Skip to main content
    SYNQ Research · Amazon US · Sep 2025 – Jul 2026

    Does Amazon's SQP Data Reveal Your Organic Rank?

    Every Amazon seller tracks organic ranks with a scraper — and every brand-registered seller also gets Search Query Performance (SQP) data straight from Amazon. We matched the two, week by week, across six US consumer brands we manage: 61,098 keyword-weeks. The question: when a product ranks well, does Amazon's impression / click / purchase share show it — reliably enough to act on?

    Rank-band gauge: works (AUC 0.83 top-3 · 0.80 top-10)
    Week-to-week move detector: too noisy (63% directional agreement)
    ± Bonus: each source catches the other's errors
    Keyword-weeks analyzed
    61,098
    20,768 matched to an SQP row
    Rank ↔ impression share
    ρ = −0.73
    Spearman, pooled · up to −0.85 per brand
    Best single proxy metric
    Impr. share
    beats click share & purchase share
    Scope
    6 brands
    44 weeks · home, outdoor, personal care & wellness

    01The relationship

    Each dot is one keyword × one week: where a daily rank tracker saw the product family rank (x) vs the impression share Amazon reported for that query (y). The dark line is the expected share at each rank. It falls tenfold between page 1 and the deep ranks — SQP only "sees" what shoppers actually scroll past, so deep ranks barely register.

    no ads on this term that week family ads active bucket median & isotonic fit
    0.03%0.1%0.3%1%3%10%30%131030100300 ≈ page-1 boundary (rank 48) rank 1-3: median 4.35%rank 4-6: median 3.72%rank 7-10: median 3.17%rank 11-15: median 2.28%rank 16-24: median 1.5%rank 25-48: median 0.9%rank 49-96: median 0.32%rank 97-306: median 0.14% expected share at each rank (isotonic fit) Tracked weekly organic rank (median, log scale) SQP impression share (log scale)

    Two things to notice. The cloud is wide — a single week's share can't pin an exact rank — but the level separates cleanly: the top-3 median sits at 4.4% impression share, page-2 at 0.9%, and beyond rank 96 at 0.14%. And the orange dots (ads running) float above the blue at every rank — sponsored placements leak into SQP impressions, which is why ads are handled explicitly in section 05.

    02The visibility ladder

    Median share by rank bucket. Why is a #1 ranking "only" ~4–5% impression share? Because Amazon counts every product tile a shopper scrolls past — dozens per search — so even the top listing owns a few percent of all tiles rendered. The ladder, not the absolute number, is the signal. Click share is steeper at the very top (a #1–3 listing takes 7.6% of all clicks); purchase share concentrates hardest but goes silent below rank ~15.

    Impression share (median)
    1-34.35%4-63.72%7-103.17%11-152.28%16-241.50%25-480.90%49-960.32%97-3060.14%>306†0.34%
    Click share (median)
    1-37.56%4-64.61%7-103.08%11-151.92%16-241.09%25-480.49%49-960.24%97-3060.13%>306†0.27%
    Purchase share (median)
    1-38.33%4-63.53%7-101.25%11-150.00%16-240.00%25-480.00%49-960.00%97-3060.00%>306†0.03%

    † The ">306" rows are keyword-weeks the rank tracker called unranked all week — yet many still show real SQP share. That's a tracker blind spot, not a real rank; see section 07.

    03Reading a share as a rank band

    The practical lookup: given a week's family impression share on a query, where does the listing actually rank? Read each row as "of all keyword-weeks in this share band, X% were top 3 / top 10 / page 1".

    SQP impression share
    (family, weekly)
    nmedian
    rank
    P(top 3)P(top 10)P(page 1)P(deeper
    than 96)
    absent/06,245109
    2%
    13%
    33%
    53%
    0-0.2%4,836118.5
    0%
    2%
    12%
    60%
    0.2-0.5%3,18677.8
    0%
    2%
    22%
    38%
    0.5-1%3,08444.5
    0%
    3%
    54%
    18%
    1-2%3,34725
    1%
    10%
    79%
    7%
    2-3%2,34213
    5%
    38%
    94%
    4%
    3-4.5%2,2679
    13%
    62%
    94%
    4%
    4.5-7%1,4226
    20%
    79%
    93%
    7%
    >7%2843
    52%
    98%
    100%
    0%

