Distributing banner fundraising year round (at lower levels of traffic)

We are seeking your input on a potential shift in our fundraising model: running banners year-round on English Wikipedia, at a lower percentage of traffic, rather than concentrating activity heavily at the end of the year.

Today, our Big English campaign, which runs in the US, UK, Canada, Australia, New Zealand, and Ireland, focuses most banner activity in December, where we show banners to 100% of traffic coming to the site. This proposal would shift toward a more evenly distributed approach, with banner traffic running throughout the year to a limited portion of traffic coming to the site.

This is in response to changes we're seeing in our data. In our most recent banner campaign, revenue declined 7% year-over-year and new donor acquisition dropped by 20%, driven largely by declining readership. At the same time, our current model remains highly concentrated: nearly 75% of banner revenue comes from the Big English campaign, with most of that generated in December.  As Wikipedia readership continues to decline, this level of reliance on a single month poses increasing risk.

Below, we outline the goals behind experimenting with this approach next year.

• A more consistent reader experience: Instead of a sharp spike in December (with banners shown on a large share of pageviews), we can distribute impressions more evenly across the year. We will maintain impression caps for banners to limit the total number of banners any reader will see. This is something that has long been important for our team, more banners drive incremental revenue so we are focused on effective banners rather than banner quantity.
• Stronger experimentation and learning: A longer time frame allows us to run more robust tests across different audiences, messages, and formats, leading to better-informed decisions than short campaign windows allow.
• Reduced concentration risk: Our current model depends heavily on a single campaign moment, with most revenue generated in December. Distributing fundraising across the year lowers the risk that any one period underperforms due to factors outside our control.
• Better alignment with reader behaviour and needs: Readers interact with Wikipedia year-round, but our fundraising presence is limited to a narrow window. This approach allows us to meet readers where they already are, rather than concentrating asks into a single period.

Risks we're thinking about

In addition to the benefits above, we want to be upfront about some of the risks we're weighing:

• Overexposing individual readers to banners. As mentioned above, we will need to be thoughtful about frequency caps and how we track cumulative exposure.
• Perception concerns. Some readers may see year-round banners as more intrusive. We have many ways to track sentiment to banners and we will keep a keen eye on these over the course of this work.
• Uncertainty in a new model. This is a significant change, next year we would like to experiment with this approach and we may need to adapt as we go or revert back to the old approach if this does not work.

What other risks or challenges should we be thinking about? Your input will help us design this responsibly.

Thank you, SPuri (WMF) (talk) 13:49, 9 April 2026 (UTC)

This sounds like a good idea. I'm particularly optimistic about the better conditions for testing. In the past, when the community has proposed ideas for banners, testing has always felt a little rushed due to the time pressure of the campaign. Showing fewer banners at once may result in less reader annoyance, but of course we should wait and see if the numbers bear that out. Spreading the banners out to avoid risk from sudden shocks to disposable income (such as trade wars or actual wars) seems sound. My only worry is that this could become an excuse to drastically increase total volume of banners – please don't let that happen! Toadspike [Talk] 14:33, 9 April 2026 (UTC)

According to the Charities Aid Foundation , the last two months of the year are prime time for charitable donations, with more Americans reaching for their wallets to support meaningful causes. A recent survey by CAF America found that 62% of Americans plan to make monetary donations during this holiday season – and 15% say they’ll be giving exclusively during this time. I'm curious to see how the new fundraising model performs. (My prediction: Not well.) Some1 (talk) 23:12, 9 April 2026 (UTC)

I see the donate button has changed. It is now in boldface and has a heart to its left. It appears to only show to users when not logged in. I'm okay with the heart, but I'm not sure how I feel about the "Donate" text being emphasized beyond "Create account" and "Log in", which are arguably more valuable to the projects. Toadspike [Talk] 09:51, 10 April 2026 (UTC)

@SPuri-WMF We are seeking your input on a potential shift in our fundraising model: running banners year-round on English Wikipedia, at a lower percentage of traffic Bad idea. Wikipedia famously is the one site on the internet where you can expect to NOT see banners. Banner blindness is a thing, and audience goodwill might drop with too many banners, especially if they are unexpected.

