by Brian Tomasik
First written: 4 Feb. 2014; last update: 9 Nov. 2016

## Summary

If you find an effective charity, write to them to ask whether they use Google Grants, and if not, suggest they sign up. Google Grants offers the prospect of immense returns for a small amount of labor, although one needs to be careful about not competing with other effective organizations and choosing keywords that draw in new people rather than preaching to the choir.

Update, Sep. 2015: Having used Google Grants for the last 1.5 years for several organizations, my conclusion is that the value of AdWords is modest. None of my organizations has found via AdWords a major donor or a promising future employee, even though our websites get high traffic volume from ads. Maybe part of the reason is that the best people don't click on ads much? Another reason is that the best people tend to be concentrated in dense social clusters, so that networking can be more effective.

Google Grants is a program that gives registered charities in the US and many other countries up to $120,000 of Google ads per year for free. This is an amazing opportunity, yet many nonprofits are still not signed up. Google Grants were popularized to the effective-altruist community by Effective Fundraising (now called Charity Science). For instance, see "Scaling Google Grants for Animal Charities." ## The value of Google Grants$120,000 per year is a lot of money. Of course, it can only be spent on ads, so the value is lower than having pure cash. Say you value ads at $0.1 for every$1 of budget. (The reason not to value them closer to $1 is that most ad traffic is noisy and bounces away.) Then$120,000 would be equivalent to $12,000 of cash per year. Google Grants require a few hours to submit an application and then at most a few hours per month to manage the campaigns. Say that's 1 hour of work per week in the long run. If you spent 1 hour a week for 50 weeks a year on ads worth$12,000, that would be $240 per hour. Louie Helm of the Machine Intelligence Research Institute reports that after initial time spent setting up campaigns, he now devotes at most an hour per month to maintenance. Especially if part of your organization's goal is outreach for important ideas, then ads to your content are not just instrumentally useful by growing your organization but also terminally useful by expanding your readership. Ads are a much more scalable and divisible form of room for funding than additional employees, who take time to be trained and come in discrete units. Additionally, unlike with hiring employees, the counterfactual opportunity cost of ads may be fairly low, although this will be discussed more later. Maintaining the ads by refining keywords is important, because many nonprofits only use a small fraction of their$10,000-per-month budget due to not investing enough attention to their bids and keyword choices. If done well, Google Grants traffic can comprise a substantial fraction of the total visitors for a charity's website, although any given visitor may have lower quality than someone visiting the site due to a referral or link. On average, 25.3% of visits to charities with Google Grants come from the program, and it could be an even higher percentage for small or startup nonprofits, or for those that manage their campaigns successfully.

If you know a charity that does important work, you can do an imperfect test to see if they're already signed up by Googling their name. If an ad shows up, they probably already have Google Grants. For instance, for {machine intelligence research institute}, you can see an ad at the top or bottom of the results page. Of course, this test may yield false negatives if the charity does have ads but not for its own name.

If you don't see an ad for the charity you like, then you should contact them and ask if they're signed up for Google Grants. For example, here's a modified version of an email I recently sent to Mercy for Animals (MFA):

Hi MFA,

I've been interested in your work ever since some friends pointed me to you a few years ago. Your investigations are amazing, and I was particularly affected by the one about catfish slaughter.

I'm writing to ask if you use Google Grants. The program offers $120,000 worth of ads per year, which can go to either your home page or a landing page with a video like Farm to Fridge. If you're not signed up yet, the application is pretty simple, and I'd be glad to assist if you like. Best wishes with your work! Brian I got a helpful reply shortly afterward informing me of MFA's status. I've written to several other EA organizations with the same question and have found that some are signed up, while others are not and are eager to investigate further. If you don't have quite such a personal connection to the charity, here's a more generic template you could use: Hi CHARITY_NAME, I'm a supporter of your work, and I wanted to write to ask if you're signed up for Google Grants. The program offers$120,000 per year of Google ads for free to registered charities in many countries. It doesn't take very long to apply. When I Googled {CHARITY_NAME}, I didn't see an ad for you folks, which is why I thought you might not be registered yet.

Let me know if you need any help or would like further info on this!

Cheers,
YOUR_NAME

Because many charities don't use most of their Google Grants budgets, being signed up is just a first step. Dedicating someone to spend a few hours per month refining the campaigns is another important task that requires more consistent attention. If you have a good relationship with the charity and can inquire about its level of diligence here, go ahead and do so.