    Turned into simple decision rules, tuned so that when the rule fires it's right ≥80% of the time:

    Decision rulePredictsPrecisionCoverage
    (recall)
    Weeks
    flagged
    imp_share >= 9.37%top388.2%5.8%68
    click_share >= 52.08%top397.1%6.4%68
    imp_share >= 4.21%top1080.3%31.9%2,031
    click_share >= 10.77%top1081.1%16.2%1,020
    imp_share >= 0.49%page180.1%75.8%12,863
    click_share >= 0.51%page180.2%61.9%10,494

    The page-1 rule is the workhorse: impression share ≥ 0.5% ⇒ page 1 at 80% precision, and it catches three quarters of all true page-1 weeks. Top-3 can only be confirmed by very high shares (rare but near-certain when seen). AUC reads as: given one in-band and one out-of-band keyword-week, the share ranks them correctly ~80–83% of the time — a strong screener, not a court verdict.

    04What SQP structurally can't see

    SQP only reports a query for your ASIN when the ASIN actually appeared in the results a shopper saw (top-of-search and early rest-of-search, top ~100 queries per ASIN per week). So absence is data. When ads are taken out of the picture, deep ranks vanish from SQP — which makes the missing row itself a usable "we're deep" signal.

    all weeks weeks with no sponsored slot seen (organic only)
    1-388%94%4-682%84%7-1085%90%11-1588%89%16-2486%85%25-4884%79%49-9682%77%97-30664%55%>30649%50% % of keyword-weeks with an SQP row for the product family
    87%When the family is absent from SQP (but the query is measured), the true rank is worse than 10.
    67%…and worse than 48 (off page 1) two-thirds of the time. Absence ≈ "not on page 1, or ads-only visibility".
    94% → 50%Organic-only capture rate falls from 94% at rank 1–3 to ~50% beyond rank 96 — the structural floor of the proxy.

    05Ads inflate the shares — but don't subtract them

    Same rank, two worlds: weeks with the family's ads on the term vs off. Ads roughly double-to-triple the measured share at deep ranks (from a tiny base) and add ~15–40% at top ranks. But naively subtracting ad impressions from SQP impressions over-corrects and makes the proxy worse (AUC drops from 0.80 to 0.74): at deep ranks SQP shows only ~0.6× the impressions the ads console reports, i.e. SQP counts a sponsored appearance only partially. Treat "ads active on this term" as a flag that shifts the calibration one band down — not as a number to subtract.

    family ads off family ads on
    0.05%0.1%0.3%1%3%1-34-67-1011-1516-2425-4849-9697-306>306 median SQP impression share (log scale)

    06Where the proxy fails: week-to-week movement

    Can SQP detect a rank move from one week to the next? Mostly no. Across 14,453 consecutive-week pairs, when rank moved ≥5 positions the share moved in the matching direction only 63% of the time — barely better than a coin flip. And the bigger the move, the smaller the share response, because big moves happen deep on the page where shares are near zero anyway.

    Share response to a weekly rank move
    0-10.30pp2-40.24pp5-90.18pp10-190.14pp20+0.09pp rank move (WoW) → median |Δ share| in points
    Why it fails

    A move from rank 8 → 5 changes expected share by well under one percentage point, while normal weekly share noise is about that size. Rank changes live on a fine scale; weekly share is a coarse, noisy instrument. Keep daily rank tracking for velocity and trend confirmation; use SQP for level, breadth, and truth-checking.

    07The two sources police each other

    SQP is measured from real shopper sessions; rank trackers scrape a geo-neutral snapshot once a day. Where they disagree, the disagreement itself is diagnostic — in both directions.

    37.0%of "unranked all week" (>306) keyword-weeks show >1% SQP impression share — the product was clearly visible to shoppers. Tracker blind spots: variant mismatch, scan depth, crawl locale. (919 rows)
    8.6%of tracker top-3 claims are contradicted by near-zero SQP share — one geo-neutral snapshot ≠ what the average shopper saw (personalization, location, timing).
    0.5%of matched weeks show >10% click share while the tracker says rank >48 — and 95% of those had family ads running: paid placement masquerading as "organic" visibility.

