Paid Growth, Step by Step: Shopify's Performance Marketing Lead on Tests, Signal and Team Design

Davis says "performance marketing" and "paid growth" can be used interchangeably, but he recommends specifying what you mean. Historically both meant online only, though he argues offline channels and affiliates can now be included. Lenny sums performance marketing up as marketing whose performance you can measure, and Davis agrees.

Asked whether every company should do it, his hot take is yes. Google now shows AI results and paid listings take the first four spots, so organic listings sit far down the page, and on Meta you almost need promoted posts to be seen. Even influencer marketing, he notes, is a paid component.

He singles out paid search as the baseline for everyone, because it is user-driven: your ad shows only when someone types a relevant keyword. Everything else, such as ads on Meta or between YouTube videos, is disruptive media that may not suit a business at every stage.

To judge whether paid could drive most of your growth, Davis asks where your users are. Your analytics will show which platforms already send you traffic. If something works organically, such as TikTok, ask what happens if you turn it up to 11 with paid.

He gives the example of Hairstory, which was strong on Google Shopping. Analytics showed many visitors coming from Meta and TikTok, then very new, so the team tested Meta with customer testimonials and TikTok with the same assets. What worked on Meta did not work on TikTok, because the user mindset differs, though he recalls the eventual results as exponential.

The lesson is not to copy creative from one platform to another. Each has a different experience, and the right test builds on data you already have.

Davis borrows a phrase from his director: signs of life. You put a little money into a platform to see whether there is any response, then pull back and build a proper campaign. Spending big just because there is a signal skips the questions of whether you have the right creative, campaign and messaging.

His top tip is to start with your own data: load your customer base into the platform to build lookalike audiences. On Meta he builds ad sets at 1%, 2-4%, 5-7% and 8%-plus, where a smaller percentage means behavior more closely correlated with your customers. The widest tier usually fails, but he still tries it. With a limited budget, he suggests running only the 1% audience.

Telling whether creative or targeting is failing is a million-dollar question, he says. Check click-through rate first to see whether the right people are engaging, then separate creative from content with control-versus-test setups. Ultimately a focus group is the only way to learn why an image does or doesn't resonate.

Test length is often a budget decision rather than a statistical one. A small test can still support a business case: if $1,000 produced a given reach and click-through rate, you can estimate what $10,000 would return. He also urges teams not to be hard on themselves, since each platform has unique behavior and "we're either winning or we're learning."

Davis says Google comes first, and he clarifies that it includes YouTube, Search and the Google Display Network. If you are just starting, he suggests Search, then Meta (Facebook and Instagram), then YouTube if you have video creative. He is bullish on video as long as you can feed a flywheel of fresh creative, and advises starting with emotion, whether comedy or happiness, because people act on what they remember.

LinkedIn looks almost three times more expensive than other channels, but its targeting by job title, industry and company is powerful. At SoftLayer, trying to win Coca-Cola's cloud business against AWS and Microsoft, the team learned the buyer worried about recency and private security. They ran LinkedIn ads on those themes to Coca-Cola employees and geo-fenced the Los Angeles office where the decision-makers sat.

The result was that the salesperson was told on the next call that security and recency concerns were understood. Davis still advises testing Google and Meta before LinkedIn unless you are an enterprise company with high-value deals.

For emerging channels he points to connected TV and podcasts, which are doing well for people who measure them correctly. He hasn't experimented with VR or AI placements and prefers to let others be the guinea pigs.

For startups, Davis begins with the goals and timeline. If you want results quickly, paid needs to start immediately because SEO takes time and a new category may have no search demand at all. In one case a hotel-gifts product bid on generic hotel-booking keywords, and moving the budget to display to build awareness, rather than search, is when the company started seeing returns.

If your product resembles an established one, he describes bidding on a competitor's name as "coattail riding." Whether paid works depends on when you expect results and how much demand exists.

On running paid before product-market fit, he cites IBM's Africa efforts. Users were interested, but the company billed in dollars across a continent of many currencies, so they weren't converting. Once it focused on South Africa and supported the rand, results improved, and he advises against buying awareness for a product people can't yet convert on, since you only annoy them.

Asked how he beats agencies, Davis says many follow copy-and-paste playbooks and don't go deep into the weeds. He runs an ops cadence covering finance, performance, structure, keywords, ad copy, quality score and targeting, each with specifics reviewed weekly, biweekly or monthly. Agencies with 50 clients tend to switch on automated bidding, run search query reports and move on without asking about landing pages or shortening a five-step conversion flow.

As a manager he avoids stretching people too thin, holds weekly one-on-ones and encourages them to say when they are overwhelmed. He believes in hiring smart people and getting out of their way.

Agencies and consultants are a good place to begin, he says, especially if the owner shouldn't spend time in ad accounts. He favors forward thinking and backwards planning: agree a milestone, such as hiring someone full-time at $50K a month in spend, with the agency up front. A partner who resists that conversation is a red flag, and sometimes the new hire keeps the agency on for specific channels at a lower fee.

Davis cites Nate Silver's *The Signal and the Noise* as his hiring lens. Platform mechanics can be taught, but accounts produce endless metrics, such as reach, frequency, CPM, CPC, conversion rate and cost per lead, and someone has to decide which matter. That requires a data-minded person.

