Your note: the results should flow as a story, open eyes, and end on "give us your money and save 8x". This is how. After the three typed fields and the four taps, the results play one reveal per click, each one a shape whose size or count is the number. The last reveal is the ask, and its N is worked out from their answers against our price list, so Coral gets 8, a small shop gets 6, a big business list gets 11, and a list cleaned last spring is told honestly it can wait. Tap the N and the maths opens.
Built: "Your results, one step at a time", right under the fields, plays these seven plus Scene E. Scene E (how often to clean) comes just after the ask, as "keep it that way", so the story peaks on the money and the habit comes second.
Seven scenes down a page, each with its own Play. The visitor scrolls and chooses. The ending is one scene among seven, and the "8 times" is easy to miss.
One stage, one change per click (or tap, or the arrow key). A dot per step shows where you are, and Back steps back one. Every change holds still until the next click, so it can be read. It always ends on the ask. (RME-115.H2: each scene now moves one pose per click instead of running its poses together.)
Scene G today: clean blocks already drop onto a year of quiet, and the count is worked out. But it is one scene among seven, it says "fits about 8 times" rather than what they save, it has no branch for a list that should wait, and every review only ever showed Coral, so it read as a fixed 8x.
The motion (press Play): the year bar fills from the left, month by month (that is the money you keep paying), then one clean-sized block drops onto it, then the next, counting 1, 2, 3 until no whole block fits. The count is N, the sliver left over is the part we do not claim. Nothing moves that is not the number.
$1,954 a year ÷ $240 a clean ≈ 8.1
One line of maths. The notes on this page still say $240, from before the price list was confirmed.
Five pictures, one per Next, drawn from their own numbers. The sum only shows as a small footnote on the last one.
A list that should wait shows one stack taller than the whole pile ("0 fit, not yet"). A list of 1,000 or fewer shows no stack to pay for. It plays at the end of the results ("How we got N, in pictures", or tap the N's "i"), and here for the four sample lists.
One button and a call.
The free first 1,000 is the main button to reviewmyemails.com/get-started. The audits go to reviewmyemails.com/magic, as the webinar offer does, one link per rung.
What N means in plain words: clean once today, and in the next twelve months you stop paying your email tool for the people who had stopped reading. That saving is N times the price of the clean. It holds when the tool bills by contacts stored (10 of 12 big tools do), so the "i" says "up to" for tools that bill by emails sent. The free first 1,000 is counted (your call). We leave out two things that would make N bigger: any better inboxing for the readers, and the sliver after rounding down.
| List | Answers | Gone quiet | A year of quiet | One clean | N | The ending says |
|---|---|---|---|---|---|---|
| Coral, Tidewell Supply Co. | 40,000 · mixed · 4 years · never cleaned · a little growth · $400 a month | 40.7% | $1,955 | $242 | 8 (8.1) | Clean once, save 8 times what it costs. |
| A small shop | 8,000 · consumers · 3 years · never · growing quickly · $110 a month | 25.2% | $332 | $49 | 6 (6.8) | Clean once, save 6 times what it costs. |
| A business list | 150,000 · businesses · 6 years · cleaned 2 years ago · flat · $1,200 a month | 48.2% | $6,935 | $581 | 11 (11.9) | Clean once, save 11 times what it costs. |
| Cleaned last spring | 25,000 · consumers · 3 years · cleaned 6 months ago · a little growth · $150 a month | 6.1% | $109 | $149 | 0.7 | Your list is in good shape. A clean will not pay back yet, check again in six months. |
The branches: N of 2 or more gets "save N times". Between 1 and 2 gets "one clean pays for itself inside a year". Under 1 gets the honest "not yet" above. A list of 1,000 or fewer gets "your first 1,000 free covers it". All four rows are the page's own engine, run on the answers shown.
N always rounds down, so the claim is never bigger than the maths.
The clean price is the price after the free 1,000, and the maths on tap says so: "your first 1,000 are free, so you pay for 39,000".
The ending has two links only: the free 1,000 at /get-started (main) and the audits at /magic.
Sources: decay bands from the calculator research (IDEA-015, consumer and mixed are our estimates), the price list from RME-Pricing.pdf (confirmed 2026-09-24), the billing share from the pricing research snapshot of 4 Sep 2026. The ending uses no Validity figure. It stays in Scene B's "i" only.
What changed after your nine notes (2026-09-24): the motion now carries the story. Every scene is shapes whose size or count is the number: 100 people who go quiet, land in spam and open, a bill square that splits, money squares that merge, a list that fills with quiet contacts month by month. Each frame shows its story as 4 to 6 key poses side by side, then plays the real motion on Play. You can also drag the slider to scrub it. Text is only a label. Every number carries a small tooltip with its source, right next to it, instead of a research shelf.
You asked us to check the sources we chose and the ones we refused against the sources themselves. We checked 33: 24 we use, 9 we refuse. Validity's 2023 and 2025 reports were read in the PDFs. The 2026 Validity report and Klaviyo's report sit behind a download form, so those were checked through their own pages and write-ups that quote them. Full notes with every link: 01-library/research/2026-09-24-calculator-source-audit.md.
We show just under 85% (2022), 84.8% (2023), 83.5% (2024), 87.2% (2025). The 2023 report says inbox rates were "just below 85 percent in 2022" and its chart says 84.8%. Same number. There is no 2024 report, and the 2025 report calls 84.8% its "last edition". So no Validity figure exists for 2023. The right line: 84.8% (2022), 83.5% (2024), 87.2% (2025)
27 billion emails and 150,000 brands is 2025 data (Omnisend's 2026 report), not 2024. We also take no rate from Omnisend, only its size, so it holds up none of our numbers.
