What a Bigger Giveaway Prize Buys

100% We looked at 16,561 competitions run by 3,651 businesses started between 1 September 2025 and 13 August 2026, plus 165,204 campaigns since 2021, to give giveaway organizers a detailed look at how other brands run competitions. We used our own platform data to show what prize money actually buys, and how far two free decisions carry a campaign: how many entry Actions the widget asks for, and whether the business runs a second one. Every figure below is a median, and every table names its sample and its window.

1. What we found
2. What other people spend
3. $65 to $645, the prize gap between plans
4. What the prize money buys
5. The two levers that move Conversion most
6. Inside the top decile
7. The second campaign beats the first
8. Duration and Action count, flat across every prize band
9. Why two thirds leave the value blank
10. What is moving in the market
11. What to change on your next campaign
12. Benchmark tables
13. Frequently asked questions
14. How we did this
15. Cite this research

Organizers ask us this more than anything else: is my prize budget reasonable? So the study starts there, with the distribution to check your own budget against. The typical competition that records a prize value records $300. A quarter spend under $90. A quarter spend over $1,000. One in ten goes above $3,000.

Prize budgets span three orders of magnitude, so the middle is narrow and the tail is long. The 25th percentile (p25), a quarter of campaigns spend less than that and three quarters spend more. P90 means nine campaigns in ten spend less. Campaign-level total, prize value times quantity, US dollars. n = 6,322 campaigns across 1,798 businesses, 1 September 2025 to 13 August 2026. The scale is logarithmic, so each gridline is ten times the one before it.

The figure has barely moved in six years. The median has sat between $300 and $528 in every year since 2020, across 167,072 campaigns. We found no prize inflation across those six years.

Start with what each tier looks like. The pattern is consistent. The bigger the plan, the bigger the prize, and the more people who enter.

Table 1 · Median campaign profile by plan, 1 Sep 2025 to 13 Aug 2026. Every figure is a median.

Entered is the share of widget views that converted to an Entrant. Impressions counts renders and not unique people, so this is a rate per view. Premium is suppressed at 47 businesses, below our 50-business minimum, and the 18.9 percent is shown for completeness. One campaign carries no plan and sits outside this table, so the rows sum to 16,560 of 16,561. Every other row clears 200 campaigns and 50 businesses on every column.

Two things move together down that table, in opposite directions. Entrants rise 3.4 times from Free to Business. The share of people who enter falls 8.7 points.

A Business campaign draws 34 percent more Entrants than a Pro campaign, 769 against 574. But once prize budget is held constant, most of that difference disappears.

Table 2 · Business against Pro at the same prize budget, median Entrants

Every cell clears 200 campaigns and 50 businesses on both sides.

Above $500 the two tiers are indistinguishable. Business campaigns draw more people because Business customers buy bigger prizes and promote harder, and the plan is a marker of that rather than the cause of it. The gap survives the prize control in one place only, the cheapest band, where a Business campaign still draws 21 percent more.

Read the last two columns of table 1 together, because either one alone gives the wrong answer. A Free campaign converts 35.9 percent of its widget views and draws 225 Entrants. A Business campaign converts 27.2 percent and draws 769. Even though the conversion rate falls by 8.7 points, the absolute number of contestants rises 3.4 times.

That is what you would expect when a larger business promotes a campaign to a broader, colder audience, while a smaller business reaches people who already know or follow them. Because Conversion Rate uses total reach as its denominator, the rate can fall as promotion expands even while the campaign attracts more entrants overall.

The reach column helps explain the pattern. Among campaigns with at least 100 Impressions, the median Free campaign generated 849 Impressions, while the median Business campaign generated 3,198. Impressions increased 3.8 times, compared with a 3.4 times increase in Entrants. When the denominator grows faster than the numerator, the Conversion Rate falls, which is consistent with the 8.7-point decline.

We should not attribute the entire decline to reach, though. Each column is an independent median calculated from a different set of campaigns, so the values cannot be divided directly into one another.

This tells us which number matters in each situation. When two campaigns reach roughly the same number of people, Conversion Rate is the better comparison because it shows how effectively the campaign turns that audience into Entrants. When reach is very different, the Entrant count matters more.

By that measure, the median Business campaign attracts 544 more Entrants than the median Free campaign, 769 compared with 225. Its lower Conversion Rate reflects the broader audience it reached, not necessarily a weaker campaign. Section 5 is where Conversion Rate becomes more useful, because it holds reach more constant and looks more directly at how the widget performs.

