{"id":2452,"slug":"seo-j-curve-and-new-site-ramp","title":"SEO J-curve and new-site ramp","kind":"reference","scope":"business","status":"current","audiences":["kevin","smb-owner","candid-team","client-prospect"],"topics":["new-site-trust-accrual","seo-j-curve-ramp"],"reference_body":"## Overview\n\nThe **SEO J-curve** is a stylised model of the cost-versus-return profile of a newly launched website's organic-search investment. It describes a curve in which front-loaded build, content, and technical-SEO costs are incurred before any return materialises; an \"invisible window\" of roughly six to twelve months follows during which organic traffic is near zero regardless of content quality; cumulative return crosses from negative to positive somewhere between month six and month twelve as rankings firm up; and — conditional on the site being genuinely good and held through the trough — evergreen pages can compound for years thereafter.\n\nThe shape itself is industry-consensus as a mechanism. Magnitudes and timing for any specific business cannot be honestly forecast to a point estimate. The curve is paired with a large body of empirical evidence on how long pages actually take to rank, what fraction of pages ever rank at all, and what variance drivers determine outcomes. Both the model and the empirical anchors are documented in the master research briefs [[research-brief-new-site-ramp-economics-june-2026]] and [[research-brief-new-site-timing-benchmarks-june-2026]], which sit alongside four sister briefs in the [[research-cluster-new-website-launch-june-2026]].\n\nA second concept commonly conflated with the J-curve is the \"Google sandbox\" — the proposition that Google deliberately holds new sites out of rankings for a fixed period. Google representatives have rejected the sandbox as a deliberate filter; the observed multi-month ramp is instead attributed to the absence of accrued trust signals. This distinction — *signals vacuum* rather than *penalty box* — is central to interpreting the empirical timing data.\n\n## Core empirical anchors\n\nThe strongest single quantitative anchor for the length and severity of the new-site invisible window is Ahrefs' May 2025 study by Patrick Stox. The study tracked one million random URLs first seen by Ahrefs' crawler in September 2023, plus an additional two million URLs created in October 2023 and filtered to non-empty English content, over the following twelve months.\n\nThe headline finding is that **only 1.74% of newly published pages reached the top 10 within a year**; 98.26% did not. Filtered to non-empty English content the figure rose to 6.11%. Of the small minority that did reach the top 10, approximately 40.82% did so within the first month, with high-search-volume terms (when they ranked at all) more likely to rank inside the first month and low-volume terms spread more evenly across the year.\n\n**Source:** Ahrefs (Patrick Stox), \"How Long Does It Take to Rank in Google?\", May 15, 2025; 1M random URLs first seen September 2023, tracked for one year. **Confidence:** Verified (named source, methodology disclosed, large sample). **Caveat:** Ahrefs is a tool vendor whose product helps with SEO, but the headline number cuts against the typical vendor incentive to oversell — a low base rate makes SEO harder, not easier, to sell. The sample of \"newly published pages\" skews toward the quality side of the web, and the figure describes the *distribution* of outcomes, not a forecast for one site. \"First seen by Ahrefs crawler\" is a proxy for publication date and may lag.\n\nFor comparison the same study reported a 2017 figure of **5.7%** — meaning the share of new pages reaching the top 10 within a year declined by roughly **3.3×** over eight years. This 5.7% → 1.74% drop between 2017 and 2025 is the single strongest large-sample signal that the environment for new pages got materially harder over the period.\n\nThe picture is markedly worse at the competitive end of the keyword distribution. In the same Ahrefs 2025 study, the share of newly published pages reaching the top 10 within a year **for high-volume keywords specifically** is **0.3%** — roughly one-sixth of the all-keywords 1.74% figure.\n\n**Source:** Ahrefs, Patrick Stox, May 2025. **Confidence:** Verified. The figure implies that targeting competitive heads on a brand-new domain is an extraordinarily long-tail bet at the page level: the realistic Stage-1 strategy is to earn long-tail and low-competition queries first while site authority accrues, rather than to chase volume terms on day one.