figure is annual contract value taken from the subscription record itself, not an estimate, and every one is a floor. We have not applied a recovery rate to the at risk figure because we have no data on this client past recovery. Nothing here prices trust, brand damage or regulatory exposure.

1. Build the renewal journey. A membership page, a delivery record and a notice at sixty and fourteen days. This is the only leak on the list that touches USD 2,229,013.
2. Fix the Draft dead end. It is the reason the renewal has nothing to show.
3. Turn on the cancellation reason. Configuration, not a build. Until it exists, nothing can be proved.
4. Make the first session produce one finished thing, because 13 per cent never reach month two.
5. Warn before the card fails, and give the member somewhere to fix it.
6. Reorder the dashboard so the core action outranks the add on.
7. Replace the scoreboard and check what was decided on the old lifetime value.

10. What we did not cover

Named so that absence is a decision rather than an oversight. Sports and Entertainment, Land and Ranch and New Homes, which are three of the four brands in the billing organisation. The three features locked behind the Classic tier. Any surface reached only after submitting a form, because we made no writes. The marketing funnel upstream of signup, because no analytics access was provided. And anything a member would tell us, because no member has been spoken to.

11. What could change the size of this job

Everything we found that could move a price, in the open. A named risk is a shared risk.

12. How we checked

Dates. Billing ledger read in full on 3 September and again on 28 September 2026. Product walked live on 28 September 2026.

What we used. Two member accounts provided by the client, one LuxVT Classic and one eXp Luxury Agent with a Sports and Entertainment membership. Read only access to the billing system. 5,309 subscription records and about 8,000 invoices were read through the Zoho Billing API and aggregated. Every money and subscription figure in this report traces to the Zoho Billing screens shown in chapter 2, captured live on 28 September 2026. No personal data of any member appears in this report, including inside the screenshots, which are captured from aggregate dashboard views and masked wherever a name could appear.

What we changed. Nothing. Every call was a read. We did not submit a form, dismiss a dialog, create, edit or delete anything, on the client instruction.

Corrections made during this audit. Six, all reflected above. We marked the claim that entry costs USD 950 as Not supported, and the claim that the Brand tier excludes syndication as Not testable. Both were our error: we had checked the billing catalogue and one marketing site, not the published plans page, which states both outright. We reported that the 60 plus against 80 plus contradiction had been fixed; reading all four tier pages instead of one showed it has moved rather than gone, and now runs the wrong way. And we have adjusted the first month churn figure downward after finding one time purchases recorded as subscriptions inside it. The three earlier corrections stand: We first recorded the Create Listing form as 38 fields on one flat screen; it is 38 across four tabs. We carried forward two July findings, the 60 plus against 80 plus contradiction and the placeholder alt text, which a live re-check showed have since been fixed. And we withdrew a finding entirely: an earlier version claimed 98 members were billed USD 1 to USD 6 a year with USD 168,040 at stake. Reading the billed amount on every one of those subscriptions showed the low figures are catalogue defaults and the members pay real prices. That finding is now leak 11, at a fraction of the size.

Every finding, and how it was verified

Each claim below was checked live in the client's own systems on the date shown, either by walking the product and reading the page, or by recomputing the figure from the full billing ledger. Nothing here rests on inference or a single earlier capture.

Palette sampled from the live member product on 28 September 2026: near black navigation with a muted gold accent. Severity colours are the report own and do not follow the client brand.