SPEC section 9 credits it to Mailgun. It is Validity's, and it is not a measurement: ESPs count sent minus bounced, bounces sit near 1.5%, so dashboards show about 98.5%. Mailgun's number is the separate "only 12% know what delivery rate means".
Not a Gmail mix-up. Two write-ups quoting Validity's 2026 report give 87.2% global ("a 3.7% uplift") and Gmail separately at 89.8%. Validity's own page says inbox rates rose in 2025. 87.2% was Gmail's figure in the 2025 report, a coincidence. It looks right, but nobody has read it in the report yet. If it holds, "one in six never reaches the inbox" becomes about one in eight.
| Source | What it really measures | What to do |
|---|---|---|
| ZeroBounce, "at least 23% a year" (2025) and 28% (2024) | The share of all addresses customers sent in for checking that year that came back bad. A snapshot of lists people chose to check, not one list followed for a year. ZeroBounce calls it yearly decay itself | Keep, it is our business floor. The tooltip already says ZeroBounce sells verification |
| Consumer bands 8 / 12 / 18%, mixed 12 / 15 / 20% | No source gives these numbers. Each source is read right but measures something else: mechanisms with no rate (AWS), how many people go online (Pew, Ofcom), unsubscribes per campaign (Klaviyo), paid cancellations (beehiiv, Reuters) | Done (RME-115.G): labelled "our estimate, built from the studies" in the quiz, the calculator template and the SPEC |
| Country tilt: OECD tenure | OECD gives average tenure. The US 3.9 years is a BLS median, which runs lower | Add "average, the US is a median", it is already marked directional |
| "30 to 40% go quiet within 12 months" | No named study. We refuse it in research/01 but still list it under the consumer middle band in research/02. It measures going quiet, not addresses dying | Done (RME-115.G): taken off the middle card in research/02, and marked refused there |
| Source | How it was measured | Why we use it | Verdict |
|---|---|---|---|
| Validity 2025: inbox 83.5%, spam 6.7%, missing 9.8% | Seed addresses Validity's customers test with at hundreds of providers, only senders who agreed, calendar 2024 | Global inbox rate | Matches |
| Validity 2025 by provider: Gmail 87.2, Microsoft 75.6, Yahoo 86.0, Apple 76.3 | Same seed data | Provider layer | Matches |
| Validity 2026: 87.2% (2025) | Same method, report behind a form | Newest year | Could not verify in the report, likely right |
| Validity 2023: $0.10 an email, $15,000 per million | A sum on the 15.2% not inboxed, not a measured sale | Revenue per email | Matches |
| Mailgun 2025: 12% know what delivered means, 13% seed test | Survey of 1,100+ senders, mid 2024 | Delivered is not inboxed | Matches |
| ZeroBounce 2026: 23%, 62% valid, catch-all over 9% | 11B addresses checked in 2025 | Business floor | Matches their words, see above |
| BLS tenure: 3.9 years | Household survey, about 60,000 homes, Jan 2024 | Business middle | Matches, but today's release says 4.1 years (Jan 2026). Update |
| BLS JOLTS: 3.2% leave a job a month | Employer survey, July 2026, monthly | Business middle | Matches |
| Klaviyo: $0.11 a campaign email, 0.2 to 0.3% unsubscribe | Klaviyo's own accounts, 325B emails | Consumer middle | Partly verified, the brand count and unsubscribe range are in the gated report |
| beehiiv 2026: 5% to 17% a month | Paid newsletter cancellations, 2021 to 2026 | Consumer ceiling | Matches, monthly, paid |
| Reuters Institute 2023: a third cancel or haggle | Online survey, 20 countries, 2022 to 2023 | Consumer ceiling | Matches, paid news |
| AWS, Pew 2011 (70% of Americans use email), Ofcom 2025 (Microsoft reaches 87%) | A blog with no rate, a phone survey, audience reach of all Microsoft services | Consumer floor reasons | Matches. Ofcom is not an email figure |
| Google and Microsoft 2 years, Yahoo 12 months | Provider policies | Why dead addresses pile up | Matches. Yahoo deletes the account itself at 18 months |
| Verisign: .com renewal 75.3% | Registry data, Q3 2025, preliminary | Lapsed domains | Matches |
| Litmus: machine opens over 50% by May 2022 | About 1B tracked opens a month | Machine opens tip | Matches |
List size has no outside source: the engine works the starting size back from the visitor's own answers. Mailchimp's benchmarks (unsubscribes about 0.22%) were checked too and match.