\n\nTwo further Ahrefs figures complete the empirical floor. **96.55% of all pages get zero organic traffic from Google**; only 3.45% earn any meaningful organic traffic at all (**Source:** Ahrefs industry analysis. **Confidence:** Verified (widely cited Ahrefs figure). **Caveat:** the denominator includes a long tail of low-effort, abandoned, and non-commercial pages; even adjusting for that, the share of pages earning meaningful traffic is small, and survivorship bias in vendor \"look at this winner\" case studies is enormous). And **72.9% of pages ranking in the top 10 are more than three years old** (**Source:** Ahrefs industry analysis. **Confidence:** Verified). The top-10 SERP is, by composition, dominated by aged content.\n\nA separate Ahrefs figure from the same 2025 study finds that the average #1-ranking page is **5 years old**, up from approximately 2 years old in 2017 (**Method:** 1.3M US keywords, top-10 URL ages via Ahrefs crawler \"first seen\". **Source:** Ahrefs, 2025. **Confidence:** Industry-consensus. **Caveat:** \"first seen\" is an age proxy that lags actual publication; the relationship is correlational, not causal). The figure describes the average page that holds #1 *today*, not how long any new page takes to get there — a new page can reach #1 sooner if it offers something the incumbent does not. The two are related but distinct measurements.\n\n### Ahrefs ranking benchmarks\n\n| Metric | 2025 | 2017 |\n|---|---|---|\n| Newly published pages reaching top 10 within a year (all keywords) | 1.74% | 5.7% |\n| Same metric, high-volume keywords only | 0.3% | 0.3% |\n| Average age of #1-ranking page | ~5 years | ~2 years |\n| Share of top-10 pages older than 3 years | 72.9% | — |\n| Share of pages with zero organic traffic | 96.55% | — |\n\nCaveats above attach to each row. The 0.3% figure for high-volume keywords specifically is the single most useful number to quote when a site owner asks why a brand-new domain is not ranking for a competitive head term after three months: 0.3% of pages do that within a *year*; volume terms are an established-site game.\n\n## The Semrush stability picture\n\nA separate 2022 Semrush study, \"How Long Does It Take to Rank Higher on Google\", tracked 28,000 domains never previously seen in Semrush's US database, all of which were ranking somewhere in the top 100 for at least one keyword at study start, over thirteen months from July 2021 to July 2022. The published results emphasise *stability* rather than initial appearance:\n\n- Approximately **41% were ranking in the top 10 after six months** at some point during the period.\n- Only **19% reached the top 10 by month 6 and maintained it to the end**.\n- Of websites that made it to the top 10, **27% remained there the entire study**.\n- **Fewer than 5% maintained first-page rankings for a year**.\n- **92% failed to stay in the top 100** across the year.\n- Just **4.2%** held at least one top-10 ranking for all thirteen months.\n- Nearly **10% appeared in the top 10 in only one month** of the year.\n- **55.1% of domains that failed to reach the top 10 had no backlinks.**\n\n**Source:** Semrush, \"How Long Does It Take to Rank Higher on Google\", 2022 (data 2021–2022). **Confidence:** Single-source for the headline percentages. **Caveat:** product-incentivised (Semrush sells SEO tooling), but the more important issue is severe selection bias — the sample is defined as domains that already ranked in the top 100 for at least one keyword, which excludes the much larger universe of new domains that never ranked at all. The figures therefore systematically *understate* time-to-traction for a truly random new site. Every Semrush figure here should be read against that filter.\n\nThe Semrush picture frames stable ranking as the exception rather than the milestone most pages reach. Read against the Ahrefs base rates, the conclusion is convergent: rare initial top-10 entry, even rarer sustained presence.\n\n## Indexing as a prerequisite\n\nRanking is downstream of indexing, and the indexing pipeline imposes its own delay distribution. Synthesising Google's own published guidance, John Mueller's recurring statements, large-sample data from IndexCheckr (16 million pages, February 2025), and Onely's submission-test work, the defensible client-facing range is that **for pages that get indexed at all, the median lands somewhere in the 1–4 week range**, with a meaningful minority taking 1–3 months and a small share taking longer. Same-day or within-48-hour indexing is real but concentrated among high-authority and news sites.