While Google ads are basically free money, they're not pure altruistic value dispensed from thin air. Ads don't create time or wealth but move it around, so their utility comes from directing people to things that are more useful than what they would have found instead. As a result, it's important to be careful about counterfactuals when encouraging Google Grants and running ad campaigns.

Spreading Google Grants to effective charities is valuable in part because it helps give them advantage over less effective charities. If you sign up ineffective charities for Google Grants, you may compete for the same keywords and divert traffic that would have gone to the better charities. This is more of a concern for ineffective charities that directly compete in some area with keyword overlap -- e.g., animal shelters competing with a wild-animal charity for phrases like {helping animals}.

In addition, Google may in the long run decide how much it's willing to spend on Google Grants based on usage volume, and signing up more ineffective charities might eventually mean less allocation per charity.

Finally, if ineffective charities have greater publicity, they may divert funds from people who would have given to better charities. Or if the charity is not merely ineffective but actively harmful, Google Grants allows it to do more harm.

### Choose keywords to attract people who wouldn't have visited anyway?

If you're a business aiming to sell products, you'll choose keywords that maximize the inflow of users to your site who make purchases. You don't care whether they would have bought the same product from another supplier. Indeed, you might prefer to steal customers away from your competition. You'll probably choose keywords that are specific to your product because those have highest click-through and purchase rates.

If you're a charity aiming to introduce people to a new set of philosophical and practical ideas about altruism, you need a different mindset. You don't care so much about directing people to your specific organization but rather wish to bring in more people to the general milieu of folks who think hard about how to make the biggest impact on the far future. So if your ad uses the keyword {effective altruism}, it's not obvious how much benefit this has. Google's organic results for that query are already full of relevant links, and it's not clear that an ad would contribute anything on top of those.

There is research suggesting that ads mostly don't take away clicks that would have gone to organic search results. This particular study was done by Google, which has a vested interest in encouraging more advertising, but it also seems to be confirmed by anecdotal reports. However, if the whole page of results is full of relevant content, I'm more skeptical that ads would, err, add any value unless you're selfishly promoting your particular organization to the exclusion of other similar organizations. So choice of keywords is, err, key.

As another example, say you're linking people to a factory-farming video, like "What Came Before." It wouldn't make a lot of sense to place an ad for the query {factory farming video} because Google's organic results are already littered with relevant videos. In any event, there are already ads for this query, so you'd be facing stiff competition. What you might try instead is targeting a demographic of people who wouldn't necessarily have found the video anyway. For instance, you could target queries related to celebrities who are veg*an, with ad text like "Find out why [celebrity] has changed his diet." Ads like these have the virtue of not preaching to the choir and instead bringing in genuinely new audiences. Of course, celebrity names may be expensive, so you'd want to play around with keywords that are both acceptably cheap and targeted toward new and relevant demographic groups.

Likewise, for an organization about risks of future technology, keywords relating to the technology itself could prove better than keywords specifically about technological risks, because it's the former group that more needs to hear the message, even if the latter group is more likely to donate.

In general, there's some balance to strike between reaching a mass audience with low interest vs. reaching an interested audience where your message has little counterfactual impact.

Unfortunately, Google makes it hard to take a counterfactual-aware approach very far. If you don't have good click-through rates, your Quality Score suffers. You also can't apply for GrantsPro status if your click-through rates are below 1%. It just doesn't work to push too hard trying to divert traffic from atypical keywords. So consider the advice of this section to only apply around the margins of your campaigns. Do still focus on click-through as a main metric.

### Reaching out to diverse charity sectors

Charities in the same sector will somewhat compete for keywords. Fortunately, in practice, we'd expect diversity among the keyword sets that charities choose, but in the limit where many charities all competed for the exact same set of keywords, signing up more charities for Google Grants would just level the playing field for the new charities versus those already signed up, because two charities can't both rank at the top ad position for the same query. This suggests that the charities we sign up should typically be better than average in their field in order for signing them up to be positive. (And here I actually mean average, not median, in case the two differ.) In practice, keywords don't completely overlap, which gives some tolerance for slightly worse charities to be potentially worth signing up as well. However, the worse charities may steal not just ad space but also donors and labor from the better charities, so there are other considerations against promoting them besides ad competition.