| Number | What the source actually has | Verdict |
|---|---|---|
| 22.5% a year (HubSpot citing MarketingSherpa) | One sentence, no link, date, sample or method. The original could not be found | Reason holds |
| 28 / 30 / 43% a year (ZoomInfo, RevenueBase) | ZoomInfo quotes others, RevenueBase gives no method | Reason holds |
| 70.8% of contacts, 65.8% of titles change a year | Landbase credits IndustrySelect for these | Reason holds, but we named the wrong firm |
| $12.9M (Gartner), $3.1T (IBM) | Estimates about data quality in general (HBR 2016) | Reason holds |
| 36 to 1 (Litmus), 42 to 1 (DMA 2019) | Marketer surveys. DMA: 197 UK marketers. Litmus now shows 2025 data where 36 to 1 is the software sector | Reason holds. Our note that Litmus left out poor measurers could not be confirmed |
| 55 to 85% check spam (Return Path) | Not found anywhere. Our kept figures check out: Mailgun 71% (receipts), ZeroBounce 76% | Reason holds |
| 22 to 25% engagement lost "by Validity, HubSpot and Return Path" | A blog sentence copied between sites | Reason holds |
| 30 to 40% dormant in 12 months | No named study | Reason holds, but we still lean on it (above) |
| Return Path 2024: 97% vs 76% | Return Path became part of Validity in 2019. Only in blog roundups | Reason holds |
Nothing on this page changes the quiz yet. The fixes above go through the ticket, and any outside figure still goes on the Almanac candidates list before it becomes our claim.
This is the first slide of the quiz: three fields you type into, then four one-tap questions. Change any of them and every scene below redraws from the real engine (engine.js, copied into this page). The defaults are the demo visitor, Coral Seaforth at Tidewell Supply Co.
In the real quiz the three fields sit on one slide (your note on the questions). The four taps follow, one per card, each with an "I don't know" chip that picks the default shown here.
This is the quiz after the three fields and the four taps. Press Next (or tap the picture, or use the arrow keys). Each Next shows one change and holds still, and Back steps back one. Change the fields above and the story redraws. From the ask on, the two links show, with "How we got N, in pictures" beside them.
| Your note | What changed | Where |
|---|---|---|
| The list should start and slowly grow to 40,000, then show the share that no longer exists, growing year after year, with a ratio line | Rebuilt as a film of the list itself: it grows from where it started to today, then the quiet band creeps in year by year, then the "still reading" line draws. Animated for real | Scene A |
| Just a number input they type into, and the same for price. One slide with three easy fields | One slide, three typed fields: contacts, emails sent a month, what you pay. Live above | First slide |
| "49 reach an inbox if 83.5% of the rest gets in" is out of context. 100 people, then quiet, then inbox, then opens, then fake opens, so people feel it | Exactly that funnel, as 100 people who change colour, drop into a spam tray, open, fake-open, and finally sort themselves into blocks. Animated for real | Scene B |
| The $4,082 and $1,954 a year only make sense if you read | Now a picture: one square per year of your bill (smaller years were cheaper), the quiet slice of each lifts out and they merge into one square. Then two 3-year squares nest to show what comes back | Scene D |
| Show "18 days" better: spend per period, what we would remove, what is left, what they save, what we cost against that. And price per contact, because a smaller list was on a cheaper plan | The month's bill as a square at your price per contact, split into readers and quiet. The quiet slice becomes a bar, the clean bar sits over it, and the months stack until they pass it. The price per contact carries into Scene D, so older, smaller years cost less | Scene C |
| The cadence scene is hard to understand, imagine it as a video about your list | Four lanes race through the next 3 years, one per cleaning habit. The quiet part of the list rises, gets scooped out at each clean, and the average lands on the right | Scene E |
| Don't ask "did you know there's a high chance", tell them. If they never cleaned, it is certain | It is a statement now. Never cleaned means part of the list has gone quiet, with their own count. A list that has already crossed gets "Most of your list has already gone quiet." | Scene F |
| Research is not a separate shelf. A tooltip next to each number throughout | The shelf is gone. Every number in every scene has its own with the study, year, sample and who made it. The "numbers we refuse" now sit inside the tooltip of the number they would have replaced | All scenes |
Each note also has a reply from the yt_claude bot on the old text, which is kept further down this page under "Round 1".
The one ideaYour list looks like it grew. Underneath, a bigger and bigger slice of it stopped reading, and it never shows up as a bounce.
Lands on41 in 100 gone quiet · 16,288 of 40,000 · 23,712 still reading
Tooltips onthe quiet count (decay band for this kind of list, with the range), the year-1 slice, the still-reading line.
Your note"it starts and slowly grows ... then over time we see the slice grow ... and a line of the ratio." That is this scene, beat for beat.
FlagThe engine works backwards from today, so "4 years at 10% growth" means the list started near 27,321, not at zero. If you want it to grow from nothing, the quiz needs a "how big was it when you started" question, or we draw the first years greyed as an estimate.
The one ideaYou pay to reach 100. By the time you take out the quiet ones, the spam folder and the machine opens, a small handful really read you.
Lands on41 quiet · 10 spam or missing · 37 never opened · 12 real readers. Your report said 22 opened.
Tooltips on41 (decay band), 10 (Validity inbox rate by year, cited not headlined), 12 (Almanac open-rate range), 22 (Almanac and Litmus on Apple's machine opens).
Your note"100 people ... 41 gone quiet, 59 still exist ... 80-something percent inbox ... 30% open ... remove the fake opens ... a small piece of the real list." Done in that order. The open step uses about 1 in 4, because that is what the Almanac supports.
FlagOnly the first step is their own number. The inbox, open and machine-open steps are benchmark ranges applied to their list, so they carry an "example" tag in the tooltip. The quiz never asks the open rate, and SPEC 13.1 says open rate is a gut-check only, never an input. Also: the library has no person icon, so these are a plain geometric figure. A candidate for the asset library.
The one ideaA slice of every bill pays for people who stopped reading, and that slice covers the price of a clean within weeks.