\n\n**Source:** Synthesis of Google Search Central documentation, John Mueller \"several hours to several weeks\" statements, IndexCheckr 16M-page February 2025 study, and Onely benchmarks. **Confidence:** Industry-consensus on the range; survivorship-flagged on every constituent number. **Caveat:** the synthesis must always be paired with the failure-tail figure — roughly 16% of even valuable, indexable pages are never indexed at all. Quoting the 1–4 week range without the tail is what produces unrealistic expectations.\n\nThe indexing distribution and its bias flags are covered in detail in [[research-brief-new-site-timing-benchmarks-june-2026]]; the underlying pipeline mechanics (discovery, crawl, render, index selection, ranking, re-evaluation) are documented separately in the sister Google-search-lifecycle brief.\n\n## Why there is no fixed settling window\n\nThe proposition that \"a new site takes [N] months to rank\" is one of the most persistent claims in SEO discourse. Adjudicated against the available evidence:\n\n- The **deliberate sandbox** — Google withholding rankings from new sites for a fixed period — is **false** as a documented mechanism. Mueller has rejected the sandbox and honeymoon framings explicitly.\n- The **effect** of a multi-month trust-building lag is **real but variable**. The \"3–6 months\" figure widely cited in practitioner discourse is a useful planning heuristic, not a measured constant. The underlying mechanism is missing trust signals, not a Google filter.\n\n**Source:** Synthesis of Mueller 2018/2021 statements rejecting the sandbox and practitioner consensus. **Confidence:** Verified on the no-deliberate-sandbox half; Directional-Speculative on the \"3–6 months\" planning number. **Caveat:** The heuristic is appropriate for setting expectations and inappropriate as a promise. The honest framing is that a multi-month ramp should be expected, the mechanism is signals accruing rather than a timer running out, and which mechanism is at work can be diagnosed month-by-month from Google Search Console movement.\n\nThree further considerations converge on the same conclusion that no fixed settling window exists:\n\n1. Mueller frames new-site ranking volatility as Google \"making assumptions\" before it has signals — assumptions that correct in *either* direction rather than constituting a deliberate boost or penalty.\n2. The Semrush 28,000-domain study (above) found fewer than 5% of new domains maintained first-page rankings for a year, only 4.2% held a top-10 keyword across all thirteen months, and 92% failed to stay in the top 100. Stability is the exception, not a milestone most pages reach.\n3. The \"honeymoon period\" — new content ranking high then dropping, or starting low then climbing — is, per Mueller, not a deliberate boost or penalty but Google making assumptions before it has enough signals, which then correct in either direction. Observed swings range from hours to a few weeks; practitioner estimates of a longer adjustment window stretch to 3–6 months.\n4. The 2025–2026 core-update context further weakens any \"rankings will settle in N weeks\" assumption. Normally SERPs settle two to four weeks after a core update completes, but the December 2025 core update did not settle cleanly — Search Engine Roundtable documented nine volatility waves in seven weeks following its rollout.\n\n**Source:** Synthesis of Mueller 2021 statements, Semrush 2022, and Search Engine Roundtable December 2025–February 2026 coverage. **Confidence:** Industry-consensus. **Caveat:** \"no fixed settling window\" does not mean \"rankings are random.\" It means the timeline distribution is wide and expectations should be steered toward distribution-based rather than point-estimate framings.\n\nThe practitioner heuristic itself — widely repeated across 2024–2026 — comes in two parts:\n\n- **New sites:** 3–6 months for meaningful traction; 9–12 months in YMYL or competitive verticals.\n- **New page on an established, high-authority domain:** ranks in days-to-weeks, inheriting crawl priority and domain-level trust.\n\n**Source:** Practitioner consensus, 2024–2026. **Method:** no controlled study. **Confidence:** Directional-Speculative. **Caveat:** anecdotal and survivorship-biased — practitioners remember the sites that worked. The \"3–6 months\" framing is a useful planning heuristic, not a measured constant. The underlying mechanism is missing trust signals, not a Google filter.\n\n## The J-curve model\n\nThe J-curve treats organic-search investment as a capital project rather than a marketing expense. The cost-versus-return profile is decomposed into four stages, each capturing a distinct phase of the curve.\n\n### Stage 0 — Upfront investment (Month 0)\n\nBuild, content, and technical-SEO costs are incurred immediately and are largely **sunk** — they precede any return. Practitioner syntheses consistently note that SEO costs are front-loaded across strategy, technical fixes, and content creation. This is the bottom-left of the J: cash out, nothing in.