Right idea, wrong mix: why a geo expansion converted at 1.8% on YouTube and 9% on Search

A geo expansion test produced a 1.8% blended conversion rate. The same cities, same offer, same period, produced a 9% conversion rate on Search. The difference was not the market. It was that 60% of the budget went to YouTube.

This is a common misread of what a channel mix test actually tells you. The expansion idea was right. The budget split buried the signal.

What the numbers actually say

A 1.8% blended rate with a 9% Search rate implies YouTube was converting somewhere around 0.3% to 0.5%, depending on how the remaining spend was distributed. That is not a YouTube problem in isolation. YouTube awareness spend converting at low single digits is normal and sometimes acceptable if the goal is reach. The problem is that this test was framed as a conversion test for a new geography, and 60% of the budget was allocated to a channel that does not close demand, it creates it.

Search in those same cities at 9% is a strong signal. It tells you that intent exists in the market. People in those cities were already searching for the product category. You did not need to manufacture demand with YouTube first. You needed to capture the demand that was already there.

The test did not fail. It answered the wrong question at the wrong cost.

Why this happens in geo expansion planning

When a team enters a new region, the instinct is to build awareness before driving conversions. The logic sounds reasonable: new market, new audience, they do not know us yet, we should introduce the brand first. So the media plan loads up on YouTube, display or connected TV, and Search gets a minority of the budget.

That logic is defensible in certain situations. It is not defensible when:

• Search volume data already shows category demand in the target cities
• The product has a short consideration cycle
• The team is running a test with a defined conversion goal and a limited budget window

In those cases, leading with YouTube is not a strategic choice. It is a default that comes from applying a brand awareness playbook to a performance question.

The other factor is that YouTube results are easier to show in a deck. Impressions, views, reach, frequency: these numbers look substantial even when conversion is low. A Search-heavy test with a smaller reach footprint can look less impressive in a slide, even if it is generating nine times the conversion rate.

How to plan v2

Step 1: Validate intent before you allocate

Before setting the channel split, pull Search Impression Share and keyword volume data for the target cities using Google Ads geographic breakdowns or Keyword Planner filtered by metro area. If monthly search volume for your core category terms is meaningful in those cities, that is your budget anchor. Search should lead.

A reasonable starting point for a conversion-focused geo expansion v2: 70% to 80% of budget in Search, 20% to 30% in YouTube or display used as a retargeting and brand recall layer, not a primary conversion driver.

If Search volume is genuinely thin in those cities, the original instinct to seed awareness first has more merit. But that should be a data conclusion, not an assumption.

Step 2: Separate the measurement goals by channel

YouTube and Search should not be evaluated on the same conversion metric in the same test window. YouTube's contribution in a geo expansion is brand recall, assisted conversions and search lift. Measure it with brand search lift studies, view-through attribution with an appropriate window (7 days is a common starting point), and incremental search volume in the target cities over the campaign period.

Search gets measured on direct conversion rate, cost per acquisition and impression share. Mixing these into a single blended conversion figure, as the original test did, hides what each channel is actually doing.

Step 3: Structure campaigns to isolate the geo signal cleanly

In v2, create dedicated campaigns for each target city cluster rather than a single campaign with geo bid adjustments. This gives you clean conversion data per market, clean budget control, and the ability to pause underperforming cities without disturbing the others.

For Search, use exact and phrase match on category-intent terms first. Broad match expansion can come in a second phase once you have baseline CPA data per city. Starting broad in a new geography inflates spend on irrelevant queries before you have enough data to exclude them efficiently.

Step 4: Set a realistic test duration

A geo expansion Search test needs enough conversion volume to be statistically meaningful. If your target CPA implies you need at least 30 to 50 conversions per city cluster to make a budget decision, size the test budget and timeline accordingly. A two-week test with five conversions per city is not a test. It is noise.

For most mid-size accounts, a four-to-six week window with a budget that can generate 30 or more conversions per cluster is a minimum viable read.

Generalizing to US regional expansion

The same framework applies when a US-based brand is expanding from its core markets into secondary regions. A brand strong in coastal metros entering the Midwest or Southeast faces a version of the same question: is there existing demand to capture, or do we need to build it?

The answer is almost always some of both, but the ratio matters. A few practical checks:

Google Trends by metro. Run your core category terms filtered by metro area. If the trend index in your target cities is 60 or above relative to your existing markets, demand exists and Search should lead.

Impression share in existing campaigns. If you are already running national campaigns, check whether your target cities are appearing in the geographic breakdown of your Search campaigns. Organic impression share in those cities, even without targeted spend, is a signal that queries are happening.

Competitor presence. Use the Auction Insights report filtered by geography. If established competitors are bidding in those cities, demand is there. You are entering a contested market, not an empty one.

When these checks confirm demand, the v2 playbook is Search-weighted from day one. YouTube can run in parallel at a minority share to support brand recall, but it should not be the primary budget vehicle until Search has established a baseline CPA you are comfortable with.

When demand is genuinely thin, a sequenced approach makes sense: four to six weeks of awareness spend (YouTube, display) to seed brand familiarity, followed by a Search ramp as you monitor whether brand search queries begin to increase in those cities. The mistake is running both phases simultaneously at equal budget before you know which situation you are in.

The compounding cost of a bad channel split

The 1.8% blended rate is not just a conversion problem. It is a data problem. When 60% of a test budget goes to a channel that converts at under 1%, the test produces very little actionable signal about whether the market works. You spend the budget, get a discouraging blended number, and either abandon a market that might have worked or re-enter it with the same flawed assumptions.

A Search-weighted v2 would have produced cleaner data faster, at a lower cost per data point, and with a conversion rate that gives the team a real basis for a budget scaling decision.

The original test also illustrates a broader issue in performance marketing: decisions made at the campaign planning stage determine what the data can tell you. A poorly constructed test does not just waste budget. It delays the learning by a full test cycle.

Where Almond fits

One reason channel mix errors repeat across geo expansion tests is that the reasoning behind the original test rarely gets captured anywhere a future planner can find it. The decision to allocate 60% to YouTube was made in a planning meeting or a Slack thread. The 9% Search conversion rate lives in a campaign report that may or may not get linked to the next planning cycle.

Almond's Business Context Engine reads ad account data alongside meeting notes, Slack and docs, and retains the test result alongside the reasoning. When v2 planning starts, the system surfaces the prior result with context: what was tested, what the split was, what each channel produced, and what was decided. The team does not rediscover the 9% Search rate by digging through old reports. It is already in the context layer before the first planning conversation begins.