We are not using Premium as a benchmark in this comparison. Only 47 businesses are represented, below our 50-business minimum. The 18.9 percent Conversion Rate remains in the table for completeness, but it should not be treated as a benchmark.

Part of the remaining difference comes from the widget itself. A Free campaign asks for a median of 5 Actions, while a Business campaign asks for 7. Section 5 looks at how that affects performance. The rest comes from audience quality and reach, neither of which changes simply because a business changes plan.

So the answer to the question at the top of this section is that plan level predicts results, but does not create them. Once prize budget is held constant, the remaining difference comes mainly from the audience a business can reach and the entry Actions available to it. Section 5 shows that those Actions do not perform equally, while Section 4 shows that reach separates large campaigns from small ones by more than prize value does.

Building a larger audience takes time. Borrowing one through a partner, creator or paid promotion can expand reach much faster.

Bigger prizes do bring more people. The useful question is how many more, for how much.

Table 3 · What each prize band costs and what it draws, 1 Sep 2025 to 13 Aug 2026. Every figure is a median.

All four bands clear 200 campaigns and 50 businesses, so the whole range is reportable. Prize value per Entrant is the campaign's stated prize value divided by its Entrants, taken as a median across campaigns. The four rows sum to the 6,322 campaigns that state a value.

Read across the two bands a planner actually chooses between. Increasing the prize 22 times, from $210 to $4,590, came with 3.2 times the Entrants, from 471 to 1,499. Across the full span it is worse: 115 times the budget, from $40 to $4,590, for 4.7 times the people.

The useful way to hold that is as a unit price. One Entrant carried $0.44 of stated prize value in the $100 to $499 band and $3.35 in the most expensive one. That is a figure you can set against what you currently pay for a click, a lead or a follower through ads. Prize money is one input to a campaign's real cost, and Gleam's guides cover the rest of the arithmetic: how to calculate a giveaway's CAC and giveaways against paid ads, a blended CAC framework. Nothing in this report is a CAC figure. This is the stated value of the prize and not what it cost the business to supply, which for a company giving away its own product is usually a different number.

There is no point at which the prize suddenly starts working. Every extra dollar came with fewer people than the dollar before it.

The last column drops as the budget rises, which looks like bad news and is mostly arithmetic. Expensive campaigns are promoted harder, so far more people see the widget. The number who enter goes up. The number who see it goes up faster.

Both bars grow with the prize budget, and the outer bar grows faster. Median Impressions rise 317 percent from the $100 to $499 band to the most expensive one. Median Entrants rise 218 percent. Impressions rising faster than Entrants is the direction of the Conversion fall, though the two medians come from different campaigns and cannot be divided into each other to size it. n = 4,698 campaigns in the three bands shown, from 1,563 businesses.

Which means a falling Conversion Rate on an expensive campaign describes a wider audience, and it is a poor signal for whether the campaign worked. Judge that campaign on how many people entered.

If prize budget is a weak lever, something else is carrying the difference between a campaign that draws 150 people and one that draws 2,600. It shows up if you sort every campaign by how many people entered and cut the list into four equal groups.

Table 4 · Campaigns sorted into four groups by size, 1 Sep 2025 to 13 Aug 2026

n = 16,542 campaigns across 3,647 businesses with at least 100 Impressions. Impressions is Gleam's name for Widget Views, the denominator used in this report.

The biggest campaigns drew 17 times as many Entrants as the smallest. Their Conversion Rate is the same: 29.8 percent at the top against 28.9 percent at the bottom. The entire gap is Impressions, 10,048 against 527.

Table 4 is the reason Impressions is the number to move. Prize money buys them at the prices in table 3, $0.44 to $3.35 each. Your email list, your own following, a partner's audience and paid promotion are the other ways, and for most businesses they are cheaper.

Campaigns asking for one to three Actions converted 44.0 percent of the widget views they received. Campaigns asking for thirteen or more converted 23.8 percent. That is a 20.2 percentage point spread, against 5.5 points across the whole prize range.

Table 5 · Entry Actions against Entrants and Conversion, 1 Sep 2025 to 13 Aug 2026. Every figure is a median.

The Entrant column rises, and it is a trap. Campaigns with long widgets are also promoted harder: median Impressions climb from 1,410 to 2,644 across the same rows. The extra people come from the extra reach.

Hold the reach still and the direction reverses.

Table 6 · Median Entrants by Action count, inside bands of Impressions

Every cell clears 200 campaigns and 50 businesses. The comparison also survives keeping one campaign per business and running inside the Pro tier alone.