\n\n**Source:** Compass research synthesis, June 2026; vendor practitioner syntheses (directional only). **Confidence:** Industry-consensus for the mechanism. **Caveat:** the *size* of upfront spend does not predict the *size* or *timing* of return. Spending more upfront does not buy a shorter invisible window; it only deepens the trough if the site does not eventually rank.\n\n### Stage 1 — Invisible window (≈Months 0–6)\n\nSearch engines must crawl, index, and accrue trust in new pages; new domains lack accumulated authority and backlinks. **Traffic is near zero regardless of content quality** through roughly the first six months. The mechanism is a *signals vacuum* — Google has not yet observed enough about the site to rank it confidently — not a deliberate filter.\n\nThe empirical anchors for the magnitude of the window are the Ahrefs base rates above: 1.74% of newly published pages rank in the top 10 within a year overall; 0.3% for high-volume keywords; and among the minority that do rank within a year, most reach the top 10 between roughly 61 and 182 days after publication. **Source:** Compass research synthesis citing Ahrefs May 2025. **Confidence:** Industry-consensus for the existence and length of the window; Verified for the Ahrefs base rates. **Caveat:** the invisible window cannot be shortened by spending more on the build or by submitting more aggressively to Google — the lever is *trust accrued over time*, not effort applied.\n\n### Stage 2 — Traction and break-even approach (≈Months 6–12)\n\nAs pages accrue authority and engagement signals, rankings firm up; the cumulative-return curve crosses from negative to positive somewhere in this window. Multiple independent practitioner syntheses converge on **3–6 months for early signals** (impressions, long-tail clicks, indexation depth) and **6–12 months for consistent organic traffic or leads**.\n\n**Source:** Compass research synthesis, June 2026; Shopify's SEO lead Arthur Camberlein (\"most sites can expect to see measurable results from their SEO efforts within three to six months\"). **Confidence:** Industry-consensus for the 3–12 month range. **Caveat:** \"results\" does not equal \"payback.\" Traffic milestones precede revenue milestones, which precede break-even. A site that \"starts ranking\" in month 6 may still be months from cumulative break-even. The range applies to leading indicators (movement on impressions, position, long-tail clicks), not to recovery of upfront investment. Confusing the two is one of the most common conversation errors during the trough.\n\n### Stage 3 — Compounding (Year 2+), conditional\n\nEvergreen pages can accumulate traffic and links over years; this is the part of the curve that justifies the upfront sunk cost. HubSpot's \"compounding posts\" data and the broader observation that content is one of the few forms of marketing with a compounding return describe the asset-appreciation mechanism.\n\n**Source:** Compass research synthesis, June 2026; HubSpot Research; Tomasz Tunguz. **Confidence:** Industry-consensus for the mechanism; Single-source / non-agency for the HubSpot magnitude (used for mechanism, not magnitude). **Caveat:** this stage is most often *assumed* in vendor ROI pitches. The honest framing is that compounding **can happen** when conditions are right, not that it **will happen** if the site owner waits long enough. Compounding is conditional, not guaranteed (see *Failure modes*, below).\n\n### The four stages summarised\n\n| Stage | Timing | Cumulative return | What is happening |\n|---|---|---|---|\n| 0 — Upfront | Month 0 | Strongly negative | Build, content, and technical costs incurred; nothing yet returned |\n| 1 — Invisible window | Months 0–6 | Deepens negative | Crawling, indexing, signals accruing; traffic near zero regardless of quality |\n| 2 — Traction approach | Months 6–12 | Crosses zero | Rankings firm up; early conversions; cumulative return approaches break-even |\n| 3 — Compounding | Year 2+ | Compounds positive | Conditional; evergreen pages accrue traffic, links, and authority — *if* the site is genuinely good |\n\n**Source:** Synthesis from Compass research, June 2026, consistent across multiple independent practitioner explainers. **Confidence:** Industry-consensus for the shape; Directional-Speculative for any specific magnitude or timing. **Caveat:** the J-curve is a *mechanism*, not a forecast. Treating it as a predictive model for a specific business is the most common analytic error.