In our most recent banner campaign, revenue declined 7% year-over-year and new donor acquisition dropped by 20%, driven largely by declining readership. I would start with the basics: Does the WMF actually need more money, what will the money be spent on, and how has the WMF handled the money it has received so far, and what does it plan to spend money on in the future? And what about cutting costs? I don't think "run more banners" is the only way forward. Polygnotus (talk) 04:05, 13 April 2026 (UTC)

Readers are shown banners quite regularly. They're just advertising Wiki Loves X or similar type campaigns. Best, Barkeep49 (talk) 14:20, 13 April 2026 (UTC)

They are? Oh. I block such things so I've never seen 'em. And an ad for Wiki Loves Scientists is probably less offensive than e-begging. Polygnotus (talk) 15:12, 13 April 2026 (UTC)

I am not inclined to tell professionals how to do their job and I am not a professional fundraiser so what do I know? That said my initial thought was along the same lines as Some1, which is that holiday fundraising for a site like Wikipedia is easier. So I guess my comment is "sure, as long as we don't end up in a scneario where we've shown lots of banners throughout the year, find it fails, and then still need to do an intensive drive at the end of the year". Best, Barkeep49 (talk) 14:23, 13 April 2026 (UTC)

Some1 and Barkeep49 make good points, perhaps the WMF could go halfway, still keeping the spike in December but decreasing it and replacing it with minimal year-round banners. As Barkeep49 said well, I am not a professional fundraiser, so I'm not going to tell you how to do your job! Feeglgeef (talk) 18:10, 13 April 2026 (UTC)

Thank you all for the input and suggestions, they are deeply appreciated.

• @Toadspike, +1! I am particularly excited about how this model will decouple testing from peak revenue pressure. Right now, trying to both hit the budget and test new ideas within a single month is limiting. A more distributed approach gives us the space to improve banner performance more meaningfully over time, rather than relying on increased impression volume which does not tend to drive significant incremental revenue per reader and is something we are actively avoiding.
• @Some1 thanks for that link, we do see that pattern in our data.
• @Barkeep49 your concern is well taken. And @Feeglgeef you nailed our vision: we do expect to retain a December peak, but with fewer banner impressions than today. The goal is to reduce lower-performing impressions during that period and reallocate them to times of year that are less saturated and potentially more effective for readers.
• On the donate button: you're right that the small heart “Donate” button has now been rolled out across all Vector 2022 Wikipedias for logged-out readers. The bolding is intentional, to make the option more visible as a lower-friction, less disruptive way to give compared to banners. That said, I'll pass along the feedback to the Product team as they continue to refine how and where readers can engage.
• @Polygnotus, to clarify: year-round testing is not about increasing the overall number of banners shown. It's about maintaining revenue while improving efficiency, especially in the context of declining readership. We share the goal that a banner-free reading experience is ideal. We estimate that the number of pageviews currently receiving a fundraising banner could decrease if impressions are spread more strategically throughout the year.
• On budget allocation, I'd encourage you to engage in the Annual Plan conversation where these trade-offs are explored in more depth.

Please keep the comments coming, this feedback is incredibly valuable. - SPatton (WMF) (talk) 18:28, 15 April 2026 (UTC)

Thanks a lot! Also seconding Barkeep's comment that the "Donate" button ideally shouldn't be bolded: the action to create an account should be just as visible, as that is how we get new editors, and, well, keep the project going. Not convinced about it being understood by readers as a lower-friction alternative, as, if I understand correctly, you are planning to do both concurrently. Chaotic Enby (talk · contribs) 23:46, 29 April 2026 (UTC)

For those following this discussion, there is some follow-up discussion at the Village Pump on bolding the donate button. Best, Barkeep49 (talk) 16:09, 30 April 2026 (UTC)

Experimenting with email signups to build longer-term support for Wikipedia

Hi everyone,

We want to share an upcoming small banner experiment we're planning in the United States on Wikipedia, and would like your input at this early stage.

What we're testing

Today, most fundraising banners ask readers directly for a donation in the moment. In this experiment, we want to test a different approach: instead of immediately asking for a donation, we'll invite a small portion of readers to share their email address so we can stay in touch over time.

The idea is simple: for some readers, a single visit may not be the right moment to give. Email may allow us to build a longer-term relationship, sharing useful Wikipedia-related content, updates about how Wikipedia works, and occasional opportunities to support financially. We've included an example of what the banner could look like.

What users would receive

People who opt in would receive a short series of emails that may include:

• Highlights of interesting Wikipedia content
• Information about how Wikipedia and the Wikimedia Foundation work
• Occasional invitations to support Wikipedia financially or in other ways

All emails will clearly identify the Wikimedia Foundation as the sender and include standard unsubscribe options.

What we are trying to learn

This will be a limited, time-bound experiment.

We want to understand:

• How many readers choose to sign up
• Whether email follow-up leads to stronger long-term donor engagement
• Whether this approach helps people better understand how Wikipedia is funded and why support matters
• Whether email provides a meaningful additional channel for engagement beyond the immediate banner visit

We will also measure whether this has any unintended impact on current donation conversion rates.

