> ## Content Index > Fetch the complete content index at: https://cascader.io/blog/llms.txt > Use this file to discover other available public pages before exploring further. # Who the Top Slot Moves - URL: https://cascader.io/blog/who-the-top-ad-slot-moves/ - Published: 2026-10-27T14:00:30.000Z - Updated: 2026-10-27T14:00:30.000Z - Description: Same-ad studies shrink the top slot's click gain to a fraction of the raw curve. Who it moves, what a new visitor costs, and how to test it. - Author: Rick Hecker - Tags: Cascader Research *Between the top two ad slots, a higher position mostly changes how many people look at you, not what a click is worth. In proportion, it adds least for the name people already came to find.* --- In Google Ads you can buy the first ad slot directly. Choose [Target Impression Share](https://support.google.com/google-ads/answer/9121108?hl=en&ref=cascader.io), set the placement to "Absolute top", and pick a percentage. Google then sets your bids "to help show your ads" there that often. Its [help pages](https://support.google.com/google-ads/answer/6381002?hl=en&ref=cascader.io) give the reason in one sentence: "Generally, an ad gets more clicks as it moves to a higher ad position." Top ads "typically" cost more per click, because they must clear [higher Ad Rank thresholds](https://support.google.com/google-ads/answer/1722087?hl=en&ref=cascader.io). The click curve in any account seems to confirm it. [Matthew Goldman and Justin Rao](https://research.chicagobooth.edu/~/media/70F205DE79214CA396DBA2ECC2304B94.p?ref=cascader.io) studied the twenty thousand highest-revenue ad-and-query combinations on Bing over four months of 2013, when Bing showed up to four ads above the results. On average, the first ad was clicked 32% of the time and the second 9%. That 23-point gap mostly compares different ads. The ads that win the first slot tend to be ones people click more wherever they appear. A brand's own ad on a search for its own name is the vivid case: it sat in the first slot in about four of every five of its impressions. To see what the slot itself does, you have to compare an ad with itself. Goldman and Rao could sometimes do that. Microsoft ran routine experiments on its ranking methods, assigning users at random, and when two methods disagreed about an ad, the same ad landed in different slots for different people. For 783 of the twenty thousand pairs this happened often enough to measure. There, moving from second to first added about seven or eight percentage points of click-through: about a third of the 23-point gap. [Sridhar Narayanan and Kirthi Kalyanam](https://doi.org/10.1287/mksc.2014.0893?ref=cascader.io) found the same shape on Google. They had nine months of daily bids, quality scores and outcomes for four large online retailers. That let them compare days when a retailer barely won the higher slot with days it barely lost. A naive comparison said the top slot more than doubled click-through. Their estimate was a gain of about a fifth. So the slot does something real, and much less than the curve suggests. To see what it does, and to whom, it helps to look first at how people search. ## Two kinds of searcher In 2011 [Kinshuk Jerath and colleagues](https://business.columbia.edu/sites/default/files-efs/pubfiles/6058/position%5Fparadox.pdf?ref=cascader.io) built a model of a search page with two kinds of people on it. Some already know which firm they want and go to it "irrespective of its position". Others do not, and start with whatever is on top. For a firm the first group already seeks, paying extra to be on top buys nothing from them. Jerath and colleagues called the added fee on those clicks "pure waste, because these consumers will click on the firm's link anyway." The clearest evidence for the second kind of searcher comes from outside advertising. For eight months in 2012–13, Expedia showed about a third of its hotel searchers their results in random order. [Raluca Ursu](https://doi.org/10.1287/mksc.2017.1072?ref=cascader.io) used that experiment. Rank told those searchers nothing about quality, yet the first hotel was clicked about 12% of the time, the second about 9%, and the tenth about 4.5%. People were barely looking. The average page held 27 hotels, and even among searches with at least one click, only 7% produced a second. Ursu tested position two ways: as a steady slope down the list, and one position at a time. Both found that rank changed which hotels people opened. Neither found a significant effect on whether they booked once they had opened one. Her reading, [now standard in research on online search](https://www.anderson.ucla.edu/sites/default/files/document/2024-03/HonkaSeilerUrsu%5F2024.pdf?ref=cascader.io), is that a higher place lowers the cost of looking. It makes something easier to examine, not more attractive. Put the model and the experiment together and you get a plain account of the top slot. It is a convenience. It helps people who were going to look at whatever came first. It does little for people who already knew where they were going. ## Who it moves Goldman and Rao's data let that account be tested on ads. They sorted ads into three kinds. For each