\n\n## When organic does not compound\n\nThe compounding case for organic search is conditional. Six identifiable failure modes are documented in the literature in which the J-curve never turns up and the trough becomes the entire story:\n\n1. **Zero search demand** for the topic — no one searches for what the business does.\n2. **Thin or undifferentiated content** that never earns authority or links.\n3. **Fundamentally weak site or product** — no product-market fit, no conversion regardless of traffic.\n4. **Algorithm update or AI-Overview shift** that erases traffic at stable rankings.\n5. **Abandonment or rebuild that resets authority** (see *Rebuild and authority loss*, below).\n6. **Highly competitive heads** where established incumbents hold the SERP — the 0.3% top-10-within-a-year figure for high-volume keywords above documents the base rate.\n\nAny one of these can flatten the J-curve; combinations can convert it into a permanent loss.\n\n**Source:** Compass research synthesis, June 2026. **Confidence:** Industry-consensus. **Caveat:** diagnosing *which* failure mode applies requires honest investigation, not just waiting longer. Pattern-recognising \"we're in the invisible window, wait it out\" when the underlying problem is failure-mode 1, 2, 3, or 6 is the sunk-cost trap in disguise.\n\n## The sunk-cost framing cuts both ways\n\nThe economic argument around premature abandonment versus persistent investment is symmetric:\n\n- **Premature abandonment** converts a temporary paper loss into a permanent realised loss. The sunk-cost literature explicitly warns that overemphasis on avoiding the sunk-cost fallacy can lead to premature abandonment of otherwise worthwhile projects.\n- **Patience as a doctrine** is the textbook sunk-cost trap when the underlying site is fundamentally weak. Throwing additional money at an asset with no search demand, no differentiated content, or no product-market fit is not \"holding through the trough\" — it is the fallacy itself.\n\nThe economically correct test is **forward-looking**: does the future payoff from continuing beat the next-best alternative use of the same money and attention? Past investment is irrelevant to the decision.\n\n**Source:** Compass research synthesis citing the sunk-cost literature. **Confidence:** Industry-consensus on the principle. **Caveat:** this is the most uncomfortable framing for both vendors and site owners because it explicitly admits that the right answer is sometimes to stop. The diagnostic threshold below operationalises when the question should be asked.\n\n### The six-month diagnostic\n\nA practitioner-level diagnostic threshold is documented across multiple sources: if, by approximately six months after launch, there is **no movement in impressions or keyword footprint** despite clean technical SEO (indexing functioning, mobile parity, no robots.txt or 5xx issues) and **genuine, differentiated content**, the appropriate response is to **re-evaluate the site itself rather than continue waiting**.\n\nThe check at month six runs through:\n\n- Are technicals clean? — Crawl Stats, mobile parity, sitemap, response codes.\n- Is there real search demand for the topic? If no one is searching, failure mode 1 is in play.\n- Is the content differentiated and substantive? If undifferentiated, failure mode 2.\n- Does the underlying product or service convert when traffic does arrive? If not, failure mode 3.\n\nIf the diagnostic surfaces a fundamental problem, an honest stop or a strategy pivot is the rational response. Continuing to invest in an asset that will not compound is precisely the sunk-cost trap the threshold is designed to prevent.\n\n## Variance drivers, ranked by evidence\n\nThe factors that determine where any individual site lands on the timing distribution are not equally well-supported. Ranked from strongest to weakest evidence:\n\n1. **Query competition and search volume — strongest.** Ahrefs documents that high-volume terms are dramatically harder (the 0.3% top-10-within-a-year figure for high-volume keywords in 2025; broadly similar pattern in 2017). Low-volume and long-tail terms rank materially faster. Semrush observed that top performers targeted slightly longer queries (3.2–3.5 words) than the overall sample.\n2. **Site authority and backlinks — strong.** Semrush reported that more than half of domains with no backlinks never reached page one; 55.1% of top-10 failures had zero backlinks. Ahrefs found high-Domain-Rating pages performed significantly better. Top-10% domains in Semrush's data averaged a Domain Authority of approximately 20.3.\n3. **Content quality — strong for indexing, moderate for ranking.