Scope and safeguards

• This will run only in the United States for a limited period
• We are aiming to collect approximately 5,000 email signups during the test
• All data will be handled under our existing donor privacy policy

As always, we'll share results back with the community once we have them. - SPatton (WMF) (talk) 19:26, 29 April 2026 (UTC)

@SPatton (WMF) When and how will this be determined to have been a success?

I assume the contents of the emails is just taken from somewhere, where does that come from?

Will the total amount of ebegging banners per year go up, or will the other types of ebegging banners be reduced?

Is the call to action in the screenshot final? "interesting articles, behind-the-scene insights, and ways to support Wikipedia" is not very enticing, nor is a Submit button (subscribe would be better). Who is the target audience? People who want interesting facts are perhaps not interested in ways to support and behind-the-scenes insights, so I would get rid of those two and write a CTA based on the idea that you could get some interesting facts or articles in your inbox once in a while. Depths of Wikipedia has a good collection of interesting stuff. Polygnotus (talk) 19:46, 29 April 2026 (UTC)

I don't think readers would give a website they casual browse their personal email addresses just to occasionally get emails about "behind-the-scene insights, and ways to support [the website]" (uninteresting and most people's inboxes are already full as is), or at least I wouldn't. I agree with Polygnotus's suggestion above. Some1 (talk) 00:00, 30 April 2026 (UTC)

How can a promise of a continuing series of emails be a "limited, time-bound experiment"? The "limited period" presumably needs to last at least a year to compare donations to background rates. There is nothing in that screenshot which indicates a limited time-bound experiment. After that year plus, is the plan at the end to send an email saying "Thanks for begin part of this donor experiment, there will be no further emails", or to continue sending emails to the experimental 5,000 Americans for a longer period? CMD (talk) 07:20, 30 April 2026 (UTC)

Yeah differentiating between "emails are the cause of the donation" and "emails are not the cause of the donation" is impossible so this experiment will always be a success or a failure depending on what you want those terms to mean.

Its not like people who donate because of the emails are gonna click a button in an email and end up at a donation screen. It is 2026 no person does that. I have thousands of unread emails in my inbox and I am pretty sure a decent amount of them are asking me for money. Polygnotus (talk) 07:30, 30 April 2026 (UTC)

Ugh...I'm usually pro-fundraising, but this seems really skeevy. It feels dishonest to present this to readers as a fun newsletter with cool info, when really it's just a fundraising tool. Can we not do the whole spam email thing? Toadspike [Talk] 12:17, 30 April 2026 (UTC)

If the newletter is properly laid out with the right content, and the donate CTA is not obnoxious, sure. Conversely, splashing the donate CTA before the content is laid out is definitely WP:SURPRISE! – robertsky (talk) 15:43, 30 April 2026 (UTC)

Are people even interested in newsletter emails today? The vast majority of people see them as spam, and measuring the negative effects of yet another pop-up banner asking to subscribe to a newsletter is going to be much harder than seeing the positive effects of having people who subscribe. As Polygnotus said, it can be quite easy to label this a success or a failure regardless, and we would like to have clear visibility on how success will be measured. Chaotic Enby (talk · contribs) 20:13, 30 April 2026 (UTC)

Thank you for your initial responses and let me address your feedback here.

• How will we determine success? This is an exploratory experiment. We’ll look at sign-up rates and downstream engagement, such as open rates, unsubscribe rates, delivery rates and donations over time. We’re not treating any single metric as definitive. The goal is directional learning about whether this channel contributes to engagement and support.
• By small-scale test,” we mean that we are running the banners for a limited period of time on Wiki. People who choose to sign up will continue receiving emails unless they unsubscribe (which is always available).
• The email content will be created by our content team and may draw on existing content. They will include both informational content and occasional fundraising asks.
• Thank you Toadspike [Talk], this is a fair concern, and we want to be clear in the sign-up process, as this will mean more readers are happy with the content they receive. We want to understand whether readers want a lighter-touch way to stay connected to Wikipedia. We’ll be looking closely at signals like opens, clickthroughs and unsubscribes, as well as what content gets the highest engagement. If readers are finding value in what we send them, that's something we will consider scaling. If the test shows limited value or negative impact, we won’t scale it.
• The screenshot is a draft to provide you with a visual idea of what we are thinking. We’ll continue refining the wording and your suggestions are very helpful, thank you.

SPuri (WMF) (talk) 21:14, 30 April 2026 (UTC)

The goal is directional learning about whether this channel contributes to engagement and support. I would recommend finding or hiring someone who knows about email marketing and research design. Saves a lot of time and work. People already know how this works and when this doesn't work.

