Upper Funnel KPIs on TV Have Downstream Implications

TV has always been a powerful driver of upper-funnel metrics like site visits, and it drives purchases too — just on a longer timeline. That gap raises an obvious question: does a change in visits actually say anything about what will happen to purchases weeks later, or do visits represent a different, lower-intent audience that just makes the upper-funnel number look better than the eventual business outcome? We decided to find out. We compared 127 million site visitors with 5.1 million purchasers across 300+ advertisers. We discovered that their household-demographic profiles were strikingly similar.

Here’s the spoiler: a metric like site visits that shows up in days is quietly foreshadowing a metric that will be measurable for weeks. This is what it means for upper-funnel KPIs to have downstream implications.

Why Visits Can be Useful as an Early Signal

Site visits appear quickly. Purchases and other conversions often take weeks to reach enough volume for a reliable signal, especially in industries with long or complex decision cycles. In the case of financial services, insurance, pharmaceuticals, and automotive, the final outcome might be an approved application, a new policy, a prescription; variables that could take weeks or even months. Therefore, outcomes can occur well after the initial site visit, making visits a useful early indication while conversion data develops.

That difference in timing creates a real tradeoff. Rely too heavily on visits and you could favor inventory that produces traffic without real sales. Wait only for purchases and several weeks of budget may pass before you know whether the campaign is reaching the right audience. The narrower question is whether visitors and purchasers are meaningfully different.

What We Measured

We studied 306 advertisers on the Tatari platform over an eight-week period. For each advertiser, we used its site pixel to identify everyone who visited and everyone who completed a purchase: 127 million visitors and 5.1 million purchasers in total. We matched both groups to modeled household attributes from a third-party data provider, then compared their profiles across 173 attributes in 26 categories, including age, household income, education, household composition, homeownership, length of residence, and technology adoption among them.

Purchase events are not standardized across advertisers: one pixel may report checkouts, another overall purchases, another separate events for first-time and returning customers. We used each advertiser's own purchase event, applied consistently across the eight-week window.

Visitors and Purchasers Look Very Similar

For every one of 173 household attributes, we asked how common it was among visitors and among purchasers, pooling all 300+ advertisers together. Take whether there are children at home: 32.0% of visitor households had them, and so did 31.9% of purchaser households.

That near-match was the rule rather than the exception. Across all 173 attributes the typical difference was about a tenth of a percentage point. The two audiences look nearly identical.

Purchaser share versus visitor share for all 173 household attributes (pooled across advertisers)

Figure 1. Purchaser share versus visitor share for all 173 household attributes. Points on the dashed line have an identical mix. Pooled across 306 advertisers, eight weeks.

Because the chart includes every measured attribute, the result is not driven by a small set selected after the analysis. The full distribution lies close to the line of equality.

Household income shows the same pattern in a more familiar format. Visitor and purchaser distributions are nearly identical across nine income brackets, although purchasers are modestly more likely to fall in the highest bracket.

Household-income distribution for purchasers and site visitors

Figure 2. Household-income distribution for purchasers and site visitors. Labels show purchaser share minus visitor share in percentage points. 306 advertisers, eight weeks.

The Pattern Holds Advertiser by Advertiser

An aggregate result can hide offsetting patterns, so we repeated the comparison separately for each advertiser. For the median advertiser, the typical difference between its purchaser and visitor profiles was minimal.

Distribution of each advertiser’s median difference across 173 attributes

Figure 3. Distribution of each advertiser’s median difference across 173 attributes. The final bar includes 13 advertisers above three percentage points. 306 advertisers, eight weeks.

The pooled result is therefore not an artifact of opposite skews canceling one another. Most advertisers show the same close alignment.

Where the Groups Differ

The largest gaps lean in a consistent direction: purchasers are slightly wealthier and slightly better educated than visitors.

The ten largest purchaser-minus-visitor differences

Figure 4. The ten largest purchaser-minus-visitor differences, excluding duplicate encodings of the same measure. 306 advertisers, eight weeks.

We checked whether these gaps were just noise: they became more visible as purchaser counts increased, a pattern more consistent with a small recurring effect than with random sampling variation. We're treating the statistical wording as provisional until a planned per-attribute equivalence test is complete.

The practical interpretation is modest: purchasers skew slightly upmarket, usually by one or two percentage points. They do not appear to be a different demographic audience.

How Marketers Should use these Findings

• Use visit-based response as an early directional signal. If site visitors resemble eventual purchasers, visit metrics can provide an earlier read on whether a campaign is reaching the intended audience.

• Make earlier changes reversible. The result supports using visits to guide low-risk, near-term adjustments while purchase volume is still developing. It does not support committing to a long-term strategy on visits alone.

• Keep purchases as the final measure. Similar demographic profiles do not make visits and purchases interchangeable. Purchases remain the outcome that determines whether the campaign created business value.
  
  

What this Analysis Does and Does Not Show

A few limits are worth naming. The visitor group is everyone the site pixel recorded, not only people who arrived from TV. Because buying requires visiting, purchasers also appear in the visitor group — though they are only 3.7% of it, and rebuilding the comparison against visitors who never purchased widens every difference by about 4%. Customers acquired before the window look like visitors here too; eight weeks cannot identify them, but even if a quarter of all visitors were past buyers, the largest difference would still be just 2.4 points. Demographic similarity is also not the whole story: two audiences can match on income and age yet differ in purchase intent, category affinity, or lifetime value.

The Takeaway

Across the advertisers we examined, site visitors and purchasers had very similar household-demographic profiles — purchasers skewed slightly wealthier and better educated, but only by a point or two. That matters because it removes the most common objection to acting on upper-funnel metrics: that visits come from a different, lower-intent crowd than the people who actually buy. They don't. Visits won't tell you how much revenue is coming, but they will tell you, within days rather than weeks, whether your campaign is reaching the kind of household that converts.

Want to understand what your data is telling you about potential customers? Let’s talk!