Lands on$400 a month at 1¢ a contact · $163 of it on quiet contacts · one clean $240 · covered in about 6 weeks · then up to $237 a month
Tooltips onthe price per contact (your two fields, divided), the quiet slice (decay band), the emails that go to them (your third field), the clean price (our price list), the payback.
Your note"how much they spend, how much we would remove, how much is left, how much less they spend, and how much we cost against that." Those are the five poses.
FlagCarried from round 1: the smaller bill assumes the tool bills by contacts stored (10 of 12 big tools do). Tiers make the real drop lumpy, so it says "up to". The emails-sent field adds 65,000 emails a month to quiet contacts, which matters most to send-billed tools.
The one ideaThe waste was small at first and grew every year, and it adds up to a real pile. Cleaning twice a year from here keeps a visible chunk of the next three years.
Lands onpaid so far for quiet contacts $4,082 · next 3 years $16,736 never cleaning against $10,526 cleaning twice a year (cleans included) · up to $6,210 back
Tooltips oneach year's square (list size that year times your price per contact), the merged square (the formula with your values), both 3-year squares, the saving.
Your note"only makes sense if you read" and "if the list was smaller before, they weren't paying $400". Area is the money, and the early years are visibly smaller squares.
The guessKept as an optional tap before pose 3: the viewer drags a square to the size they think, and the real one lands beside it. Not in the prototype.
FlagA square's area is the dollars, so a square twice the money is only about 1.4 times as wide. That is the Billion Dollar-o-Gram rule and it undersells the gap a little on purpose.
The one ideaA list starts going quiet again the day after a clean. How often you clean decides how much quiet weight you carry on average.
Lands onaverage quiet over 3 years: never 48 · once a year 20 · twice 11 · three times 8 in 100
Tooltips oneach average, the 15 in 100 line (our "healthy" mark), the cleaning cost per lane, the Almanac's "every three to six months" (005.014.004, 020.009.002).
Your note"hard to understand, imagine it as a video about your list." Each lane is their list, the same bars as Scene A, running forward in time.
The one ideaLeft alone, the quiet ones end up outnumbering the readers. It is a date, not a risk.
Lands onthe quiet ones outnumber the readers around year 6 of the list, about 2 years from now
Two versionsNot crossed yet (Coral): the bars race to the crossing. Already crossed (the button, a business list 5 years old, never cleaned): it opens on "Most of your list has already gone quiet." No question in either.
Tooltips onthe crossing year (engine maths, with "on our maths" and the flat-compounding caveat), the provider policies (Google deletes personal accounts unused for 2 years, Microsoft and Yahoo similar) as the did-you-know.
Your note"why are we asking, we are telling them, it's 100% if they never cleaned." Every list that was never cleaned has quiet contacts on it, because every band in the research is above zero. So we state it, with their count.
The one ideaOne clean is small next to what the quiet ones cost in a year.
Lands ona year of quiet $1,955 · one clean $242 after the free 1,000 · Clean once, save 8 times what it costs.
ThenTwo links, nothing else (decision 2026-09-24-calculator-n-rules): "Clean your first 1,000 free" to reviewmyemails.com/get-started as the main button, and the audits at reviewmyemails.com/magic. No "book a call" here.
FlagBranches: N of 2 or more says "save N times", N between 1 and 2 says "pays for itself inside a year", under 1 says "not yet, check again in six months", and a list of 1,000 or fewer says "your first 1,000 free covers it". N always rounds down.
Scene B uses the Almanac's ranges (about 1 in 4 open, machine opens add 5 to 20 points to the report) on top of their own list. Keep them as an "example" step, or add an optional open-rate field?
The engine starts Coral's list at 27,321 four years ago. Growing from zero needs one more question. Keep the engine's start?
Validity's 87.2% for 2025 is unchecked, the Almanac's decay numbers disagree with the research, and our price breakpoints need your okay. The full table is in round 1 below, unchanged.
IDEA-015-list-decay-calculator/engine.js, copied into this page, so every shape is drawn from simulate, savings, timeline and cleanPrice with the inputs above.envelope-spam-teal.svg, envelope-open-teal.svg (Scene B), prop-treasure-chest.svg (Scene D). Icons are at least twice the height of the text they label.Round 1 as you reviewed it. Its Play buttons are switched off.
Updated after your note (2026-09-24): decay and inbox numbers now come from the calculator's own research and Validity's benchmarks, and every one of them can be found in the story (tap a number, or open the research shelf in frame g). Six quick questions, then a short animated story made from the visitor's own numbers. No scrolling. Every page unfolds, parts pop up, and small paper tabs open tips, facts with their Almanac number, and a share button. It ends on how we help and the app. Press Play on any frame to see its beat.
The inbox line uses Validity's global average by year: 84.8% (2022 data), 83.5% (2024 data). The 87.2% for 2025 has only been seen in write-ups about the 2026 report, not in the report itself. It is not a Gmail mix-up (the source audit checked), but it stays unchecked. It is drawn dashed until someone downloads the report and checks it.
Our estimate from the research puts a mixed list at 12% to 20% a year (middle 15%). The Almanac says 20% to 30% (005.011.003). A visitor who taps through to the Almanac will see a different number. Update the Almanac, or add a line explaining the difference.
Scenes c, d and h assume the bill grows with contacts stored. The pricing research (10 of 12 big tools bill that way, snapshot 4 Sep 2026) backs it, but tiers mean the real saving is lumpy. The money card asks how their tool bills, and says "up to" otherwise.