At the same reach, a campaign asking thirteen Actions ends up with 35 to 41 percent fewer Entrants than one asking three. Only the smallest campaigns, under a thousand Impressions, show no difference, and those are too small for the gap to show.

Conversion falls by nearly half across the range. The share of people who enter runs from 44.0 percent at one to three Actions down to 23.8 percent at thirteen or more, while Entries per Entrant climbs from 1.3 to 9.5. n = 16,542 campaigns across 3,647 businesses.

So the trade is clear. Each Action you add is a filter on the people who were going to enter anyway, and a multiplier on how many times the survivors enter. If entries are a scoring mechanism for a draw, the extra entries are worth nothing to you and the lost people are real. If each Action is an outcome you actually want, a follow you need or an email you are collecting, then the Conversion cost is the price of that outcome.

The tables above say nothing about which Actions to keep. So we held the number of Actions still and compared campaigns that used a given Action against campaigns at the same widget length that did not. Among campaigns asking for 4 to 20 Actions, the choice of Action moved Conversion across a 10.9 percentage point range.

Table 7 · Conversion difference associated with each Action, holding the number of Actions constant

Each row compares campaigns carrying that Action against campaigns with the same number of Actions that do not. Bands of 4 to 5, 6 to 8, 9 to 12 and 13 to 20 Actions are compared separately and combined by campaign count, so a longer widget is never compared with a shorter one. Every row clears 1,000 campaigns and 100 businesses, and no business holds more than 10 percent of any row. Six further Actions were measured and suppressed for failing the dominance rule.

An earlier version of Table 7 carried a second column showing how many more or fewer Entrants came with each Action, and reported that an Email Subscribe came with 89 percent more. Holding the number of Actions constant does not hold reach constant, and campaigns asking for an email drew 83 percent more Impressions before anyone entered anything. Repeating the comparison inside bands of Impressions, that 89 percent becomes plus 6, plus 24, minus 3 and plus 2 percent across the four bands. It changes sign, so no single figure represents it.

The same test applied to every other row produced the same result. Viral Share moved from 76 percent to plus 2, plus 9, minus 7 and minus 7. Twitch Follow moved from minus 43 percent to minus 2, minus 21, plus 2 and plus 14. The Entrants column measured how widely a campaign was promoted, so we removed it. A caveat on that column would not have traveled with it.

The Conversion column survives that test better, because its denominator moves with its numerator. Three rows held their direction in all four Impression bands: X Follow positive, Viral Share negative and Chat Members negative. The rest changed sign in at least one band, so read this column as a ranking that still needs confirming and not as twelve settled figures. The full band-by-band test is scheduled for the next edition.

So the length of the widget is only half the decision. Two campaigns can ask for the same eight Actions' worth of effort and land 11 points apart on Conversion, because of which eight they picked.

One practical constraint sits behind this table. Which Actions you can use at all depends on your Gleam plan, so the tiers in table 1 differ in which Actions they make available as well as in what their customers spend. Across 27,256 Free and Hobby campaigns since 2021, none used an Email Subscribe, because the Action is not available on those plans. That is a statement about availability and says nothing about outcome. If a row here looks worth having and is not in your dashboard, check the plan before you change the widget.

What this table does not settle is why. Businesses that ask for a given Action differ from businesses that do not, in audience, category and what they are running the campaign for, so part of every row is who chose it. Nothing here says the same Action would behave that way on your campaign. It does say that Actions are not interchangeable. Section 10 shows organizers have been quietly rebuilding these stacks for six years.

Sort every campaign in the window by how many people entered and take the top tenth. Those 1,656 campaigns, run by 559 businesses, collected 57 percent of all Entrants on the platform. The top fifth collected 72 percent.

Table 8 · What campaigns look like at each level of the distribution

Prize is the median stated value among the campaigns in each group that record one: 629, 637, 2,018 and 3,038 campaigns respectively.

Read down the Actions column. Seven, seven, eight, six. The biggest campaigns and the smallest ask for almost the same amount of work. Read down the Days column. Fifteen, sixteen, fifteen, thirteen. They run for the same length of time.

Two columns do move. The prize is ten times larger at the top, $1,600 against $150. Impressions are twenty-four times larger, 20,195 against 825. And the Conversion Rate is flat across all four rows, 30.9 against 27.9.

So a top-decile campaign is built the same way. The same widget, the same length, shown to twenty-four times as many people, with a prize ten times the size. Section 4 already priced that prize at $0.44 to $3.35 per Entrant, which accounts for a fraction of a twenty-four-fold difference in reach.