kind they plotted every ad's gain from moving to the top against how often it was clicked in second place, and drew a line through the points. The simplest line, drawn through zero, reads as one proportional gain per kind: If columns extend beyond the page, scroll sideways. With a keyboard, focus the table and use the arrow keys. | Ad and query | Example | Proportional gain from moving to the top (line through zero) | | ---------------------------------------------- | ------------------------------------- | ------------------------------------------------------------ | | A brand's own ad, on a search for its own name | a Samsung ad on "Samsung smart phone" | about 30% | | An ad on a generic product search | any ad on "smart phone" | about 45% | | A rival's ad, on a search for another brand | a Nokia ad on "Samsung smart phone" | about 62% | The pattern fits the two kinds of searcher. A search for "Samsung smart phone" is full of people already headed for Samsung, so Samsung's own ad gains least from being first. A Nokia ad on that search has no such head start. For many of those people it is visible only if it is in the way, so it gains most. Not every gap in that table is equally firm. The brand row rests on 91 ad-and-query pairs, and a line through zero is only one way to fit the points. Goldman and Rao's second line adds a fixed amount for every ad of a kind, whatever its second-slot rate. That means ads that start low gain proportionally more. Brands start high on their own names: in the pooled figures, a brand's ad in the second slot was clicked roughly a quarter of the time, and other ads about a tenth. At those usual rates, the second line gives brands about 42%, generic ads about 50% and rivals about 69%. Under either line, at each kind's usual rate, the brand gains least in proportion. Brand and generic would be level only if a generic ad started as high as a brand. The firmest gap is brand against rival. [Goldman's dissertation](https://escholarship.org/content/qt2hb1942q/qt2hb1942q.pdf?ref=cascader.io) tests it within the same query strings. On 32 brand queries where both could be measured, rivals gained more on 25, and the difference was individually significant on 12 of those. Of the queries that went the brand's way, only one was significant. Among ads that were not the searched-for brand, more-visited sites gained less from the top, which led Goldman and Rao to call popularity "a substitute" for position. On generic searches, the top slot also did more for sites whose visitors generally stay rather than leave at once (a site-wide traffic measure). The authors' label is that quality is "a compliment to position" (their spelling). Both findings are about how many people click, not about what their clicks are worth. Two older studies point the same way. On Microsoft's search engine in 2007, [Przemysław Jeziorski and Sridhar Moorthy](https://www.jeziorski.me/files/search%5Fprominence.pdf?ref=cascader.io) followed the same retailers' ads as their positions shifted, sometimes within minutes. On camera-brand searches, they report, a retailer outside Alexa's top hundred websites got 30–50% more clicks in the first slot than in the second. A top-hundred retailer got 0–13% more. Narayanan and Kalyanam's smaller retailers, and their keywords that had historically drawn few clicks, showed larger gains than their largest retailer and their high-click keywords, though those subgroups cannot be statistically told apart. [Stanford's business school](https://www.gsb.stanford.edu/insights/sridhar-narayanan-google-search-ads-higher-isnt-always-better?ref=cascader.io) headlined its account of Narayanan and Kalyanam's paper "Higher Isn't Always Better", adding that "especially for well-known advertisers, a higher position doesn't always pay." The direction was right. The evidence that carries it comes from Bing. ## The brand's extra clicks There is a catch in reading this as "the top slot does nothing for famous names". Those high starting rates cut the other way in raw clicks. Apply each kind's lines at its usual rate, and the brand's ad gains roughly 7.5 to 10 extra clicks per hundred impressions. Other ads gain roughly 4.5 to 7\. The brand gains as much as anyone, or more. Its proportion is small only because its base is large. So the top slot does produce extra clicks on the brand's own ad. The question is who those people are. On a search for your own name, your site usually sits in the ordinary results just below the ads. Some of the searchers who click your ad when it is first would, with the ad lower, have clicked your ordinary listing instead. They are the informed searchers of Jerath's model, changing their route rather than their destination. How many do that, no study of ad position measures. [Cascader's earlier piece on pausing brand keywords](https://cascader.io/blog/what-a-brand-ad-pause-actually-measures/) takes up the related question of advertising on your own name at all. Google's first worked example for Target Impression Share, a brand showing "100% of the time" on its own terms, is about appearing, not about being first. Placement is a separate setting. ## Are the added clicks worth less? If the top