Full list of missing facts at the end. Files: 02-projects/review-my-emails/ideas/IDEA-015-list-decay-calculator/quiz-story/ (brief, UX, storyboard, premortem and copy review).
Coral Seaforth, marketing manager at Tidewell Supply Co. (both from the demo-names library, RME-109, still waiting on your okay). Her answers: 40,000 contacts · building it 4 years · a mix of businesses and consumers · never cleaned · growing a little (10% a year) · pays $400 a month. Every number below is what the engine gives for those answers, using our estimated band for a mixed list, built from the research: 12%, 15% or 20% a year, with 15% as the middle (IDEA-015 SPEC section 2).
On screenThe book lies closed, then opens. Headline, one line of promise, a Start button, the privacy line.
BeatFold (the two pages open from the gutter), then Pop (headline stands up), then the button.
InteractionStart, or press Space. Nothing else on this page.
Numbernone
Sourcenone, no email fact on this page
Share line
MobilePortrait: one page, not a spread. The book opens bottom to top instead of from the gutter.
On screenOne card per question, six in all: how many contacts, how long you have been building the list, who is on it, when you last cleaned it, is it growing, what you pay your email tool a month (plus an optional "which tool"). Full table in 02-ux.md.
BeatCard folds in, question pops, chips pop left to right. Tapping a chip turns it teal, fills the next dot and folds to the next card.
InteractionTap a chip or press 1 to 6. Every card has "I don't know", which picks a default and tags the number "example" for the rest of the story. The money card also asks how the tool bills: by contacts stored, by emails sent, or not sure (added by the premortem).
Numberdefaults: 40,000 · 4 years · a mix · never · 10% a year · no bill (scenes c and d open with an example you drag)
Sourceno email fact asked. The band each answer picks is in frame a.
Share line
MobileChips stack in two columns, thumb height. The money question becomes a number pad, not a slider.
On screenNot a line going up. The whole list is a grey hill, the living contacts are a teal hill inside it, and the gap between them is the ghost. The hatched strip is the honest range, low to high, with the middle as the headline.
BeatGrey hill pops (your list as your tool shows it). Teal pops inside it and flattens year by year while the grey keeps growing. The hatched range fades in last, and the headline counts to 41.
InteractionDrag across the years to read any point. Tap "low / middle / high" to move the headline. Tap the band to see the study behind each edge.
Number41% gone quiet (range 34% to 51%) · 16,288 of 40,000 (13,583 to 20,267) · still reading 23,712 · by year: 14%, 25%, 34%, 41%
Formulaengine.js simulate(), backward, monthly steps. Monthly decay = 1 − (1 − yearly)^(1/12). Mixed list 12% / 15% / 20% a year (our estimate). Growth 10% a year. List age = the shorter of age and time since last clean.
ResearchSPEC section 2 and research/01 to 02: consumer 8% / 12% / 18%, our estimate built from these studies (Klaviyo, 183k brands, Omnisend, 27B emails, beehiiv, Reuters Institute), business 25% / 28% / 30% (ZeroBounce 11B addresses "at least 23%" a year, BLS median job tenure 4.1 years, JOLTS 3.2% separations a month), mixed is our stated blend, also our estimate (research/02 recommends about 15%)
Did you knowCatch-all domains, over 9% of addresses, accept mail for mailboxes that no longer exist, so a lot of it never bounces. (ZeroBounce Email List Decay 2026)
Share line
FlagThe Almanac says 20% to 30% a year (005.011.003) for every list. The research is lower for consumer and mixed lists and about the same for business lists. Tapping through to the Almanac would show a different number. Vendor figures (ZeroBounce, Klaviyo) are first-party, so the page names that next to them.
MobileRotate the chart: years run top to bottom and the hills grow sideways, so the headline sits on top.
On screenA 10 by 10 grid, one square per 1% of the list. A stack of three paper strips adds it up as the grid changes. Under it, a small line of Validity's global inbox rate by year, with the year used picked out.
BeatAll 100 pop in teal. Then 41 fade to empty outlines, row by row (the gone-quiet share from scene a). The 59 left close up, and 10 of them turn amber: the inbox loss applies only to what is left. The stack drops one strip per step, and the year line draws last.
InteractionTap a year on the line to use that year's rate. Tap "by mailbox" for Gmail 87.2%, Yahoo 86.0%, Apple 76.3%, Microsoft 75.6% (Validity 2025). A "what if you cleaned" toggle: the outlines vanish and the stack shows 100 still reading, about 84 in.
Number100 → 59 still reading → 59 × 83.5% ≈ 49 reach an inbox · delivered 98.5% against inboxed 83.5%
Formulastill reading = 100 × (1 − gone-quiet share). Reach inbox = still reading × inbox rate. Default rate = Validity's latest checked global average (83.5%, 2024 data).
ResearchValidity Email Deliverability Benchmark, 2025 edition (2024 data): inbox 83.5%, spam 6.7%, missing 9.8%. 2023 edition (2022 data): 84.8%. The 2025 edition also explains the delivered 98.5%: an email tool counts sent minus bounced, and bounces sit near 1.5%. It is Validity's illustration, not a measurement. Almanac 016.014.001 agrees that delivered does not mean inboxed.
Did you knowIn 2024, spam placement almost doubled between the first quarter (4.5%) and the fourth (8.6%). (Validity 2025) · 88% of senders could not correctly define "delivered". (Mailgun 2025)
Share line
Flag2025's 87.2% is unchecked (seen only in write-ups of the 2026 report), so it stays dashed. Validity's numbers come from seed accounts at its own customers, so the page says that. The extra drag the quiet contacts put on the rest of the list is not modelled, because we will not fake it.
