Why average position misleads, and how to build a visibility score from position weights, keyword demand and coverage to compare yourself with competitors.
"What's our average rank?" is the most common question in Amazon rank tracking, and one of the least useful. An average position hides the keywords you lost, treats every position step as equal, and ignores how much demand sits behind each search. This guide explains what to measure instead: a visibility score, or your share of page one, built from position, demand and coverage, and how to use it to compare yourself with competitors.
Why average position misleads
Average position has two structural flaws.
It only counts keywords where you rank. Google's Search Console documentation is explicit about this for web search: "A link must get an impression for its position to be recorded." Queries where you never appeared show a dash, not a bad number. Most Amazon rank averages work the same way. The consequence is selection bias. If you fall off page one for a weak keyword, that keyword leaves the average, and the average improves, while your actual visibility shrank.
It treats all steps as equal. Moving from position 1 to 2 and from position 30 to 31 both count as "one place." Shopper attention doesn't work like that, as the research below shows. A plain average also gives "dog crate" and "dog crate for 50 lb dog" the same say, regardless of how many people search each.
So a seller can see the average rise from 8.5 to 4.7 in a week where traffic fell. The worked example later in this article shows exactly how.
Share of voice: where the idea comes from
Marketing measured visibility as a share long before e-commerce:
- Share of voice (SOV) classically means a brand's share of all advertising in its category. John Philip Jones argued in Harvard Business Review (1990) that ad budgets should be set by comparing a brand's market share with its share of voice.
- SOV versus share of market. The IPA's summary of Les Binet and Peter Field's effectiveness research puts the pattern simply: when SOV exceeds share of market, brands tend to grow; when it falls below, they tend to shrink.
- Impression share is the digital version. Google Ads defines search impression share as impressions received divided by the estimated impressions you were eligible to receive. Amazon Ads launched a similar metric in 2021, top-of-search impression share, for Sponsored Products and Sponsored Brands.
Two cautions. First, the SOV research is about paid advertising and is correlational; applying it to organic visibility on Amazon is an analogy, not a proven law. Second, none of these definitions is Amazon's formula for organic search. Amazon doesn't publish one. What follows is a measurement method, not a reconstruction of Amazon's algorithm.
What research says about attention by position
A visibility score needs a position weight: how much attention position 1 gets relative to position 5 or 25. Several lines of research inform that choice.
Web search: attention decays fast. In an eye-tracking study of Google results, Joachims and colleagues (SIGIR 2005) found that users "click substantially more often on the first than on the second link," even though they viewed both about equally often. They called this trust bias. They also saw a break around ranks 6 and 7, where results fell below the visible screen. Craswell and colleagues (WSDM 2008) perturbed a major search engine's rankings and found that a "cascade" model, where users scan top to bottom and stop at the first good result, explained early-rank position bias better than a simple decay curve.
Position has a causal effect. Correlation alone can't separate "ranked high because it's good" from "clicked because it's ranked high." Raluca Ursu (Marketing Science, 2018) used randomized hotel rankings at Expedia and found that rankings causally affect what consumers search, but don't directly drive the purchase once a product is examined. Visibility gets you considered; the listing still has to win the sale.
Amazon-specific evidence:
- Farronato, Fradkin and MacKay (NBER, 2023) tracked a browser-extension panel of Amazon shoppers. In 72.1% of searches, users did not click past the first results page, and scroll data suggested only about half of the products on a full page were actually seen. It's a small panel (184 users), but the direction is clear.
- Rock, Strauss, O'Reilly and Mazzucato (Information Economics and Policy, 2024) analyzed results from more than 2,000 Amazon queries. Position, adjusted for on-screen prominence, strongly shaped clicks, and in the top five results, where typically four were ads, users were less sensitive to relevance and price.
Grids behave differently. Eye-tracking work on grid-based image search (Xie et al., WWW 2019) found a middle bias within rows, slower decay than in lists, and row skipping. That's image search, not Amazon, but it warns against assuming one curve fits both grid and list pages.
The honest summary: attention drops steeply with position, the shape depends on layout, and Amazon publishes no click-through curve. Any weight you choose is an assumption. Pick a reasonable one and keep it fixed.
The four ingredients of a visibility score
1. Position weight. Information retrieval offers ready-made curves. Discounted cumulative gain (Järvelin and Kekäläinen, ACM TOIS 2002) divides each result's value by the logarithm of its rank, commonly 1 / log2(position + 1). Rank-biased precision (Moffat and Zobel, 2008) uses geometric weights: the user looks at each next result with a fixed probability. Log weights decay gently; geometric weights decay hard. Either is defensible if used consistently.
2. Demand weight. A top position on a high-demand keyword is worth more than the same position on a niche phrase. Use coarse demand tiers rather than exact volumes, which are estimates anyway. Brand Analytics' Search Query Performance report includes search query volume for queries relevant to your brand, which helps set tiers. Our guide on how many keywords to track covers building and tiering the set.
3. Coverage. The share of keywords, or of demand, where you appear on page one at all. Missing keywords count as zero visibility, which fixes the selection bias.
4. Organic only. Amazon Ads says sponsored products can appear "at the top of, alongside, or within shopping results." Ads depend on bids and budgets. Mixing them into an organic score means your visibility moves when a competitor changes a bid. Track sponsored share separately if you need it, as explained in how an Amazon rank tracker works.