** Indexing Insight's March 2025 1.7M-page study found that 88% of not-indexed pages were quality-driven. Google's March 2024 core update was described as removing 45% more low-quality, unoriginal content from results.\n4. **Content depth and length — moderate, correlational.** Semrush found top-10% domains averaged 846 words against 243 for lower performers; top-ranking content ran roughly 3.5× longer. **Mueller has cautioned that Google does not use word count as a ranking factor**, and the correlation should be treated as such, not causation.\n5. **Internal linking and crawl efficiency — moderate, mechanism-backed.** Google and Onely both emphasise prominent internal links and clean architecture as the principal levers for moving pages from \"Discovered — currently not indexed\" or \"Crawled — currently not indexed\" into the index.\n6. **Content velocity and publishing frequency — weak, Directional-Speculative.** The \"publish weekly to train the crawler\" claim is frequently asserted but not backed by controlled large-sample data. Google's stated position is that crawl demand responds to popularity, staleness, and quality, not to a publishing-cadence quota.\n\n**Source:** Synthesis of Ahrefs 2017/2025, Semrush 2022, Indexing Insight 2025, and Google statements. **Confidence:** Industry-consensus on the ordering; each individual figure carries its own source's confidence label. **Caveat:** the depth-and-length correlation is the most-abused number in this list — top-performing pages are long because they are comprehensive, not high-ranking because they are long. Treating word count as a direct ranking lever is a category error.\n\n## Ramp economics\n\nThe J-curve has direct consequences for how the investment is financed and amortised. Three economic mechanisms recur across the literature.\n\n### Amortisation over useful life\n\nA website is a capital asset with a finite useful life. Spreading the build cost across that life converts a lumpy upfront number into a per-period cost. Huemor's Wayback Machine analysis of INC-5000 sites found an average site lifespan of **2 years 4 months** before substantive rebuild. HubSpot has reported that **71% of marketing leaders redesign every one to three years**. Slower-moving local and service sectors can run three to four years or longer before rebuild.\n\n**Source:** Huemor (Wayback-Machine analysis of INC-5000 sites); HubSpot. **Confidence:** Single-source for the Huemor figure; Industry-consensus for the one-to-three-year rebuild pattern. **Caveat:** Huemor is a web-design agency with a vendor incentive to argue for longer-lived rebuilds. INC-5000 sites are growth-stage US companies and are not representative of small-business or local-service markets. The figure is directional for the rebuild-cycle pattern, not a target lifespan for any specific operator.\n\nThe implication, paired with the typical 6–12 month invisible window, is that the rebuild cycle and the J-curve invisible window are uncomfortably close in duration. A site rebuilt at 28 months has had only roughly 16–22 months of productive ranking life after the invisible window — a poor return on the upfront sunk cost.\n\nPractitioner sources note that **disciplined incremental refreshes** (rather than full rebuilds) can extend useful life to 4–5+ years and improve amortisation substantially.\n\n**Source:** Compass research synthesis, June 2026; web-design practitioner sources. **Confidence:** Industry-consensus on the framing; Single-source on specific lifespan figures. **Caveat:** \"useful life\" is partly a choice, not a fixed number. The lifecycle decision (rebuild versus refresh) materially changes amortisation.\n\n### Invisible window as fraction of useful life\n\nIf a site takes 6–12 months to gain traction and is rebuilt at 30–36 months, then **roughly one fifth to one third of the useful life produces little-to-no organic return** — which raises the *effective* per-productive-month cost and makes premature rebuilds particularly destructive. A simple arithmetic sketch:\n\n| Useful life (months) | Invisible window (months) | Productive months | Invisible-window share |\n|---|---|---|---|\n| 30 | 9 | 21 | 30% |\n| 36 | 9 | 27 | 25% |\n| 48 | 9 | 39 | 19% |\n| 60 | 9 | 51 | 15% |\n\nExtending useful life via incremental refresh drops the invisible-window share to roughly 15–20%, materially improving the amortisation picture.\n\n**Source:** Synthesised from Compass research, June 2026. **Confidence:** Directional-Speculative — the arithmetic is exact given the assumptions, but the assumptions span wide ranges. **Caveat:** this is the strongest single quantitative argument for the \"rebuild less often, refresh more often\" pattern that prevails in disciplined web-development practice. It does not specify how often \"less often\" should be in any given case; it specifies the trade-off the choice produces.