The obvious way to answer this is to take businesses with a long run of campaigns and line those campaigns up in order. That method gives a clear answer and the answer is wrong, for a reason worth setting out before the table, because it is the same reason most published benchmarks read high.

A business only has a tenth campaign if its earlier ones worked well enough to keep going. Selecting on a long run therefore selects businesses whose first campaign already succeeded. Measured across every business on the platform, a first campaign from one that went on to run ten or more medians 758 Entrants. Every other first campaign medians 405. So a sequence drawn only from the long-run businesses starts from an unusually high first campaign and has nowhere to go except down.

Here is the sequence across all 25,009 businesses instead.

Table 9 · Median outcome by campaign number, every business, since 2021

All campaigns from 2021 onward, so sequence position is real and not truncated by the twelve-month window. No business contributes more than 3.9 percent of the campaigns in any row. Conversion and Impressions use campaigns with at least 100 Impressions.

Entrants rise from 435 on a first campaign to 565 by the eleventh to twenty-fifth, which is 30 percent. Conversion rises alongside it the whole way, 25.8 percent to 28.0. Both metrics move together, so this is not one number improving at another's expense.

The last row turns down. Businesses past their twenty-sixth campaign median 514 Entrants, still above a first campaign and below their own eleventh-to-twenty-fifth. Two readings fit. Campaigns at that frequency may be smaller by design, run as routine promotions instead of events. Or an audience that has been asked twenty-six times does start to thin. This data cannot separate them, and either way the figure stays above where the business started.

An earlier internal cut of this section reported the sequence falling from 736 Entrants to 508, and recommended planning each campaign down from the last. That cut used only businesses with ten or more campaigns. The direction reverses once every business is included, and the recommendation reverses with it. We are stating this because the earlier figure circulated internally before publication.

Table 9 measures one campaign at a time, which answers a different question from the one a marketer is asking. Line the same businesses up by how many campaigns they have ever run and add up everyone who entered.

Table 10 · Total Entrants across every campaign a business has run, since 2021

All 25,065 businesses that started a campaign from 2021 onward. Median campaigns is the median within each band, so the last row describes a business that has run 47 campaigns and not exactly 26.

A business that ran once reached a median of 334 people. A business that has run 26 or more reached 58,506. The per-campaign figure and the accumulated figure both point the same way, which is the part the earlier cut of this analysis got backwards.

Cadence has a sharper effect than count.

Table 11 · Median outcome by gap since the business's previous campaign

A campaign launched within thirty days of the last one converts best of any group, 30.2 percent, and draws 16 percent more Entrants than a first campaign. A campaign launched one to three months after the last one draws the most people of all, 610, because it reaches 41 percent more of them.

So a tight cadence came with higher Conversion and a monthly-to-quarterly cadence came with more reach. Both sit above a first campaign and both sit above leaving it more than a year.

One consequence of all this belongs next to every benchmark on this page. Inside the twelve-month window, businesses on their eleventh campaign or later make up 27.4 percent of the businesses and 73.6 percent of the campaigns. The platform median of 511 Entrants describes them.

A first campaign inside the same window medians 362 Entrants at 26.6 percent. If you are planning your first, that is the pair of numbers to plan against, and the platform figures elsewhere on this page describe where a business gets to after several years of practice.

Two decisions look as though they should scale with the prize. Neither does, which means you can take both off the list of things to worry about when the budget changes.

How many Actions to ask for. The median campaign configures 7 to 9 Actions at every prize level from under $100 to over $2,500. There is no budget at which other organizers start asking for more.

How long to run. Median duration runs 22, 14, 15 and 21 days across the four prize bands, so it moves by eight days with no order to it. The cheapest prizes and the most expensive run the longest. Campaign length does not track Entrant numbers either. Set the run length from how long your promotion can keep pushing, and nothing in this data argues with that.

Table 12 · Duration and Action count against prize budget, 1 Sep 2025 to 13 Aug 2026

Every dollar figure in this report describes the 38.2 percent of campaigns in this window that record a prize value. That is worth understanding before you use any of them, because the third who fill it in are not a random third.

If recording a value were simply a sign of a careful organizer, the bigger plans would do it more. They do not. Business records a value less often than Pro.

Table 13 · Share of campaigns recording a prize value, 1 Sep 2025 to 13 Aug 2026

What separates them looks like whether the number tells the reader anything new. On a gift card the number is the prize, so it gets typed. On a games console most people already have a figure in mind. On a curated bundle or a trip there is no single price to type, and on a limited-run item a modest number would undersell something whose appeal is that you cannot buy it.