His love of data comes partly from baseball: he played, then blew out his arm and stayed with the stats, such as which hitter belongs where in the lineup. Early in his career he took every project he could, including SEO, email and affiliates, then fell for paid when he saw it show exactly what keyword was bid on, how many saw the ad and how many clicked.

In interviews he throws a pile of data points at candidates and asks how they would optimize the account. Overwhelmed candidates miss the point, while the ones he likes ask about the campaign's purpose. If the goal is conversions, impressions and reach barely matter, whereas awareness campaigns count viewers and consideration campaigns count funnel entries.

On metrics like CAC and LTV, Davis says to hope for a strong finance partner who sets guardrails. For account analysis he screen-shares Google Ads exports comparing brand and non-brand on expected click-through rate, landing page experience and ad relevance. In his example, brand ad relevance lagged, and a single poor-strength ad with a high CPC could be paused, since he believes quality score also operates at account level.

Google itself recommends fixes, such as adding unique headlines or more keywords, and he says people often ignore that. For benchmarks he advises talking to platform reps rather than seeking generic numbers. If you name your top competitors, reps will say whether your click-through rate, conversion rate and cost per click are above or below an anonymized threshold.

He distinguishes impression share, which he calls ego marketing, from click share, which reflects reaching the right user. In a 2023 test, changes left impression share flat but lifted click share. He also recommends a PPC Hero article by Jacob Brown that combines Auction Insights metrics to show how often a competitor truly outranks you. He admits much of his analysis is manual and asks listeners to share any AI automation.

Davis favors multi-touch attribution with time decay, since people forget quickly where they first met a brand. The first touch deserves some credit but isn't the reason they converted. Attribution by itself is also biased because it doesn't say whether the person would have converted without the ad. Netflix and eBay ran incrementality studies, and he says eBay cut nearly all brand spend around 2012 after learning users would find it organically.

Most big companies he has worked with, such as Pinterest, Shopify and IBM, build attribution tools in-house, so he has no third-party favorite. He advises leveraging in-platform tools because the main thing in performance marketing is the signal you send the platform. If the optimized action doesn't match business results, you are feeding it noise.

Incrementality should come from geo experiments or conversion lift tests, whose results feed what he calls an incrementality adjusted factor for each channel. In a lift test you deliberately don't show ads to some winning auctions, which creates a control group and a spend holdback. Platforms will partner on this if you spend enough, but below about $50K a month it's more work than the signal is worth.

Davis says to start conversations about hiring around $50K a month in spend, since filling a role can take three months. The first hire is the data-driven person who works in the platforms, followed by a creative with a graphic design and branding background, and third a dedicated data scientist for incrementality tests and reporting. He wants ad copy to be collaborative, and recalls an ADT test where a simple zero-setup, zero-install ad beat a carefully packed value-prop ad.

The first hire's title is usually growth marketing specialist or manager, because the person will wear many hats, with specialists added as the team grows. At Shopify the culture is "GSD," meaning hands-on keyboards in the ad accounts, with a calendar of tests that ends in a readout of what was learned.

To avoid bloat, his manager built a calculator where each person totals meetings, time off, optimizations and reporting against the days in the quarter. If someone is in the red for several quarters, the team first cuts meetings and then builds a business case for a hire. The ops cadence is a spreadsheet of big rocks, such as keywords, and smaller pebbles, such as negatives, with their frequency, which also lets cross-functional partners see how often each task is done.

He tries to cut time to impact for new hires from about 90 days to 45 or 30, citing *The First 90 Days*. He gives them early ownership of a campaign, shows them his own workflow live in one-on-ones, and doesn't expect perfection.

On Apple's ATT changes, Davis relays his mobile teammate's answer: as long as SKAdNetwork can provide attribution and measurement for iOS, the team is fine. Asked about creative, he says people vastly underestimate it, citing Dollar Shave Club's video. Strong creative tells a story and creates emotion, whether through comedy or heartstrings.

When competitors appear to steal your traffic, he says not to assume malice. Close-variant matching may be showing their ads, or they may simply be mismanaging their account. Check the competitive report, see whether the pattern is consistent, and file claims for trademark violations or misspellings of your brand name.

On AI, he relays Google's reminder that Smart Bidding and recommendations are already AI, so the conversation is partly a reset. He sees a big impact on content and creative, such as Google's dynamic ad builder. He also warns that AI could be abused to slip ads through Google's review in small drips.

If a creative agency gives you everything you want, you may not need an in-house hire, Davis says. In his experience, though, in-house creative matches tone and look better and can iterate quickly without agency hour or budget caps. AI could help a lone marketer, as with Google's dynamic ads and likely similar Meta tools, though he says scrappy output isn't always the best route.

His closing advice is "forward think, backwards plan." Decide where you want to be, such as on every platform, then work back to what you can do now, starting with search, which needs only keywords, and then add creative. At Shopify an emerging channel earns "performing" status by hitting set thresholds, like a thousand conversions a month at a given spend and lift, with milestones along the way across ad formats such as feed, conversation, video and carousel.