The email content will be created by our content team and may draw on existing content. Plenty of excellent content is already available so there is no need for a content team to create anything. Polygnotus (talk) 21:56, 30 April 2026 (UTC)

Thank you @Polygnotus. We have an email marketing team within Fundraising. You can read about our email marketing program in the latest Fundraising Report. I agree, there is lots of amazing content already. SPuri (WMF) (talk) 14:26, 1 May 2026 (UTC)

@SPuri-WMF Why isn't the email marketing team involved? They would have already pointed out the tracking difficulties and the weak Call to Action. If an experienced team exists then there is no need for "directional learning", right? Anyone in that line of work can predict the outcome.

Comparing donations from those 5k sign-ups against the general public is meaningless because the 5k is already far more likely to donate. Without a Randomized Controlled Trial where half of the sign-ups are randomly assigned to a control group that receives no emails (or emails without a donation CTA) the results will be nothing more than a vanity metric.

"Directional learning" is just corpo-speak for a lack of hard KPIs. Without a clear threshold for failure, this experiment can never be called a "success". If the WMF does this without setting a clear goal then the WMF will learn nothing from the experiment failing.

A newsletter is a recurring overhead, not a one-time expense. What is the exit strategy when the overhead inevitably exceeds a hypothetical increase in donations? "Sorry people who signed up we will stop sending email"? Will the WMF kill a "successful" channel if it's a net financial loss? People who are not familiar with interpreting such stats in a commercial setting are easily convinced that the project is a success based on engagement numbers alone. But engagement != money.

Offering a "low-friction" option to support often distracts users from the "high-friction" goal of donating. If sign-ups increase but immediate donations drop even slightly, the result is negative. This plan has no way to distinguish between genuine new support and support that was delayed or lost. Polygnotus (talk) 15:37, 1 May 2026 (UTC)

And please tell me you do the industry-standard double opt in? Anyone can fill in any email address and you usually can't tell if someone is from a GDPR country based on their email address (or at least not using software). And will the email banners replace ebegging banners or is this a plan to increase the total amount of banners? Polygnotus (talk) 15:58, 1 May 2026 (UTC)

Is there a clearer indication of what informational content means? Explaining how Wikipedia works? An email version of the Main Page? This could be good, but it will take some effort to put together a newsletter worth reading. Toadspike [Talk] 20:21, 1 May 2026 (UTC)

Hi, just a quick note that we did see your responses but last week and this week many of us were/are travelling for work. We will get back to you soon. Best, JBrungs (WMF) (talk) 09:28, 12 May 2026 (UTC)

While I share some of the concerns Polygnotus has raised, I want to note my overall support for this kind of work. For me - and this is perhaps true of some of my fellow editors - because of my support I see it all the more important to "get it right" so that it isn't deemed a failure after a flawed test. Best, Barkeep49 (talk) 16:17, 12 May 2026 (UTC)

Indeed. I am not opposed to experimentation in this area; I just want to be realistic about the challenges ahead. A badly designed experiment is worse than no experiment.

We need a hypothesis and a hard target (in USD, not engagement), decided before the experiment starts.

Randomly split sign-ups so half get emails with a donation CTA and half get emails without one otherwise the data will show that people who are willing to sign up to a newsletter are more likely to donate than those who are not.

And it is important to decide ahead of time what happens if and when the plug is pulled.

Most importantly: put someone in charge who has serious experience with both research design and email marketing. Proper A/B testing is not as easy as ChatGPT makes it seem; there are a lot of pitfalls that are non-obvious for people who are new to this stuff.

"Exploratory", "directional learning", and "not treating any single metric as definitive" all mean the same thing: there is no number at which this experiment fails, which means it's not an experiment but a waste of time and money. Done properly this could actually bring in some money. Polygnotus (talk) 16:52, 12 May 2026 (UTC)

Thank you everyone for the thoughtful feedback, we are listening. It is a privilege to have you thinking with us about how to make this pilot meaningful and well-designed.

As a digital fundraising team, we already rely heavily on controlled testing and measurement in our work, and we see the importance of applying that rigor here as well. Motivated by this discussion, we are planning to structure this more explicitly as a controlled experiment, and we will share more of that structure here as we refine the design.

One approach we are considering is testing at least two different email tracks: one focused more heavily on engagement content, and another with more direct fundraising asks. This will help us better understand both reader engagement and fundraising outcomes over time.

One point from this discussion especially resonated with me: if we ask readers for space in their inbox, the emails need to genuinely provide value. On that point, I’m really confident that we can offer readers and potential donors something meaningful in their inboxes. We have years of testing around our ‘Remind me later’ signup process, and engagement content, to back that up.