Work on price knowledge suggests the middle case is shakier than it looks. Dickson and Sawyer found that more than half of supermarket shoppers could not correctly name the price of the item they had just put in their cart. If people cannot price something they are holding, assuming they know what your prize is worth is optimistic.

Two practical consequences. First, the dollar tables here lean toward gift cards, cash and vouchers and away from hardware, experiences and merchandise, and we cannot correct for that. Second, filling the field in costs you nothing and gives Entrants the number they are guessing at anyway.

The obvious follow-up is which kind of prize draws the right people. The tables below get half way. They report what each vertical hands over and what that costs. They do not report what happens when the same audience is offered a different type, because almost no account in this window ran both. Prize type and audience move together in this data, so any figure comparing them would be reading the vertical as much as the prize.

Hold that next to what the money does buy. Twenty-two times the prize budget came with 3.2 times the Entrants, and every dollar after the first few hundred bought fewer people than the one before it. A lever that weak is unlikely to be the main thing separating a campaign that works from one that does not. The remaining candidates are prize fit, who you put the campaign in front of, and how much you ask them to do. This study measures the third one, and it is large. The first two are the next two studies in this series.

A dollar figure without a prize type is half a benchmark. So we classified prize names into categories and put the money next to the outcome. Only 47.1 percent of the 6,322 valued campaigns carry a prize name specific enough to classify. Our own rules require 70 percent. So the table below covers under half the platform, and should be read that way.

Table 14 · Prize type against value and outcome, 1 Sep 2025 to 13 Aug 2026

Eleven categories are listed with their counts and suppressed on every other column, because each falls below the 200-campaign or 50-business minimum. They are shown so you can see everything measured, including what did not clear a threshold. The median across all valued campaigns in the window is 550 Entrants, which is the baseline each row should be read against.

Two rows clear the thresholds and they say something useful. A gift card or cash prize runs a median of $125 and draws 449 Entrants, below the 550 baseline. Tech hardware runs $499, four times the money, and draws 1,030, which is 2.3 times the people and above the baseline. Hardware also converts better, 26.1 percent against 22.4.

Four times the money, 2.3 times the people. That is close to what the same step buys inside the cash range in section 4, where 4.8 times the budget came with 2.2 times the Entrants. Read the two rows as a description of the hardware and gift card segments rather than as evidence that the prize type is doing the work.

A platform median of $300 describes almost nobody. We classified 2,849 customer sites by hand against the 21 verticals Gleam publishes at gleam.io/industries, reading the domain of each account rather than matching keywords. That covers 95.2 percent of the campaigns in this window, and the spread it exposes is wider than any other cut in this study.

Table 15 · Prize budget and outcome by industry, 1 Sep 2025 to 13 Aug 2026

Nine verticals clear the 300-campaign and 75-business minimum. In five of them one account held more than 10 percent of the campaigns, so that account is removed from the row and from its campaign count: E-commerce, Cryptocurrency, B2B, Hobbies Crafts DIY and Esports. Removing it moved the median Entrants by less than 5 percent in three of the five. It moved Cryptocurrency from 458 Entrants to 343 and the B2B median prize from $629 to $200, which is why the rule exists. Four more verticals are listed with their counts and suppressed on every other column. Prize columns use the campaigns in each row that state a value, between 157 and 1,623 of them.

The median prize runs from $100 among content creators to $720 in sports, a seven-fold spread. Set that against the platform median of $300 and the planning error shows up. A creator budgeting $300 plans to spend three times what their peers spend. A sports brand budgeting the same plans to spend under half.

The outcome columns do not follow the money. Sports runs the highest prize at $720 and draws 1,122 Entrants. Cryptocurrency runs $513 and draws 343. Smart Wear Gadgets runs $430, less than either, and draws 1,056. E-commerce runs $300, under half the sports budget, and draws 673. Rank the nine verticals by prize and by Entrants and the two orders barely agree, which is the same result section 4 found inside a single spectrum of cash values.

Budget is the smaller half of the decision. The prize itself is the other half, and it is not the same object from one vertical to the next. We read the prize name on every campaign in the window and sorted it into one of 17 physical categories, then took the two most common in each vertical.

Table 16 · Most common prize types by industry, 1 Sep 2025 to 13 Aug 2026

A prize type is listed only where at least 20 distinct businesses in that vertical used it, and a median value only where at least 30 priced prizes from at least 15 businesses support it. Shares are the percentage of that vertical's campaigns whose prize name fell in the category, so they do not sum to 100: the remainder is spread across the other 15 categories and across prize names too short to classify.