MobileGrid on top, the stack as one line of three numbers, the year line under that.
On screenA paper month of 30 days. The clean's price fills days one by one. A second card pops: how much of each month's bill goes to the contacts who stopped reading.
BeatDays pop in, then fill teal one at a time until the clean is paid for (18). The count keeps pace. The $236 card stacks in last.
InteractionDrag the monthly bill and the days refill live. If they answered "I don't know", the bill starts at an example and wears an "example" chip. Optional "which tool?" pops a one-line note on how that tool bills.
Numberclean $240 · $400 ÷ 30 = $13.33 a day · $240 ÷ $13.33 = 18 days · quiet share of the bill $400 × 41% = $163 a month
Formuladays = cleanPrice(N) ÷ (bill ÷ 30). cleanPrice is Review My Emails' own price ($0.0080 down to $0.0026, $8 minimum, from the BIBLE). Dead share of bill = bill × dead share.
Researchresearch/04 ESP pricing (snapshot 4 Sep 2026): 10 of 12 big tools bill by contacts stored, only two by emails sent, and storage runs about $9 to $15 per 1,000 contacts a month at middle tiers.
Almanac005.001.010 ("paying for 50,000 contacts but 15,000 of them are invalid, unsubscribed, or completely unengaged, you're paying roughly 30% of your bill") · 009.001.001 (some tools charge per contact stored) · 005.006.001 (a hard bounce still counts toward the billing tier)
Did you knowA contact that hard bounces "still counts toward your billing tier even though you can never email" it. (005.006.001)
Share line
FlagBills do not grow smoothly with contacts, most tools price in tiers. "Bill × dead share" is the Almanac's own framing (005.001.010) but a real saving depends on dropping a tier. The engine's price tiers between $0.0080 and $0.0026 need your check.
MobileCalendar as 5 rows of 6, the $236 card slides up from the bottom.
On screenA half dial and four chips. After the pick, the needle swings to the real number and a strip says how close they were.
BeatChips pop. After a tap, the arc fills amber to the pick, then the needle overshoots to the real value and settles.
InteractionOne tap. "Skip" is always there. If they guessed within 25% of the answer, the strip says "close, most people guess low".
Number$4,082 over 4 years (range $3,345 to $5,229)
Formulasum over every past month of (bill × list size then ÷ list size now) × dead share that month. Same monthly series as scene a.
Source005.001.010 (the paying-for-the-dead framing). The rest is the visitor's own inputs.
Share line
Flag"Most people guess low" can only be said once real guesses are stored. Until then the strip only says how far off they were.
MobileChips become a vertical list, the dial sits above.
On screenA running counter, a bar per year so far, then the fork: never cleaning against cleaning twice a year, and what they get back. Treasure chest from the illustration library (prop-treasure-chest.svg) sits by "Up to $6,210 back".
BeatCounter rolls up while the year bars stack one by one. Then the two paths pop, the red one first, and the chest pops open on the difference.
InteractionA time slider from today to 5 years ahead, the counter runs with it. Tap "twice" to switch between once, twice, three times a year.
Numberso far $4,082 · from here about $1,954 a year ($400 × 12 × 41%) · next 3 years: never $16,736, twice a year $10,526 incl. $1,186 of cleans · up to $6,210 back
Formulaengine.js savings(): monthly loop, bill scales with list size, each clean costs cleanPrice(size then) and resets the dead to zero. Back = never − cadence.
Sourceresearch/04 (stored-contact billing) · 005.001.010 (paying roughly the quiet share of the bill) · 020.014.003 ("20,000 people you're paying to mail, paying to store data on")
Share line
FlagThis is the most aggressive number in the story. It assumes the bill drops in step with the contacts removed. Premortem item 2 suggests showing it as "up to" with the tier caveat on tap.
MobileCounter on top, year bars as a single row, the two paths stacked.
On screenSaw-tooth lines, one per cadence, each drawing itself. A paper dial of 0 to 3 cleans a year. The picked line goes bold, the others fade.
BeatThe red "never" line draws first and keeps climbing. Then each cadence draws and drops to zero at each clean (the saw). The average share counts beside each one.
InteractionTurn the dial. The headline changes with the pick. "Why not every month?" tab: shows three times a year buys only 3 more points than twice (8% vs 11%) for 43% more cleaning spend.
Numberaverage dead share, next 3 years: never 48% · once a year 20% · twice 11% · three times 8%
Formulaengine.js savings() avgShare per cadence at 15% a year and 10% growth. Twice a year is the fewest cleans that keeps the average under 15% (the green line in SPEC section 6).
Source005.014.004 ("every three to six months", a "90-to-180-day cadence") · 020.009.002 (M3AAWG: "Periodic audits quarterly or semi-annually") · 005.001.003 (by volume: over 100K a month "verify monthly", 10K to 100K "quarterly", under 10K "at least every six months")
Did you knowA list "sitting inactive for more than six months" should be cleaned "before your next send". (005.004.002)
Share line
FlagThe Almanac's advice is by sending volume, and the quiz does not ask volume. Proposal: headline says "every 3 to 6 months" (the Almanac's range) and the dial shows the visitor's own maths. Also: savings() and simulate() average the share slightly differently (30% vs 27% for once a year). Make it one function before the build.