One solution to keyword competition was suggested by Nick Beckstead in a comment on "Scaling Google Grants for Animal Charities": Charities could coordinate to each use different keywords. In part this may already be done by pricing mechanisms, since the popular keywords will show higher required bids. Pricing is more likely to avoid keyword duplication the more charities optimize their campaigns to lower cost per click rather than blindly picking keywords and sticking with them indefinitely. Beyond that, coordination could theoretically be carried out more directly, especially for close collaborators, though it seems pretty messy in practice.

Another way to help avoid the problem is to sign up charities in diverse sectors: some in animal welfare, others in international relations, others in improving governance, others in popularizing social science and philosophy, others in far-future research, and so on. These domains are less likely to share the same keyword space, so as long as the charities you sign up are better than the average of those already signed up in those domains, it's a win.

The web has no shortage of AdWords tutorials and tips, so I recommend checking them out. This section gives a few suggestions from my (limited) experience, but they're not meant to substitute for a more basic introduction (such as this video: "Google for Nonprofits Workshop - Google AdWords").

That said, don't feel like you need to become an AdWords expert before you set up Google Grants. Chances are you can get something workable with just a few hours of fiddling around, and then you can improve it over time as you learn more. Don't let the best be the enemy of the good, and don't shy from getting Google Grants at all just because you're not going to invest the time to optimize them heavily.

### Log-in and editing requirements

This is the single most important point to heed because your account may be suspended if you neglect it. According to "Active account management policy":

I keep a Google doc with the date of last login and date of last change, to make sure I'm meeting these requirements. You can see the date of last change using "Change history".

To get a sense of what words Google thinks are in your landing page, you can click on Tools > Keyword Planner > "Search for new keyword and ad group ideas" and enter your url. Of course, you can also just look at the text of your page.

Now, create an ad group with an ad text containing some of those words. For example:

Giving wisely
effectivealtruism.ch/
Learn which charities save the most
lives and stop the most suffering

Then create keywords for the ad group that contain word matches against the ad text, to improve the ad's relevance score for those keywords. For the above ad, I used these keywords:

• {giving} -- unigram match, high-volume keyword, reasonable cost per click
• {giving wisely} -- bigram match, lower volume, excellent relevance score for keyword to the ad text
• {smart giving} -- unigram match, lower volume, theoretically could have been separated out with a different ad group for higher relevance, but volume was too low for me to bother in this case.

In general, the keywords should have at least some match with the ad text, ideally a lot of match. Ad title matches are probably slightly better than ad body matches. If you want to use a keyword that doesn't match well with the current ad group, create a new ad group with its own text to better match that keyword. Indeed, the ideal situation might be to create an ad group for every keyword for maximal match with that keyword, but this seems slightly excessive.

I'm told by experts that you should generally have fewer than 10 keywords per ad group to maintain high relevance. One expert recommends 5-15. Some people recommend one keyword per ad group, though this takes a lot of work.

It's also recommended to have several ads in each ad group because Google by default rotates them to pick the best-performing ones.

You can use the Search terms report to optimize your keywords based on actual user queries.

Here are some of the other ads I created for the Effective Altruism Switzerland campaign:

What's the best charity?
effectivealtruism.ch/
Learn what science & rationality

Keywords: {charity}. Very simple and high volume, yet cost per click wasn't that bad.

Where should I donate?
effectivealtruism.ch/
Learn which charities help
the most people & animals

Keywords: {donate}, {where should I donate}.

effectivealtruism.ch/
Where you can do the most good
with your money, time, and career

Want to end speciesism?
effectivealtruism.ch/
Some animal charities are
more effective than others

Keywords: {speciesism}, {end speciesism}.

Note that most of these keywords are also good in the sense of not preaching too much to a choir of already-persuaded effective altruists.

## Acknowledgments

I first learned of Google Grants from Jesper Östman in 2012. In Jan.-Feb. 2014, Xiomara Kikauka and Joseph Savoie indirectly inspired this piece by their assistance with the Google Grants for Animal Ethics. Xiomara, Joseph, and Louie Helm gave helpful feedback on this piece, and Louie was inspired to write this blog post in reply: "MIRI's Experience with Google Adwords."