    The starkest pattern: broad, heavily personalized queries. In one case a supplement product showed a tracked rank of #1 on a ~18,000-searches-per-week wellness head term — two separate weeks — while Amazon reported a click share of ~0.4%. A true #1 takes 20%+ of clicks. The scraper wasn't lying about the SERP it saw; it just wasn't the SERP most shoppers got. If a bonus, a budget, or a strategy hinges on "we rank #1", confirm it against SQP first.

    08Does one calibration fit every brand?

    Shape transfers, level doesn't perfectly. Every brand shows the same downward staircase, but category density shifts the curve (different verticals render different numbers of tiles and ad loads). Discrimination is strongest for focused catalogs (Brand B: 0.96 AUC) and weakest for broad multi-category catalogs (Brand A: 0.74) — worth a per-brand fit where precision matters.

    Brand ABrand BBrand CBrand DBrand EBrand F
    0.05%0.1%0.3%1%3%1-34-67-1011-1516-2425-4849-9697-306 organic rank bucket median impression share (log)
    BrandAUC top-10AUC page-1
    Brand A0.7350.77
    Brand B0.9620.994
    Brand C0.9710.79
    Brand D0.963
    Brand E0.8910.792
    Brand F0.8790.93

    09Spotlight: one hero product, 40 weeks

    A category-leading product on its highest-volume keyword. Rank (top panel) and SQP shares (bottom) move together — Spearman −0.77 — including a spring slide from rank ~3 to ~8 mirrored by impression share stepping down.

    impression share click share
    151010111201020304050607 organic rank (1 = top) 0%5%10%15%20%10111201020304050607

    10What we're doing with this at SYNQ

    1 · Visibility bands from SQP alone

    The section-03 calibration runs as a computed layer in our platform: per (product family, query, week) we classify Dominant / Page-1 / Marginal / Invisible from impression share + absence, with the ads-active flag shifting the band.

    2 · Coverage where no tracker is watching

    A rank tracker follows the keywords you told it to follow — a few hundred per brand. SQP reports thousands of queries per brand (in this study, up to ~30–60× the tracked set). Band-level visibility on the untracked long tail is free — and absence is meaningful too.

    3 · Divergence alerts as data QA

    We flag weeks where the tracker and SQP disagree hard (top-3 claim vs ~zero share, or "unranked" vs >1% share) — that's ~9% of top-3 claims today. Each one is a prompt to check variants, locale, or personalization instead of silently trusting either source.

    4 · Daily tracking stays — for velocity

    Section 06 is unambiguous: weekly SQP cannot replace daily rank tracking for detecting movement. The two are complements — the tracker answers "did we move this week", SQP answers "where do we truly sit, on everything, as shoppers saw it".

    11Method & caveats

    • Data: six anonymized US brands managed by SYNQ (home, outdoor, personal care & wellness), Sep 2025 – Jul 2026. Organic ranks from a daily third-party rank tracker; share metrics from Amazon's Search Query Performance (Brand Analytics / SP-API); ad impressions from search-term-level advertising reports. Brand names anonymized (A–F); example keywords masked.
    • Join: daily ranks aggregated to Amazon's Sunday–Saturday SQP weeks (censored weekly median; a day beyond the tracker's scan depth counts as rank 307). Keywords matched to SQP queries exactly (lowercased). Child-ASIN SQP rows summed to the listing's variation family; query-wide market totals deduplicated (MAX, never summed).
    • Shares = family counts ÷ query-wide totals ("share of all product tiles rendered", so top ranks land at 4–7%, not 50%+). Purchase share is zero-inflated below rank ~15 (small counts) — a dominance signal only.
    • Ads attribution: search-term-level ad impressions mapped to product families via ASIN tokens in campaign names (77% of ad volume attributed); available for the last ~6 months of the window — the ads split uses that subset, the sponsored-slot flag covers the full year.
    • Universe caveats: SQP reports ≤100 queries per ASIN per week (Amazon's cap), so absence on a very long-tail query is partly "not in top 100" rather than zero impressions. Pooled statistics weight brands by their keyword counts — one broad-catalog brand contributes about half the panel.
    • Precision numbers are in-sample across these 6 brands / 44 weeks on the US marketplace; treat the calibration as marketplace-specific and re-fit per brand where it matters (section 08).

    © SYNQ · research note · panel built 2026-08-07

    Frequently Asked Questions About SQP & Organic Rank

    What the data says — impression share benchmarks, rank bands, and where each data source can and can't be trusted