MobileDial becomes four chips under the chart.
On screenA paper seesaw. Living on one side, gone quiet on the other. The quiz picks the version from the visitor's own numbers.
BeatVersion 1: the quiet side fills one outline at a time while a countdown ticks down, and the seesaw creeps toward level. Version 2: the seesaw drops with a thud, then "did you know" pops up line by line.
InteractionVersion 1: "what if I clean now" restarts the countdown from a clean list. Version 2: "when did it happen?" drags back along the years to the crossing point.
NumberCoral: the quiet ones take over at about year 6.1 of her list (range 3.9 to 9.2), so about 2 years from now · a list cleaned today and never again: about 6 years (range 4 to 9) · a business list (28% a year), 5 years, never cleaned: crossed around year 2.5, now 68% quiet
Formulaengine.js timeline().crossoverYear (first month quiet > living), and simulate(forward, startClean).markers.crossover.
ResearchEngine maths on the SPEC section 2 bands. SPEC section 3 has the compounding table (at 15% a year, 44% of a static list is left after 5 years).
Did you knowGoogle deletes personal accounts left unused for 2 years, Microsoft marks them inactive after 2 years, and Yahoo after 12 months. (Each provider's own policy page, SPEC section 9)
Share line
FlagSPEC section 3 marks the declining-hazard curve for cohorts older than 2 years as still to do. Until then flat compounding slightly overstates old lists, which matters most for this scene. Fix it before the build, or say "on our maths" out loud (it does).
MobileSeesaw fills the width, the text sits above it.
On screenFour paper pockets: your answers, the research, the Almanac, and the numbers we refuse to use. Each research line opens a card with the study, its year, its sample size, how it was measured and whether the source sells something (SPEC sections 9 and 10).
BeatThe pockets pop left to right. Every "how we got this" link in the story flips to this shelf and opens the right card, so anyone who plays around finds the whole evidence base without looking for it.
InteractionTap a line to open its card. From a card, "see it in the story" jumps to the scene that uses it. Tap an answer to change it.
Numberall of the above, 16 anchor sources from the SPEC section 9 table
SourceIDEA-015 SPEC sections 9 (master citation table) and 10 (stats we do not use), research/01 to 04
Did you know"Numbers we refuse" is its own pocket on purpose. Naming the folklore stats (22.5% a year, 36 to 1 ROI) and saying why we don't use them is the trust move.
Share line
MobilePockets become four tabs you swipe between, and cards open as bottom sheets.
On screenThe story ties up. The cost of one clean against what the quiet contacts cost a year. Then the three verdict emblems (the real verdict-keep / monitor / suppress.svg in the build, drawn as dots here) and the CTA.
BeatThe two bars stack, the "12 times" line counts up, then the three verdicts pop in order and the button rises last.
InteractionMain button to reviewmyemails.com/get-started. Softer "book a call". "Replay with different answers" chip strip under it.
Number$240 clean vs about $1,954 a year ($400 × 12 × 41%) ≈ 8.1, shown as "about 8 times"
Formularatio = (bill × 12 × dead share) ÷ cleanPrice(N). If the ratio is under 1, the page says so and does not push a clean (see premortem).
SourceSPEC section 2 (mixed list, 15% a year middle, our estimate) · research/04 (stored-contact billing) · 005.001.010 · 005.014.010 ("If your validation cost is less than 10% of what a single affected campaign is worth to you, clean the list")
Share line
MobileNumbers on top, verdicts in a row, the button pinned to the bottom of the screen.
On screenThe three tabs every page carries, always in the same corners. The share card the LinkedIn post shows (one card per scene).
BeatTab lift: a small tilt up on hover or tap, the card slides out of the tab like a pull-out.
InteractionShare opens LinkedIn's share window with the quiz URL and copies the line. "Include my number" is off by default, so nobody posts their list size by accident.
Numberthe fact only, the visitor's own numbers are opt-in
Sourcethe study and year are printed on every card, so the post carries its own citation
Other did-you-knows ready in the AlmanacSpam complaints: 30 complaints on 10,000 sends is already 0.3% (019.007.005), Gmail tracks it "as a rolling average" (019.002.013), and it takes "4 to 8 weeks of clean sending" to fade (011.002.010). Duplicates: "you think you have 50,000 subscribers, really it's 42,000" (005.002.015). Several lists: "If you run three lists and someone subscribed to all three, they might be getting three emails" (020.008.006). Catch-all: on a typical B2B list "15 to 35 percent" of addresses sit on catch-all domains, which "accept and then silently drop" (005.003.010).
MobileThe tabs sit in a bar above the thumb, the card opens as a bottom sheet.
On screenNot a page the visitor sees, a map of which ending they get. Plus the thin CTA bar every page carries, so nobody has to reach page h to find the button.
RulesSmall list (under 1,000): the free credits cover it, say so. Cleaned in the last 6 months: a short "you're ahead" ending with the next check date (005.014.004). Ratio under 1 (a year of quiet contacts costs less than one clean): "not worth it yet" with the reason. Everyone else: the full story.
Numberratio = (bill × 12 × dead share) ÷ cleanPrice(N). For Coral, 8.1, so the full story.
Source005.014.004 (the 3 to 6 month recheck) · BIBLE (1,000 free credits, $8 minimum)
WhyPremortem item 7: telling a 1,500 contact newsletter to clean, or telling someone who cleaned in June that their list is rotting, turns the quiz into a sales trick. Saying "you don't need us yet" is the strongest trust move we have.