slot brings in people who would not otherwise have looked, it is natural to suspect they are browsers who buy less. Between the top two slots, the best-identified evidence says hardly. Goldman and Rao measured a click's value by how long the searcher stayed away before returning to the results page (if they returned at all), calibrated against conversions. Moving an ad from the second slot to the first *raised* that value by about 1.7%. Their estimates ranged from −0.4% to +3.3%, depending on method. That is the average across all of the ad's clicks, old and new, and the top slot added 30% to 69% more clicks in their data, depending on the kind of ad and the line. For the average to move so little under that many additions, the added clicks must be worth close to what the existing ones were. Close, here, means a few percent, mostly on the plus side. Assuming the clickers the ad already had behaved the same, the added clicks were worth about 4–7% more than the existing ones at the central estimate, and between about 2% less and 14% more across the estimates. The authors caution that position changes who clicks, so this is not an effect on any fixed group of people. That fits what George Michie, then at the Rimm-Kaufman Group, [wrote in Search Engine Land in 2008](https://searchengineland.com/why-position-bidding-wastes-money-14841?ref=cascader.io): "the value of traffic doesn't vary by position." His was an unaudited claim from his firm's own study. The single-advertiser academic studies disagreed with each other. - **[Agarwal, Hosanagar and Smith](https://www.cmu.edu/entertainment-analytics/documents/technology-impact-on-entertainment/location-location-location.pdf?ref=cascader.io)** randomised one online pet-food retailer's bids in 2009 and found conversion rising lower down. Their estimate came through a model built on roughly fifty to eighty orders. A second, non-randomised data set from another retailer pointed the same way. - **[Anindya Ghose and Sha Yang](https://pages.stern.nyu.edu/~aghose/paidsearch.pdf?ref=cascader.io)**, using an observational model over ranks 1 to 131 for a large chain, found conversion higher at the top. One signal still points against "no difference". In Ursu's random hotel list, the first position's conversion per click is the lowest of the first four, in her figure and in her position-by-position estimates. None of those estimates is significantly different from her reference position, and the paper reports no test comparing the first position with the next three. The dip is unresolved, neither confirmed nor ruled out. It fits Michie's aside that position 1 tends to be "slightly lower than position 2". The top may catch a few people who would not otherwise have bothered. Account reports seem to say more than this because they compare different auctions, ads and queries. In Goldman and Rao's pooled figures, conversion was 3.3% in the first slot and 3.9% in the second. Yet the same ad's clicks were worth slightly *more* in the first. Google makes a similar point about [its top and absolute-top impression rates](https://support.google.com/google-ads/answer/7501826?hl=en&ref=cascader.io), which "may decrease as bids increase … higher bids may allow you to enter more competitive auctions in a worse location." Further down the page the picture changes. When an ad moved up from Bing's old right-hand sidebar into the fourth top slot, the same ad's clicks were worth 6–9% less, by non-experimental estimates. The authors suggest this "may indicate that users who have gone to the trouble of finding a sidebar ad, simply have a higher purchasing intent." Google's desktop pages no longer have that sidebar. Between the top two slots, the people the slot adds behave much like those already arriving. ## The price of a new visitor That makes the cost side a matter of arithmetic, and it is easy to underestimate. Moving up adds clicks, but it also raises the price of every click, including the ones you were already getting from second place. Write the top slot's extra clicks as a proportion *g* of your current clicks, and its cost-per-click premium as a proportion *p*. Then each extra click costs your old cost per click × \[1 + *p* × (1 + *g*) / *g*\]. That is the extra spend divided by the extra clicks. Now carry the two kinds of searcher into it. If a share *s* of those extra ad clicks would have reached you anyway through your ordinary listing, each genuinely new visitor costs that figure ÷ (1 − *s*). **A constructed example.