### Open Google Grants in another browser

Candidate ads that might appear on a given search are ranked by bid * Quality Score. If you win the auction, you pay the smallest amount P such that P * (your Quality Score) is higher than bid * Quality Score of all ads below you.

An important component of Quality Score is the click-through rate (CTR) that your ad got on previous searches. If we approximate Quality Score by CTR, then your bid of some cost per click (CPC) is ranked according to

CPC * CTR = (some amount of money / click) * (clicks / searches) = some amount of money / search.

This makes sense, because Google wants to maximize its revenue per search and doesn't care whether you get lots of clicks at a low CPC or few clicks at a high CPC (see p. 1 of "Online Ad Auctions"). Of course, Google also cares about long-term customer satisfaction, and so additional components are included in the Quality Score, including the page quality of the ad's target and the relevance of the ad text to the query.

If you use narrow keywords that preach to the choir, you may get a high CTR and hence higher Quality Score. You can then pay less per ad, though of course, you'll have more total clicks. Ads targeting new populations require you to pay for fewer clicks but may have low Quality Scores, so this could hamper their ability to be shown. This is especially problematic for Google Grants where the max bid is capped at $2, so you can't rely on high bids to make up for your low CTRs. If keywords operated in isolation, then there wouldn't be harm in having lots of low-quality keywords so long as you set the bids low enough that they weren't particularly expensive. However, in practice, the volume-weighted aggregate of Quality Scores across your account affects your future Quality Scores, so it may not be acceptable to have really low Quality Scores even if the ads are cheap. Thus, in practice, you'll have to strike some balance between maintaining diversity and reaching new audiences vs. not degrading your Quality Score too much. ### Lowering cost per click Given an AdWords budget of$330 per day, your first goal as a charity is to create enough ads to use the whole budget each day. If you do that, your next goal is to minimize the average cost per click (CPC) so that you can get more total clicks. If you have conversion tracking enabled, you could aim to minimize cost per acquisition (CPA), and the analysis would be similar as for CPC with the added consideration of conversion rates.

I see two main ways to reduce CPC. Several of the sub-points in each section parallel standard advice, like the suggestions in "4 Practical Ways to Lower Your AdWords CPCs."

If you have the potential to trigger for more impressions, the total number of auctions that you happen to win will be higher, so to get your ad to show the same number of times, you can lower your bid on any given impression.

The main way to expand coverage is to create more ads and keywords. Expand out into the tail of keywords, and think outside the box for new search terms. When you add new keywords, create new ads containing those keywords so that ad relevance remains high. There are tools to help brainstorm new keywords, like the search terms report.

The main risk with creating lots of keywords is that you might degrade your account-level Quality Score at the outset if the ads don't perform well, so make sure the keywords you do create are relevant. Using just a few keywords per ad group while focusing on scaling out ad groups should help in this regard. Often it's advised that you start with just a few strong keywords to get a green Quality Score rating at the beginning, to prove your account's overall trustworthiness, and then you can scale out to more and more ad groups to lower CPC.

#### Improve Quality Score

Since your rank in the auction is bid * Quality Score, you can get the same rank with a lower bid if you have a higher Quality Score. This means

2. Higher ad relevance: As discussed previously, the main way to achieve this is to have many different ad groups, each with a small set of keywords that are targeted to have text matches with the ad(s) in that group. For instance, if you have this ad:

How to best help animals
effectivealtruism.ch/
What science and rationality say

then for keywords you might have {animal activism}, {help animals}, {activism for animals}, and so on. They key is to choose phrases that strongly match the ad text (unigram match, bigram match, etc.). Creating lots of ad groups and tailored keywords takes some more work but may be worth it, both for higher relevance and for higher CTRs because your ad will look more worthy of being clicked to whomever issued the query.

3. Higher landing-page quality: Improving your landing page helps for all ads at once and is good to do anyway, although it's also unclear how much various changes actually matter to Quality Score. Google has a page explaining what they look for, including relevance of the landing page to your keywords and ad text, which implies that having more matching terms on the landing page helps. That said, it's not worth ruining the design of your page just to add relevant keywords.