Three things fall out of that table. Gift cards lead in three of the nine verticals and are the cheapest thing in the study, at a $45 median among creators and $50 in B2B. Hardware leads where the audience is already shopping for it, and the prices match. Esports accounts give away consoles and PCs at a median of $1,900. Gadget brands run the same category at $144, because they mostly give away one accessory instead of a full machine. And the second-place type is usually more expensive than the first, sometimes by a factor of twenty, as in B2B at $50 against $1,137.

That last pattern is the prize-fit argument in numbers. In every vertical the common prize is the cheap generic one, and the expensive specific one runs second. Neither section 4 nor section 8 found the expensive option buying proportionate reach, so the cheap generic prize holds up better than it looks. Picking one that the vertical's audience specifically wants is the decision that pays.

One row carries a warning. Cryptocurrency converts at 45.4 percent against a platform median near 29, on the shortest widget of any vertical at five Actions. That is either the most efficient audience on the platform or an artifact of automated entry. The invalid-entry rate for that vertical has not been measured, so this row is published as a description of the segment and is not a benchmark for anyone outside it.

1,135 sites in this window remain unclassified, which is the 4.8 percent of campaigns the tables above do not cover. They average under one campaign each, so finishing them changes coverage more than it changes any number. Four verticals sit close to the reporting minimum, and Podcasting, Baby, Pets and Wine and Spirits are barely present. The vertical labels in this edition are unaudited. An earlier pass of the same method measured a 6 percent error rate. Nobody has run a blind audit on the labels used here. Treat every vertical cut as carrying errors of that size.

Prize budgets are stable. What organizers ask Entrants to do moved. Three Actions have grown and eight have shrunk over six years, and the direction of travel is away from asking people to broadcast and toward asking them to arrive from somewhere you already own. If your Action stack came from a guide written a few years ago, this is the table to check it against.

Table 17 · Share of campaigns using each Action, by start year

165,204 campaigns, 2021 to 13 August 2026. Only Actions available for the whole period are shown, so the movement is organizer choice. Instagram follow, TikTok follow and Facebook like are excluded: each sits at zero until the quarter it shipped, so their rise records a product release.

X reposts more than halved from their 2022 peak. Instagram profile visits fell 27 points. The only Action in that table growing by more than two points is the Secret Code, which an Entrant can only complete if they met your campaign somewhere off the widget.

Eight declining rows invite one reading, that organizers are stripping Actions out. The median campaign says otherwise. It asked for 7 Actions in 2021 and it asks for 7 today, unchanged in every year between.

The three Actions excluded from table 17 are where the movement went. Each shipped partway through the period, so a six-year trend line would record a product release rather than a choice, which is why they sit outside the table. Their adoption since launch:

Table 18 · Share of campaigns using the Actions that shipped mid-period

Same 165,204 campaigns. 2026 covers 1 January to 13 August, so it is a partial year. Each Action reads n/a before the quarter it shipped, because a zero there would describe Gleam's release schedule rather than organizer behavior.

Instagram Follow reached 46.7 percent of campaigns in its second year, which is higher than every row in table 17 except X Follow and Custom Action. So the widget is the same length as it was six years ago and the slots are filled differently: broadcast Actions out, follow and arrival Actions in. That is substitution, and it is the thing to check your own stack against.

A single trend line across the whole platform is the wrong unit for this decision, because the verticals do not build the same widget. Using the industry classification from section 9, here is what each one asks for.

Table 19 · Share of campaigns using each Action, by industry, 1 Sep 2025 to 13 Aug 2026

Campaign counts per column: Creators 4,081, E-commerce 2,403, Smart Wear Gadgets 1,547, Cryptocurrency 1,167, B2B 841, RPG Board Games 702, Sports 491, Hobbies Crafts DIY 445, Esports 341. Every column clears 300 campaigns and 75 businesses, and carries the same dominance trim as Table 15. Columns do not sum to 100 because a campaign uses several Actions.

Read across the Email Subscribe row. Sports asks for an email on 51.9 percent of its campaigns and cryptocurrency on 6.3 percent, an eight-fold difference in the one Action that collects an address the organizer keeps. Read across Answer a Question: 62.6 percent in cryptocurrency against 5.6 percent in esports.

Two verticals build a widget nobody else builds. Cryptocurrency runs on X Follow at 86.1 percent, Telegram at 70.1 and Answer a Question at 62.6, and almost nothing else. B2B runs Bonus Entry at 52.3 percent and Custom Action at 35.6, the highest Custom Action share in the table, which is a vertical building entry paths that Gleam does not ship by default.

