How to test more ad spend when direct is your top channel

When Direct is your top channel, test before you spend more. Usually that means finding the channel that opens journeys ending in Direct, and setting your break-even ROAS. Then raise that channel's spend in some regions only. Judge it on total sales against the regions you left alone, not on Direct's share.

Step by step

You need GA4 with purchase marked as a key event, a few months of data, and an ad account where you can set locations. Steps 1 to 6 take an afternoon. The test itself runs for weeks.

1. Confirm Direct leads in the session view. In GA4, open Traffic acquisition and keep its default dimension, Session default channel grouping. Note Direct's share of key events and of total revenue for the last full quarter. Session dimensions use the paid and organic channels last click model, so this row is a last-click label. Menu path: Reports > Acquisition > Traffic acquisition.
2. See how those buyers first arrived. Open User acquisition, which shows how new users find your site for the first time. Its dimensions start with "First user", such as First user default channel group. If Direct is far smaller here than in step 1, other channels bring people in and Direct is how they come back. Menu path: Reports > Acquisition > User acquisition.
3. Find the channels that open journeys. Open the paths report; Google's help also lists it as Key event attribution paths under the Key events dropdown. Its chart shows which channels initiate, assist and close key events, using Paid and organic last click by default. Switch the chart's model to Data-driven, then read the Early touchpoints segment. Menu path: Advertising > Attribution > Attribution paths.
4. Set your break-even line. Divide 1 by your gross margin and write the result down before the test starts. Shopify's sales reports define gross profit as net sales minus product cost, so your margin is gross profit divided by net sales. Menu path: Shopify admin > Analytics > Reports > Category: Sales.
5. Split your regions into test and control. Open Total sales by billing location, which shows the country or region of each order's billing address. Pair groups that rose and fell together over recent months. Keep each group large, because Google warns that very small location targets might show ads only now and then. Menu path: Shopify admin > Analytics > Reports > Category: Sales > Total sales by billing location.
6. Raise spend in the test group in Google Ads. Go to Locations within the Campaigns menu, click the blue pencil icon, select the campaign and enter the test locations. Leave the control group on its usual budget. Google calls location targeting a best effort and does not guarantee full accuracy, so expect a little spillover. Menu path: Google Ads > Campaigns menu > Locations.
7. Or raise it in Meta. Open Meta Ads Manager, edit or duplicate the ad, go to the Ad set level, then Audience, then Locations. If you pick a city or region, Meta may tick "Reach more people likely to respond to your ads" by default. Untick it, or the test leaks into control. Menu path: Meta Ads Manager > Ad set > Audience > Locations.
8. Read total sales first, then Direct. After at least one full buying cycle, compare both groups in Total sales by billing location. Then, in GA4's Attribution models report, click Add filter and select Region under User to see whether Direct grew where you spent. Menu path: Advertising > Attribution > Attribution models > Add filter.

A worked example

For illustration, say you sell in six countries and Direct holds half of your total revenue in Traffic acquisition. In this worked example, User acquisition shows Paid Social as the top first-user channel. The paths chart also puts Paid Social in Early touchpoints most often, so it is the channel to test.

Say your gross margin is 40%. On one store's Break-even sheet, 1 divided by 0.40 gives a break-even ROAS of 2.5x, the same arithmetic you would do.

Say the test group is three countries and the control is three others, each selling €20,000 a week before the test. In this worked example, you add €2,000 a week of Meta spend in the test group only, for four weeks.

In this worked example, control sells €21,000 a week during the test, a seasonal lift of 5%. If the extra spend had done nothing, the test group would likely have risen 5% as well. In this worked example it sells €26,000 a week instead.

So, for illustration, the extra spend bought about €5,000 a week of sales for €2,000. That is a return of 2.5x in this worked example, exactly your break-even line. The money paid for itself and left nothing over.

Direct rose in the test group too. That tells you Paid Social was feeding Direct, which is useful to know. It does not change the verdict: at break-even, more of the same spend adds work, not profit.

Then check the length. On one store's Journeys sheet, journeys of 2 to 3 touches took 12.5 days to buy. If your multi-touch buyers are that slow, four weeks covers two full cycles, and one week would not cover one.

What if the test result looks wrong?

• Direct rose, total sales did not. The spend changed how buyers arrived, not how many arrived. Some people who would have typed your address clicked an ad first. That is credit moving, not demand growing.
• Control regions rose as well. Location targeting is a best effort, so some ads reach people outside the test group. Meta can also reach past your locations if the reach-more box stays ticked. Compare the gap between groups, not each group's growth.
• An ad platform reports a big win that Shopify does not show. Ad platforms count conversions for their own ads. Google's data-driven model, for one, looks only at interactions on Google's Search, YouTube, Display and Demand Gen ads. Trust Shopify's order totals for the verdict.
• A sale or a new drop landed mid-test. A promotion in one group swamps the spend change. Run promotions in both groups or in neither, or move the test.
• The test group barely spent its budget. Small locations might not meet Google's targeting criteria, so ads show intermittently or not at all. Merge regions until the group can spend the budget.
• GA4 and Shopify place buyers in different regions. Shopify sorts orders by billing address, while GA4 derives location from IP addresses. Travellers and VPNs blur the line a little. Judge the test on Shopify's totals and use GA4 only for the Direct check.

What to do this week

1. Check Direct in two views. In GA4, note Direct's share of key events in Traffic acquisition, then in User acquisition. Pass: Direct is much smaller among first visits, so other channels feed it. Fail: Direct leads both, so tag your links before you test any spend.
2. Pull recent sales by billing location. In Shopify, open Total sales by billing location and sort your countries or regions into two sets. Pass: the two sets track each other week by week. Fail: one set swings on its own, so swap regions until they match.
3. Build the test campaign and leave it paused. In Google Ads or Meta, point one campaign at the test group with the extra budget. Pass: control regions are excluded and Meta's reach-more box is unticked. Fail: the test can reach control regions, so fix the targeting before launch.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: Traffic acquisition report (Google); User acquisition report (Google); Scopes of traffic-source dimensions (Google); Get started with attribution (Google); Key events attribution paths report (Google); Key event attribution models report (Google); Demographic details report (Google); About data-driven attribution (Google); Target ads to geographic locations (Google); Use location targeting (Meta); Sales reports (Shopify)

Related answers

• All Direct traffic questions
• Should I spend more if direct traffic is my best channel?
• Should a fashion store spend more if direct traffic leads?
• Holdout test, defined
• Incrementality testing, explained