MobileThe CTA bar sits above the thumb, clear of the LinkedIn in-app browser's own bar.
Per your note (2026-09-24), decay and inbox numbers come from the calculator's research (IDEA-015 SPEC and research/01 to 04) and Validity's benchmarks. The Almanac still backs cadence, billing framing and several did-you-knows. The decision is in 00-inbox/decisions/2026-09-24-calculator-quiz-uses-idea-015-research.md.
| # | What | Where it bites | Options |
|---|---|---|---|
| 1 | Validity 2025 (87.2% global) is unchecked. It has only been seen in write-ups about the 2026 report, and it matches Validity's 2024 Gmail figure exactly. | b (the year line) | Download the 2026 report and check it. Until then it is drawn dashed with a "not yet checked" chip. |
| 2 | The research and the Almanac disagree on decay. Research: consumer 8/12/18 and mixed 12/15/20 (both our estimate, built from the studies), business 25/28/30 (sourced). Almanac: 20% to 30% for every list (005.011.003), and it also disagrees with itself (30-50% after five years in 005.001.006, 5-8% a year in 005.014.010, B2C 22 to 30 against 15 to 25). | a, and any Almanac link from the quiz | Bring the Almanac answers in line with the research (then run refresh.sh), or have the shelf say plainly why the two differ. |
| 3 | RME-108 moved the calculator template onto the Almanac bands earlier today. The quiz now uses the research bands. | the template and the quiz disagree | If "research for the calculator" covers the template too, RME-108 switches back. Flagged to the Super-PM. |
| 4 | How bills scale. The pricing research (dated 4 Sep 2026, several middle tiers marked estimated) says 10 of 12 big tools bill by contacts stored. Tiers make the real saving lumpy. | c, d, h | Ask how their tool bills, and say "up to" otherwise. Re-check the prices before any ad spend. |
| 5 | Our price breakpoints between $0.0080 and $0.0026 in engine.js. | c, d, h | Confirm them. |
| 6 | The declining-hazard curve for cohorts older than 2 years is still to do (SPEC section 3), so flat compounding slightly overstates old lists. | a, f | Build it before the quiz, or keep "on our maths" on the page. |
| 7 | Vendor sources. ZeroBounce, Klaviyo and Validity sell things. SPEC section 11 says to name that. | shelf cards | Every card carries a "who made this and why" line. |
| 8 | Validity and the voice rule. "Never lead with Validity" is a hard content rule. Here Validity is a cited source inside scene b, not the lead. | b, share card | The share card leads with the fact, and Validity is the citation line. Say if you want it kept off the share card. |
Seven failure stories, one agent each, run in parallel. Full write-up in quiz-story/04-premortem-and-copy-review.md.
Phone visitors from LinkedIn wanted one number. Ten pages with a tap each, CTA only on page h, so they left at page d. Fix: scene a comes straight after the questions, and a thin CTA bar sits on every page.
A peer set our decay number next to a different one (the Almanac's, or a vendor's) on LinkedIn. Fix: every number carries its study, the research shelf is one tap away, and the Almanac is brought in line (facts item 2).
Bills price in tiers, so "$236 a month" never showed up. Fix: ask how the tool bills, say "up to" otherwise, show the tier caveat on tap.
Sharing "my list is rotting" is a confession. Fix: share lines are about the fact or the action, not the flaw. One static preview image per scene, made at build time.
LinkedIn opens links in its own in-app browser. The 16:9 stage shrank and swipes fought the browser. Fix: design the 9:16 version first, tap not drag, and test inside LinkedIn's in-app browser.
The public site is the developers' platform now. Fix: before launch, get a tested /get-started that gives 1,000 free credits, plus a signup event we can count (IDEA-013).
A 1,500 contact newsletter was told to clean. Fix: the four endings in frame x.
The hidden assumption: that a scary, precise personal number is persuasive. The research behind IDEA-015 says the honest range and the visible source are what persuade. So the range, "how we got this" and the "not worth it" ending are the product, not decoration.
| Check | Result |
|---|---|
| Em dashes, semicolons | pass none in any on-screen line or share line |
| "Review My Emails" spelled out, no "RME" in visitor copy | pass |
| No inbox guarantee | pass scene b calls the inbox rate Validity's average, never "you will inbox X" |
| Never lead with Validity | check Validity is a cited source inside scene b, never the headline. See facts item 8 |
| No spam-trap claims on a marketing surface | pass the Almanac has trap timelines (005.009.006 and others), kept out on purpose |
| No label count, no vendor name | pass |
| Curiosity over accusation | revise on-screen copy says "gone quiet" and "stopped reading", not "dead". "Dead" stays in the notes only. Share lines never shame the sharer. |
| Numbers over adjectives | pass every page leads with the visitor's own number |
| Firemen, not salesmen | revise "You keep $6,210" reads as a sales promise. Changed to "up to $6,210 back", with the tier caveat. |
| Market as if live | pass CTA "Clean your first 1,000 free" to reviewmyemails.com/get-started, "book a call" softer |
IDEA-015-list-decay-calculator/engine.js (simulate, timeline, savings, cleanPrice) with its own research bands. Every number here came from it.prop-treasure-chest.svg (scene d), prop-stamp-approved.svg (closing), envelope-inbox-teal.svg and envelope-spam-teal.svg (scene b tabs), the three verdict emblems (scene h).