Cross-read this against Table 16 and the stacks line up with the prizes. The three verticals that hand out gift cards most often, creators, e-commerce and B2B, are also the three that ask for the fewest social follows. The two that hand out hardware and game keys, gadgets and esports, run the longest social stacks at nine Actions. Both halves of the campaign are picked for the same audience, which is the argument section 5 makes about Action choice, seen from the prize side.

So the platform-wide decline in the table above is an average of nine different movements. If your vertical is in this table, plan from your column. If it is not, the platform figures are the best available and they are describing somebody else's campaign as much as yours.

What this does not show is that any of these stopped working. Gleam's own customer mix changed across six years, and a platform can withdraw an Action through an API change without anyone choosing anything.

Start here

Twenty-two times the prize budget came with 3.2 times the Entrants. A return that poor means the budget is a smaller lever than reach or entry Actions. Decide what your intended audience actually wants, then use the next row to work out how much of it to buy. We have not measured prize fit directly and say so in section 9, so this is the one recommendation here that rests on what the data rules out.

Biggest measured effect

Two decisions sit here and both are free. One to three Actions converted 44.0 percent of Impressions against 23.8 percent at thirteen or more, and at the same reach the longer widget ends with 35 to 41 percent fewer Entrants. Then, at any given length, using one Action instead of another came with a further 10.9 point spread on Conversion. Cut the Actions that are not outcomes you want, then rank the ones you keep against table 7.

Set the budget

Each Entrant carries $0.44 of stated prize value in the $100 to $499 band and $3.35 above $2,500. Decide how many people you need, then find the cheapest band that gets you there. The median is $300 overall, and on your own plan it is $140 on Hobby, $312 on Pro and $645 on Business.

Read the result

A bigger prize gets promoted harder, so more people see the widget than enter it. Across the prize range, Impressions rose 317 percent and Entrants rose 218 percent, which pulls the rate down 5.5 points. On an expensive campaign, count Entrants.

Structure the prize

Campaigns with five or more prizes spend a median $1,632 against $240 for a single prize, so 6.8 times the budget. Their top prize moves from $150 to $400, which is 2.7 times. The extra money buys more winners, and 81 percent of the competitions that state a prize value run a single prize. n = 226 campaigns from 126 businesses at five or more prizes and 5,094 from 1,343 at one, of the 6,322 that state a value.

One thing this study cannot tell you is whether raising your prize would raise your reach. Bigger prizes and bigger audiences travel together in this data, and the businesses that can fund one usually already have the other. Treat the numbers above as what campaigns like yours achieved, and plan against the P25 column when you need a number you are confident of clearing.

These are the canonical tables for this edition, restated next edition on the same definitions.

Table 20 · Prize budget, full distribution, 1 Sep 2025 to 13 Aug 2026

Table 21 · Prize budget planning ranges by plan, 1 Sep 2025 to 13 Aug 2026

Premium is left out at 80 campaigns and 47 businesses stating a value, below the reporting minimum. Free carries 547 campaigns from 219 businesses, which clears it.

Table 22 · What to expect by prize band, 1 Sep 2025 to 13 Aug 2026

Plan against the P25 figure if you need a number you are confident of clearing, and against the median if you are setting an expectation. The gap between P25 and P75 is what campaigns on the same budget actually vary by.

Table 23 · What to expect by entry-Action count, 1 Sep 2025 to 13 Aug 2026

Table 22 uses the 6,322 campaigns that state a prize value. Table 23 uses the 16,561 campaigns in the window for the Entrant columns, and the 16,542 with at least 100 Impressions for the Entered column. Every campaign in both now carries an Entrant count, so neither is size biased. Read them as ranges for a campaign of that size, and shade down if yours is smaller.

Everything above rests on the decisions in this section. It sits at the end so the findings come first, and it is written to be checked.

Headline figures cover campaigns starting between 1 September 2025 and 13 August 2026, by campaign start date. 13 August 2026 is the last campaign start date in the export. Entrant counts come from a second, complete dump supplied by Gleam engineering on 10 September 2026, because a serialization fault in the first export omitted the Entrant count from two thirds of records. 379 campaigns carry different Entrant counts in the two sources, 0.2 percent of the export, and this edition uses the September dump for all of them. That is 11.4 months and we label it as the window itself instead of calling it a trailing year, because eighteen days of a full year would contain no campaigns. Every table on this page carries its own window in the caption. The whole export runs from February 2017 to 13 August 2026, and 99.3 percent of it starts in 2020 or later. Structural cuts run 2021 to 2026 and say so where they appear. Year-on-year comparisons in the next edition will compare the same calendar months.