Step by step: a simple visibility score
The numbers below are an illustrative example for a generic "lunch box" product, not real data. Weights use the log curve, demand tiers are Core = 4, Important = 2, Standard = 1, and a keyword where you don't appear on page one scores 0.
| Keyword | Demand tier (weight) | Organic position | Position weight 1/log2(p+1) | Weighted |
|---|---|---|---|---|
| lunch box | Core (4) | 9 | 0.301 | 1.204 |
| lunch box for adults | Core (4) | 3 | 0.500 | 2.000 |
| insulated lunch box | Important (2) | 2 | 0.631 | 1.262 |
| bento lunch box | Important (2) | not on page one | 0 | 0 |
| lunch box for kids | Standard (1) | 20 | 0.228 | 0.228 |
- Record the organic position for every keyword in a fixed set, from the same location and setup each time.
- Convert positions to weights with your chosen curve. Missing keywords get 0.
- Multiply by demand weight and add up: 4.694.
- Divide by total demand weight (4 + 4 + 2 + 2 + 1 = 13): 4.694 / 13 = 0.361, so a score of 36 out of 100, where 100 means position 1 on every keyword.
- Report coverage alongside it: you appear on 4 of 5 keywords (80%), or 11 of 13 demand points (85%).
Now the trap. Next week you drop off page one for "lunch box for kids" and nothing else changes. The average position of ranked keywords improves from 8.5 to 4.7, because the worst rank left the average. The visibility score falls from 36.1 to 34.4. The score tells the truth; the average doesn't.
From score to share of page one
A score out of 100 tells you how close you are to the theoretical maximum. A share tells you how much of the available attention you hold relative to everyone else.
To get it, apply the same weights to every organic product on the page. For each keyword, divide your weighted attention by the sum of weights across all organic positions on that page, then demand-weight across keywords as above. The shares of all products on a keyword add up to 100%.
This also fixes a layout problem. Grid and list pages show different numbers of results, so "position 12" means different things. Scoring each position against the whole first page of that keyword keeps keywords with different layouts comparable. If you track variants, add up the family, since shoppers see child ASINs of one product as the same choice.
Pitfalls
- Layouts. Position weights from list-style web search don't map cleanly onto grids. Prefer per-keyword normalization over raw positions, and don't compare raw scores across keywords with different layouts.
- Personalization and location. Amazon's help page says Featured results consider "customer preferences, for example, items you previously purchased or clicked on." A snapshot from one fixed, signed-out setup is a consistent reference, not what every shopper sees.
- Unweighted averages. Equal weights let long-tail noise outvote your core terms.
- Changing keyword sets. Adding or removing keywords changes the score even when nothing on Amazon moved. Freeze the set for comparisons, and recompute history on the new set when you change it.
- Switching the weight curve. A score built on log weights and one on geometric weights aren't comparable. Choose once.
- Single snapshots. Rankings wobble day to day. Judge trends over a week or more.
- Treating it as Amazon's number. Brand Analytics' Search Query Performance reports your brand's impression share and click share per query, measured on real shopper sessions. It's a useful cross-check, but it only covers your own brand and differs in method from a page-one snapshot.
Using visibility to compare competitors
Visibility becomes most useful when you compute it for every relevant brand on the same keyword set with the same weights.
- Rank the market by share. Who holds the most weighted attention in your niche, and how concentrated is it?
- Decompose changes. When a competitor's share rises, check whether it came from coverage (new keywords on page one), position (moving up on existing keywords) or demand mix (gains on high-demand terms). Each points to a different cause, as described in how to track competitors on Amazon.
- Compare share of page one with share of sales. Borrowing the SOV logic, a competitor whose visibility runs well ahead of its sales may be investing to grow, and one whose visibility lags may be living off past momentum. Treat this as a hypothesis to check, not a rule.
- Watch trends, not levels. A stable 12% share is less alarming than one that fell from 18% in a month.
How JustRanks measures visibility
JustRanks records the whole organic first page of Amazon.com search results for more than 200,000 keywords over time, from a fixed US delivery address, overnight US time, and ads never count. Keywords saved in any user's market are captured every day, all others every 48h. A Market combines up to 100 keywords per product to show visibility across a niche; positions are scored on the whole first page of each keyword, so grid and list pages stay comparable, and variants are grouped as one product family. Head-to-Head comparison and Signals cover competitor moves, and no Seller Central access is needed.
Sources and further reading
- What are impressions, position, and clicks? (Google Search Console Help)
- Ad Spending: Maintaining Market Share (Jones, Harvard Business Review, 1990)
- The Long and the Short of It: presentation (Binet & Field, IPA)
- Top-of-search impression share metric (Amazon Ads, 2021)
- Amazon Brand Analytics overview (Sell on Amazon)
- Accurately Interpreting Clickthrough Data as Implicit Feedback (Joachims et al., SIGIR 2005)
- An Experimental Comparison of Click Position-Bias Models (Craswell et al., WSDM 2008)
- The Power of Rankings (Ursu, Marketing Science, 2018)
- Self-Preferencing at Amazon: Evidence from Search Rankings (Farronato, Fradkin & MacKay, NBER, 2023)
- Behind the clicks: Can Amazon allocate user attention as it pleases? (Rock et al., 2024)
- Cumulated Gain-Based Evaluation of IR Techniques (Järvelin & Kekäläinen, ACM TOIS 2002)
- Searching and Browsing for Items: Featured factors (Amazon Help)