\n\n### Bridge spend deepens the trough\n\nWhen paid acquisition is used as a bridge during the Stage-1 invisible window, the paid spend is a real, additive cost. It **deepens the J-curve trough**; it does not replace the trough.\n\nFor modelling purposes, bridge cost belongs on the early-window cost line of the J-curve. The total upfront economic burden during the invisible window is **build + content + technical + bridge paid through the invisible window**, not just the build figure.\n\n**Source:** Compass research synthesis, June 2026. **Confidence:** Industry-consensus on the principle. **Caveat:** failing to budget the bridge is one of the most common small-business planning errors. The owner sees the build invoice, does not budget the six to twelve months of paid spend that the J-curve dictates, and runs out of runway at the bottom of the trough — converting an asset-building project into an abandoned project just before it would have turned.\n\nThe rule that follows is direct: when modelling the J-curve, include the bridge paid spend in the trough. The total economic burden during the invisible window is build + content + technical + bridge paid, not just the build figure. Refusing to budget the bridge is a flag that the underlying project may be under-capitalised.\n\n## Rebuild and authority loss\n\nA website rebuild that breaks URL structure — changing slugs without redirects, changing the domain, abandoning category structures — can forfeit \"the most valuable marketing channel\" and take \"months or years to recover.\" Accumulated link equity is discarded; canonical clusters that Google had stabilised are reset. The literature on [[web-migration-mechanics]] documents the specific failure patterns.\n\nThe implication is that full rebuilds inside the first roughly 24–36 months are particularly destructive because the invisible-window cost has been paid but the compounding has not yet materialised. The rebuild discards the asset just before it would have started producing.\n\n**Source:** Compass research synthesis, June 2026, citing practitioner sources. **Confidence:** Industry-consensus. **Caveat:** \"rebuild\" is not a binary category. Incremental refreshes — refreshing copy, updating design tokens, swapping the page-builder under the hood while preserving slugs and information architecture — preserve authority. The damaging pattern is the **full URL-structure rewrite**, which is also what most agency-led \"redesigns\" deliver. Migration discipline and the trade-offs around [[platform-lock-in]] sit alongside this consideration.\n\nThe corresponding practitioner rule is to avoid full rebuilds inside the first 24–36 months unless **performance data** — not aesthetics, not the perception that the site \"feels dated\" — demands it, and to prefer incremental refreshes that preserve URL structure when changes are needed. When a rebuild request originates from \"we want a new look,\" a design refresh that keeps the underlying URL structure intact preserves the asset value built during the invisible window.\n\n## Engagement structure: the multi-year frame\n\nThe duration mismatch between typical project-based agency engagements and the J-curve has been documented as a recurring failure pattern. Focus Digital's 2026 work on agency lifespans found that project-based engagements average approximately **24 months** in length, while retainer-based agencies average **56 months**. Twenty-four months is roughly the point at which a healthy site would be entering Stage-3 compounding.\n\nThe implication is that project-based engagements end at the moment the asset becomes valuable. The client pays the upfront cost, sits through the invisible window, *almost* reaches compounding, and then either ends the engagement (discarding the asset) or hands the site to an in-house team that does not know how to maintain it. The retainer structure — or any multi-year commitment — is what captures the compounding stage.\n\nThe corresponding framing is to treat the engagement as a multi-year asset-building exercise from the outset rather than a one-shot project: the first 6–12 months are mostly cost; value appears in years two and three if the line is held. The engagement structure that matches that economic shape is multi-year, not build-and-handoff.\n\n## How the model interacts with the 2025–2026 search environment\n\nSeveral documented 2025–2026 discontinuities affect how the J-curve and its empirical anchors should be read.\n\n- **AI-Overviews and click compression.