** The premium of 30% and the shares are illustrative inputs. The gains are Goldman and Rao's, on both lines, each at its kind's usual starting rate. The table gives the price of each extra click and each new visitor, in multiples of the old cost per click. If columns extend beyond the page, scroll sideways. With a keyboard, focus the table and use the arrow keys. | Ad and query | Gain *g* | Per extra click (*s* \= 0) | Per new visitor if *s* \= 25% | Per new visitor if *s* \= 50% | | -------------------------------------------- | -------- | -------------------------- | ----------------------------- | ----------------------------- | | Brand's own ad, own name (line through zero) | 30.0% | 2.30× | 3.07× | 4.60× | | Brand's own ad, own name (second line) | 41.8% | 2.02× | 2.69× | 4.04× | | Generic product search (line through zero) | 45.0% | 1.97× | 2.62× | 3.93× | | Generic product search (second line) | 50.2% | 1.90× | 2.53× | 3.80× | | Rival on a brand search (line through zero) | 62.0% | 1.78× | 2.38× | 3.57× | | Rival on a brand search (second line) | 68.9% | 1.74× | 2.31× | 3.47× | Two things in this table are evidence and two are not. The gains are measured, though on 2013 Bing pages. Under either line, the brand's gain is the lowest, and its gap with rivals is the firmest. The premium and the share are yours. No source here measures the premium for moving one ad up a slot, for any kind of ad. [Google says](https://support.google.com/google-ads/answer/1722122?hl=en&ref=cascader.io) higher-quality ads "can often lead to lower CPCs". Goldman and Rao found that first-slot clicks brought Bing the least revenue per click, because under the auction's pricing rule more clickable ads pay less per click. Both statements concern the level of cost per click across ads, not the premium for moving up. Which way that premium runs for a brand on its own name is unmeasured. The share is the quantity the two kinds of searcher turn on, and no position study supplies it either. What the arithmetic does show is how the pieces combine. A small proportional gain makes each extra click expensive, and a large rerouted share makes each new visitor more expensive still. On a search for your own name, the first is firm against rivals. The second is plausible but unmeasured. Two single-retailer studies, both weakly identified, tried to carry position through to profit, and neither found the top most profitable. - **Agarwal and colleagues** found lower positions more profitable for almost all of one retailer's keywords. That was at an assumed $10 average product value, on the thin data noted above. - **Ghose and Yang** found profits often highest in the middle positions. The top slot pays on a keyword when a new visitor there is worth more than that price, worked out with your own premium and your own share. ## What the slot is for The practical question an advertiser can ask of any keyword is which kind of searcher it mostly brings. Three things in an account bear on it: - **Whether the search names you.** - **Whether your ordinary listing sits just under the ads for it.** - **What happens to visits from paid and ordinary results combined when the placement changes.** The first two tell you which way to lean. The third is the only way to learn the share, and it needs a control. A simple before-and-after comparison would fold in season, demand and rivals' bids, the same kind of confusion that inflates the click curve. Instead, switch the placement on and off across alternating comparable periods, or apply it in some matched regions or campaigns and not in others. Then compare total visits against the side left unchanged. [Cascader's earlier piece on pausing brand keywords](https://cascader.io/blog/what-a-brand-ad-pause-actually-measures/) describes that kind of design. Michie's advice in 2008 was to bid what the traffic is worth and let position land where it may: "position 1 — that's great, position 6? Okay." The research since has largely borne out his view of value, at least for the top two slots. It has complicated his calm about position, because position decides how many people look at you, and that depends on who they are. Most searchers look at very little. Some arrive knowing where they want to go and find it wherever it sits. Others take whatever is put in front of them. The top slot doesn't make anyone want you more; it makes you easier to look at. For people already looking for you, that mostly changes the route they take, and what you pay for it. For people who weren't, it can be the difference between being seen and being skipped. --- *A note on the evidence.* The causal estimates come from pages that no longer exist: Microsoft's engine in 2007, Google before 2016, Bing in 2013\. Google removed the right-hand column of desktop ads in 2016 and allowed up to four ads above the results. That account is second-hand, from [a Google spokesperson quoted in Search Engine Land](https://searchengineland.com/google-no-ads-right-side-of-desktop-search-results-242997?ref=cascader.io) and [a 2020 report by the UK's competition regulator](https://assets.publishing.service.gov.uk/media/5fa557668fa8f5788db46efc/Final%5Freport%5FDigital%5FALT%5FTEXT.pdf?ref=cascader.io). The pattern of who gains is consistent in direction across three engines and three study designs, but its statistical weight comes from one study, Goldman and Rao's. That study exists as a 2014 working paper and in Goldman's 2015 dissertation, not as a journal article. Narayanan and Kalyanam's figures come from a 2014 working version: its table gives the top-slot gain as 20.6%, its text says 22.1%, and the published values may differ. Goldman and Rao's click value is a proxy built from return-to-results timing, not purchases. No study measures what any searcher actually knew; "brand on its own name", traffic rank and historic click-through stand in for it.