Dave Guarino

Why I find Google Search ads valuable for user research

Recently, I shared a write-up of some work I worked on with a few wonderful colleagues (Anne and Justin) at the California Office of Digital Innovation, where I've done varied work for the past ~1.5 pandemic years or so.

This work was fundamentally about identifying — in an acute and actionable way — barriers Californians were facing in accessing the Emergency Broadband Benefit (EBB), a pandemic benefit providing $50-$75 per month for internet to lower-income Americans.

Anne made a comment on Twitter:

So I figured I would briefly detail why I really like having Google Search ads ("AdWords") in the toolbelt for doing user research.

1. You can reach people right at the point that they have the acute need you're looking to understand (aka "already activated") #

Talking to someone who, 15 minutes ago, searched for "cheap internet" or "internet discount" is very different than talking to, for example, someone who is generally eligible for an internet subsidy because they live on a lower income.

The slightly more jargon-y way to put this is the person is already activated: the need is top of mind, and they're pursuing it — right now.

Talking to people who may want but don't have broadband internet at home is also useful; it's just getting at a different piece of the puzzle.

In our case, we wanted to understand what might be standing in the way of a person who is eligible and has already decided they have this need.

And that's precisely the research context where connecting with someone based on a search query they made is so useful.

2. With ads, you can control volume (# of people who see the ad and who you reach out to) #

You can turn ads off and on at will. You can also set a daily budget target that will generally constrain your spend (and how many people see it.) It's not perfectly precise throttling, but it's absolutely good enough for most cases.

Controlling volume is particularly valuable if you want to have fairly involved conversation with someone, like we did actively helping folks through the application as far as they wanted.

It's also helpful if you're, for example, not wanting to drive a massive amount of traffic to the landing page because this is your very first exposure of users to whatever content you put there. Many government sites are so high-volume that a direct link from, say, a main page would drive an immense amount of traffic. While you should work hard to make whatever landing page content is there valuable, no amount of thinking and working hard on that content is a substitute for getting view into how actual users engage with it. (To riff on an old saying: not even the strongest content design survives first contact with actual users.)

While search ads are one way to control volume, you can also achieve that in other ways for this kind of recruitment:

3. While you might not have set up AdWords, other parts of your organization (even in government) may already have done that #

Google Search ads are generally a tool for marketing and digital outreach.

While a research or technology team may not already have done that, a public affairs team or another department might have.

Plenty of government agencies use AdWords for digital outreach already in fact. (As an anecdote, here in California I have seen search ads for getting a vaccine from a public health agency.)

So even in organizational contexts where it's difficult to buy things like pay-as-you-go SaaS, look around and you may well find some smart colleague over in another spot has already navigated all that (albeit not for your research uses!)

Aside: even if you are in a small, nimble org, setting up AdWords may be a little intimidating (it's a very powerful tool and design for power does not necessarily lead to simplicity.) It may not be the most zero-friction onboard, but you can definitely do it.

4. You can reach users without creating work for others in your org (or in a partner org) #

A common way to recruit people for research on a web site is to put that intercept on the page. Let's say I'm trying to understand user' unmet needs in navigating the Department of Some Benefit (DSB.) Your first thought might be to put an intercept on the DSB home page. And yes, that's great!

But in doing that, you're creating work for others when you don't have to. If you step back and think about your work with the incredible point of contact you have at DSB as a design problem itself, you're actually making the choice a little bit harder for them because now they may have to go and try and get this prioritized, and communicate with other folks, etc. Sometimes that's worth it to the end of, for example, creating deeper buy-in.

I happen to be of the opinion that if you have not talked to at least one user though, choose whatever path absolutely minimizes the Time To User (TTU.) It's also nice to be able to say to a partner, "we're going to talk to some users — no action needed on your end, let us know if you have any questions or concerns!"

5. You are reaching people before they hit any friction or barriers in the actual service that may push them out #

For a lot of people, a web search is the first step to anything. And it's on the whole an extraordinarily low friction step to take.

So if you cast your net there, you can reach a really wide swath of users.

By contrast, intercepting users later in some "journey" may add selection bias of in favor of users able to jump over some hoop or point of friction.

One example: intercepting users at a certain question screen in an online benefits application is great to understand how people might struggle with that question. But did users have to create an account to get there? Now you're maybe losing vision into those folks' experience.

There's no perfect approach, but the point is that there are such low barriers to making a web search that you get people before any kind of friction in the process pushes anyone out. That's pretty valuable, particularly for services with "heavy front doors" (say, complex account creation or identity proofing steps early on.)

6. Being forced to write a landing page with some value for users is a good forcing function to actually "work the problem" (the MMM-MVP) #

For our EBB research, we didn't just run ads saying "we're doing research please talk to us!" We wanted to provide actual value to Californians who were searching for a need they had, since this benefit was now available to them.

So we had to both make ad copy language and write a landing page (in plain, friendly language) that actually communicated the benefit in a way that could create understanding and clarity on (if they wanted) how to go get it.

And doing that led to very practical knowledge development. Whereas before it had been pretty abstract thinking about the benefit, being forced to write content actually forced us to grapple with substantive questions (and similar to what users themselves would have to tackle.)

Example: Do families with kids receiving free and reduced lunch qualify? Yes, but what about the special pandemic situations where families who don't normally get it did? Have to dig into that!

It's not a lot of work, but it's very different work than background research because of the practical motivation of trying to clearly explain this thing to real people.

Limitations and constraints #

Search ads are just one tool. And of course it has limitations! (If you've ever met me, you'd know any totalizing view of some single lever "solving" a messy complex problem is decidedly not my jam.)

Depending on your aim, you can and should also talk to people who prefer offline (in-person) services; who don't go to search first; who search in languages other than ones you speak; etc.

So thinking about how this complements other approaches is good. Particularly, how does this stack up against how your current default way of finding people to talk to? What different sampling biases does this present relative to the sampling biases in those approaches?

(For example, in our case I wanted to make sure we reached users out in more rural parts of California — instead of that being a massive lift and only being limited to one locale we could visit, we were able to talk to multiple people from very different parts of rural California. On the flip side, we didn't really get to talk to people who were completely cut off from internet whatsoever: by definition we were reaching people who often were using smartphones to search or a library or work computer.)

But a big virtue of this particular tool is — as we found in our EBB work — it's one of the lowest-friction ways to "get out of the building" and talk to the people who are the only true experts in their own experience of a thing. And speaking only from my own experience, the longer you wait to do that — and the more you let get in your way — the more likely you are to develop mental models that are detached from that experience.


If this was useful to you, you can sign up for very occasional emails from me on things like this.