#### Summary

In general, creating lots of ad groups and then lowering your bids on each group seems like the best way to optimize CPC because it solves all the other problems at once, except for landing-page quality:

1. More ads and keywords means more coverage.
2. More ads and keywords means you have more candidates to try out. When you try lots of things, you can just see empirically which ones tend to have lowest CPCs and keep those while cancelling the more expensive ones. This allows you to optimize CTR and ad relevance jointly.

#### Rough heuristic: Adjust bids to approximately equalize average CPCs

Suppose you have several ad groups, each of which has several keywords. All have associated average CPC values. In the long run, it makes sense to adjust your bids to make these average CPCs more equal. This is only a rough suggestion whose accuracy could be improved upon by more detailed calculations.

In particular, what you actually want to do is equalize incremental cost per click (ICC) -- or as the economists would call it, marginal cost -- across keywords and ad groups. You can compute ICC using Bid Simulator. However, this seems somewhat tedious for large numbers of ad groups and keywords, so an approximation can be to just look at average CPC, which is shown directly on the AdWords dashboard. In the limit that the marginal-cost curves looked the same across ad groups and keywords, equalizing average cost would imply equalizing marginal cost. This figure illustrates the relationship between marginal and average costs familiar from economics, although in the case of ad bidding, marginal cost only increases, rather than decreasing at the beginning, because increasing your bid to increase click volume only makes you win more and more expensive auctions.

Suppose you win an ad auction. The amount you pay is based on what the person below you bid and is just enough to allow you to beat her in ranking. So (your CPC) = (her bid) * (her Quality Score) / (your Quality Score), rounded up to the nearest cent, so long as this number doesn't exceed your bid. From this formula we can see that if you find that average CPC is near your bid price, this may mean it's a competitive keyword where other people bid high or that your Quality Score for that keyword is low. For other ad groups and keywords, your average CPC may be much lower than your bid, even if the bid is set the same as it was for the more expensive ad groups and keywords. Equalizing your average CPCs across ads and keywords then means lowering your bids in the expensive cases more than in the cheap cases. For example, if for one keyword you're bidding $2.00 and paying an average CPC of$1.70, while for another keyword you're bidding $2.00 and paying an average CPC of$1.10, you might lower the bid for the first case to, say, $1.30, so that its CPC will then hover closer to$1.10 as well.

One complication besides the difference between average and marginal cost is that different clicks may have different value. For instance, clicks in the USA are more expensive than in India, yet you might be willing to pay more for USA clicks because people in the USA generally have more money to donate, are more influential on world culture and norms, and so on. Theoretically you can check the countries from which you're getting clicks and account for this, although in practice it's probably easiest to assume the country distribution for different keywords is comparable. In any event, this is just one illustration of why purely equalizing CPCs across your keywords and ad groups is not a foolproof procedure.

### How much tweaking is optimal?

The most important part of Google Grants is signing up and using the full budget, but if you have extra bandwidth, you should spend some time optimizing your campaigns to reduce CPC. But at what point should you stop putting effort into creating more ad groups?

Suppose you have $120K of AdWords per year that you're optimizing. Maybe your tinkering with the campaigns will provide value for the next 2 years, after which the ad landscape becomes too different for what you do now to have value. And let's say you can spend an extra 11 hours to lower your average CPC from$1.50 to $1.40 for those 2 years. How much did this pay off? If you hadn't done the optimizing, you would have gotten 2*$120K/$1.50 = 160K clicks. With the optimization, you got 2*$120K/$1.40 = 171K clicks. That's an extra 11K clicks in 11 hours, or 1000 clicks per hour. If you value clicks at, say,$0.10 each, that's like $100 per hour. So this optimization seems worth doing. You could keep estimating the marginal gains from additional effort and stop when they drop below some threshold. Of course, only bother with this once the fundamentals are in place. And it's plausible that encouraging other effective charities to get Google Grants is more valuable per hour than tweaking a few more cents away from your average CPC. ### Grantspro$40K/month limit

In June 2014, Google reopened a program called Grantspro, which allows qualifying charities to increase their spending caps from $10K/month to$40K/month if they meet certain conditions. These criteria require more skill than those for a standard Google Grants account. They include setting up conversion tracking, having hit your budget cap in previous months, and maintaining a CTR above 1% (which is not trivial to do; only two of the four charities whose Google Grants accounts I've helped with have achieved it). It might be worth investing in a professional pay-per-click company to help set up these features in return for the added value that Grantspro provides?