In the coming weeks, we'll follow up with more details around hypotheses, safeguards, and how results will be evaluated and shared back with the community. - SPatton (WMF) (talk) 21:28, 13 May 2026 (UTC)

Sharing our banner tests

Hi all,

During the December 2025 campaign, we heard the request to share more of our banner tests. Over the past few months, we've been considering the best way to create more transparency and accessibility into our banner testing. We run over 700 tests in just six months, so rather than maintaining an unwieldy list that ultimately may not be very useful without context, we propose a slightly pared down version detailing tests that end up becoming part of our control banners.

Our team already keeps detailed internal logs of changes made to each banner type, including content adoptions, design changes, bug fixes, and the like. Starting this July, we are happy to start posting these logs on a frequent basis on a designated page to keep you informed of how our banners are evolving. Posts would include: a brief description of the change made, the code diff, a link showing a banner preview diff, time and date of change.

Each banner type already has a revision history available on metawiki that can be watched if you have a metawiki account (large desktop banner, small desktop banner, large mobile banner, small mobile banner).

Thank you, -KMorrow-WMF (talk) 19:39, 29 April 2026 (UTC)

@KMorrow-WMF Some screenshots would be nice. I see a bunch of CSS and JS but no screenshot. Polygnotus (talk) 19:48, 29 April 2026 (UTC)

Hi, @Polygnotus. When I start sharing banner updates in July they will be in an easier to read format with a short description. They'll live on a dedicated sub-page to the fundraising hub and I will share the link when it is available. The CSS/JS shared was to show that this information is available if you wanted to take a look before July. Best, KMorrow-WMF (talk) 13:47, 30 April 2026 (UTC)

@KMorrow-WMF Ok, thank you! Polygnotus (talk) 21:38, 30 April 2026 (UTC)

"Thank you" Badge for donors

Earlier this year we ran a [experiment] on English Wikipedia for a small selection of readers who had recently donated, where we offered a small Wikiheart badge that would appear on their page. Response to the badge was overwhelmingly positive from donors who completed an optional survey, and the majority of donors who enabled the badge kept it for the duration of the experiment. We’ve refined the technical setup and would like to retry this experiment. We plan to run this badge test again on the English, Japanese, German, Danish, Greek, and Norwegian Wikipedias, as these are the language Wikipedias for our [banner and email fundraising campaigns]. As before, we’ll only surface the badge to recent donors who have the standard ‘donate’ cookie in their browser. We’re evaluating success based on user feedback and how many donors interact with the badge. Happy to hear your thoughts on this. Best, AKanji-WMF (talk) 16:54, 8 June 2026 (UTC)

Current Best banners

Example large banner (preview)

We owe you an explanation.

%date%: An important update for readers in %country%.

You deserve an explanation, so please don't skip this 1-minute read. It's %weekday%, %date%, and you only had a small chance of seeing our message. This year, Wikipedia has seen fewer visitors and fewer new supporters, so your visit today means a lot. We hope that Wikipedia has given you at least %local-anchor% of knowledge recently. If everyone who finds Wikipedia useful gave %local-anchor%, we'd hit our goal in a few hours.

It's been 25 years and Wikipedia is still free. It's still created by people, not machines, and we don't run ads or put up paywalls because we're not here to make a profit off your attention. In other words, it's still the internet we were promised.

Less than 2% of our readers donate, so if you've never given and Wikipedia has provided you with at least %local-anchor% worth of knowledge, donate today. If you are undecided, remember any contribution helps.

Example small banner (preview)

%date%: Wikipedia still can't be sold

We're sorry we've made several attempts to reach you. It's %weekday%, %date%, and Wikipedia is seeing fewer readers this year so your visit means a lot. If everyone reading this gave %local-anchor%, we'd hit our goal quickly. Most readers don't donate; only 2% do. Wikipedia is run by a nonprofit, not by a billionaire. So if Wikipedia has given you $2.75 worth of knowledge, please give. Any contribution helps, whether it's %local-anchor% or %local-high-anchor%.

{Following text is hidden until click to reduce initial banner height}

Thank you for joining the 2% of readers who give.

25 years ago Wikipedia was a dream. A dream built piece by piece by people, not machines. Now, with 65 million articles and 260,000 global volunteers, Wikipedia is a place of free, collaborative, and accessible knowledge—run by a nonprofit, not a corporation or billionaire. Your donation isn't just supporting a website; it's investing in the world's largest collaborative project of human intelligence—written by humans, for humans. Please join the small percentage of readers who give what they can to help keep Wikipedia strong and growing. Thank you.