** Independent measurements across Bain (February 2025), Pew Research Center (July 2025, 68,879 real searches), and Ahrefs (December 2025, 300,000 keywords) document material reductions in click-through rates when AI summaries are present. Older click-based payback studies are accordingly upper bounds in the current environment. Citation behaviour patterns within AI surfaces are documented separately at [[ai-overview-citation-patterns]].\n- **Core-update volatility.** The December 2025 core update did not settle cleanly; Search Engine Roundtable documented nine volatility waves in seven weeks. The \"rankings settle in 2–4 weeks after a core update\" planning rule is therefore not safe to apply blanket.\n- **Mass de-indexing.** Indexing Insight's May 2025 work documented approximately 25% of two million monitored pages removed from the index, with individual sites losing between 15% and 75% of their previously indexed page sets. The indexing prerequisite is no longer safely assumed to be stable once achieved.\n- **Quality bar tightening.** Google's March 2024 update was described as removing 45% more low-quality, unoriginal content. Failure modes 2 (thin content) and 3 (weak underlying site) have higher base rates in 2025–2026 than in earlier years.\n- **Crawler-side AI integration.** Cloudflare's July 2025 measurements found Googlebot volume up 96% year-on-year and that the same crawl pipeline serves both classical search and AI Overviews / AI Mode.\n\nThese conditions interact with the empirical anchors at the top of this page. The 1.74% top-10-within-a-year figure is itself a 2025 measurement; older studies (such as the 5.7% 2017 figure) represent a materially easier prior environment. Page quality and technical foundations including [[core-web-vitals]] and rigorous [[editorial-discipline-and-sourcing]] remain prerequisites; they have become more binding rather than less.\n\n## How to read timing claims honestly\n\nThree through-lines emerge across the empirical record:\n\n1. **Distributions over averages.** Google publishes ranges, not point estimates. Vendor \"average\" claims are usually survivorship-biased. The 1.74% top-10-within-a-year figure carries the same epistemic weight as Mueller's \"several hours to several weeks\" indexing range — neither licenses a point estimate for any specific site.\n2. **Mechanism versus magnitude.** The J-curve shape, the signals-vacuum framing, and the variance-driver ranking are well-supported as mechanisms. None of them licenses a specific numeric forecast for one site. Treating mechanism-grade evidence as forecast-grade evidence is the most common error in vendor pitches.\n3. **Vendor-incentive flags.** Every figure originating from a seller of the thing it measures — SEO tool vendors (Ahrefs, Semrush, IndexCheckr, Onely, Indexing Insight), web-design agencies (Huemor), CRO platforms — carries a caveat. The Ahrefs base rates above are usable in part because the low base rate cuts *against* the typical vendor incentive to oversell; figures that cut *with* the vendor's commercial incentive require more scrutiny.\n\nThe defensible client-facing framing of timing questions is therefore not a number but a distribution: indexing within days to weeks for most pages on a technically sound site, with roughly 15–20% probability that any given valuable page is never indexed; reaching the top 10 within the first year is below 10% per page and any faster result should be treated as upside; meaningful organic traction expected at 3–6 months and consistent traffic at 6–12 months on the practitioner heuristic, with the 6-month threshold operationalised as a diagnostic decision point rather than a milestone.\n\n## See also\n\n- [[research-brief-new-site-ramp-economics-june-2026]] — master ramp-economics brief\n- [[research-brief-new-site-timing-benchmarks-june-2026]] — master timing-benchmarks brief\n- [[research-cluster-new-website-launch-june-2026]] — the six-brief launch cluster\n- [[core-web-vitals]] — technical-foundation prerequisites that interact with the indexing prerequisite\n- [[ai-overview-citation-patterns]] — citation behaviour within AI search surfaces\n- [[editorial-discipline-and-sourcing]] — content-quality prerequisites that govern failure modes 2 and 3\n- [[platform-lock-in]] — engagement-structure considerations adjacent to the multi-year frame\n- [[web-migration-mechanics]] — rebuild-and-authority-loss mechanics in detail\n","rationale_body":null,"metadata":null,"links":{"outgoing":[],"incoming":[]},"created_at":"2026-06-25T17:23:11.113Z","updated_at":"2026-06-25T17:23:11.113Z","resolved_via_alias":"ramp-when-organic-does-not-compound","resolved_via_alias_anchor":"when-organic-does-not-compound"}