• 33 campaigns in the window are excluded for a duration outside 0 to 365 days. Across the whole export 1,026 records fail that test, including one recorded at 694,015 days.
• Prize values in currencies other than US dollars are excluded rather than converted, because we hold no dated exchange rate. They are 134 of 92,606 prize records, or 0.1 percent.
• Campaigns under 100 Impressions are excluded from Conversion, because the denominator is too small to carry a rate. That removes 19 campaigns from the Entrant sample, leaving 16,542.
• Any cell below 200 campaigns or 50 businesses is suppressed and shown as n/a. Industry cuts use a higher bar of 300 campaigns and 75 businesses, and Tables 15 and 19 say so. Sixteen cells on this page are suppressed: eleven prize types in Table 14, four verticals in Table 15, and Premium planning ranges in Table 21. Premium Entrant figures in Table 1 are shown with a warning rather than suppressed, at 47 businesses.
• No account dominates a reported cell. Among the 6,322 campaigns stating a value the largest business holds 5.4 percent and the top three hold 12.2 percent. Across all 16,561 campaigns in the window the largest business holds 2.1 percent. Table 7 also reports the largest business share on every row and suppresses any Action above 10 percent.

Campaign results are extremely skewed, and averages here are not describable. In the twelve-month window the mean stated prize value is $113,326 against a median of $300, because a handful of records carry implausible values, one of them at $100 billion. An average built on that tells you nothing about your campaign. Every headline figure on this page is a median, the benchmark tables carry the 25th and 75th percentiles, and every cell carries its campaign and business count.

Every comparison was run four ways before publication. Across the full sample, on one campaign per business, inside a single plan tier, and inside bands of entry-Action count. A difference that appears only in the first is reported as not holding. Where two factors move together, such as plan tier and prize budget, the comparison is repeated with one of them held inside a band, and the controlled figure is the one published.

• It is not all giveaways. It is Gleam competitions, run by businesses that chose Gleam.
• It carries no causal design. Nothing here shows that changing a prize, a plan or an Action count causes an outcome. Every comparison is between campaigns that already differed.
• It is not adjusted for inflation. Values are nominal.
• Its industry labels are unaudited. 2,849 sites were classified by hand from their domain, covering 95.2 percent of campaigns, and no blind re-classification audit has been run on that pass. The prior keyword pass measured a 6 percent error rate.
• It does not measure revenue, sales or list quality. Entrants are Entrants, and this study does not claim they are qualified leads.

Three customers are named on this page: MAINGEAR, Delta Sonic and Muckender Towels. Every figure quoted for them comes from their own published Gleam case study, which each of them agreed to. No customer is named anywhere in the benchmark tables, and no single campaign in the dataset is described in a way that would identify the business behind it.

Edition 1, published 14 September 2026. No prior edition. Corrections to this page will be published as a new version with the change named. Numbers are never altered silently.

According to the Gleam Giveaway Benchmarks 2026, the median Gleam competition that states a prize value states $300, based on 6,322 campaigns run by 1,798 businesses in the window of 1 September 2025 to 13 August 2026.

• Laporte, S. and Laurent, G. (2015). More prizes are not always more attractive. Journal of the Academy of Marketing Science 43, 395-410. Five experiments. Directional comparison only, because the experiments test stated preference and this study measures configured campaigns.
• Dickson, P. R. and Sawyer, A. G. (1990). The price knowledge and search of supermarket shoppers. Journal of Marketing 54(3), 42-53. In-store observation with intercept interviews. Context only, and 36 years old.

We did not benchmark Conversion against a published industry average. The giveaway benchmarks we found either do not state their denominator or measure a different one, so a comparison would put two unlike rates side by side.

About this series. Gleam publishes studies on competitions, lead capture, user generated content and rewards, built on first-party platform data. Every edition ships a methodology section, a changelog and the comparisons that did not hold. Published numbers are never changed silently; corrections create a new version.

Next in this series. A study of where competitions get seen, using daily referrer data held for 159,103 campaigns. It is the analysis that would separate promotion from prize money in section 4.

Known data limits. A serialization fault in the export omits the Entrant count from two thirds of records. Gleam engineering supplied a complete dump on 10 September 2026 and every Entrant figure in this edition uses it, so the earlier size bias is gone. 379 campaigns carry different Entrant counts in the two sources, 0.2 percent of the export, and this edition uses the September dump for all of them. Two campaigns carry corrupt dates and are excluded.