Measuring Upper-Funnel Impact: How to Prove What Video and Display Deliver

Upper-funnel campaigns such as video, display, connected TV and social prospecting have a reporting problem, not necessarily a performance problem. They are built to create awareness and interest among people who are not ready to buy yet, but they are usually judged with click-based tools designed for the last step before a purchase. The result is predictable: they look expensive, budgets drift toward the bottom of the funnel, and growth slows months later without an obvious cause.

The short answer: measure exposure, not just clicks, and connect it to what customers do weeks later. In practice that means tracking upper-funnel metrics in layers, proving cause and effect with a geo holdout test or lift study, and using marketing mix modeling for budget decisions. Measuring upper-funnel impact this way turns awareness spend from an act of faith into a line item you can defend.

This guide explains why click-based reporting undervalues awareness, which metrics are worth tracking, how attribution, lift tests and marketing mix modeling fit together, what to do on a small budget, and how to turn the results into a case your finance team will accept. It does not assume upper-funnel spend always works. Some awareness campaigns fail, and good measurement will show that too.

Key takeaways

• Click-based attribution undervalues video, display and connected TV because people rarely click awareness ads. They buy weeks later through search or a direct visit, and those channels take the credit.
• Track upper-funnel metrics in layers: exposure (reach, frequency), attention (viewability, completed views), demand (branded search, direct traffic, share of search) and business outcome (new customers, incremental revenue).
• No single method is enough. Use attribution for weekly optimization, geo holdouts or lift studies to prove cause and effect, and marketing mix modeling for annual budget decisions.
• One well-designed geo test usually does more to protect an awareness budget than months of dashboards.
• Small budgets can still get a read from on and off tests, a “How did you hear about us?” field and branded queries in Google Search Console.
• Match lookback windows to your real buying cycle, and agree on the decision rule before any test starts.

What Is Upper-Funnel Marketing, and Why Is It Hard to Measure?

Upper-funnel marketing, also called top-of-funnel (TOFU) marketing, is advertising aimed at people who are not shopping for a product yet: video and connected TV ads, programmatic display, paid social prospecting, podcasts, sponsorships and out-of-home. The job of these brand awareness campaigns is to make a brand known and remembered, so it is on the shortlist when a need appears. Lower-funnel activity, such as branded search and retargeting, captures demand that already exists.

That job is hard to measure for a simple reason: the effect shows up later, somewhere else. Most buyers are not in the market when they see an ad. Research by John Dawes of the Ehrenberg-Bass Institute for LinkedIn’s B2B Institute estimated that only about 5% of business buyers are in the market for a given product in any quarter, because most companies switch suppliers only every few years. The finding is now known as the 95:5 rule. An awareness campaign speaks mostly to the other 95%, who may not act for months.

Privacy changes make it harder still. Browser limits on third-party cookies, Apple’s App Tracking Transparency and the walled gardens of Meta, TikTok and YouTube mean that much of the exposure data a marketer would need never reaches their analytics tool.

Why Click-Based Reporting Undervalues Awareness

Four biases work against upper-funnel channels in standard reporting.

The click bias. Most attribution and analytics setups record clicks and sessions. People rarely click a pre-roll video or a display banner, but many of them remember the brand. When they later search for it or type the URL directly, the credit goes to search or direct traffic. The channel that created the demand gets none. The gap keeps growing as more discovery happens inside AI answers that send no click at all, a shift we cover in our guide to generative engine optimization for startups.

The time-lag bias. Awareness takes time to turn into purchases. Someone who sees a campaign today might buy in three or six weeks. Short attribution windows and weekly performance reviews cut that link off, making upper-funnel spend look like cost without return.

The last-touch bias. Lower-funnel channels, especially retargeting and branded search, sit right before the purchase. They collect the conversion, even when the buyer had already decided. Their efficiency looks outstanding, partly because they are harvesting demand that upper-funnel campaigns created.

The platform bias. Each ad platform reports its own view-through conversions using its own rules. Add up what Meta, YouTube and your programmatic platform each claim, and the total often exceeds the sales you actually made, because several platforms take credit for the same customer. Teams that notice this start distrusting every upper-funnel number, which pushes budget even further down the funnel.

Put together, these biases are the core problem with upper-funnel attribution: the further a channel sits from the purchase, the worse it looks in the report, regardless of its actual contribution.

The Cost of Getting It Wrong

When reporting systematically favors the bottom of the funnel, budget follows. Every quarter, a little more money moves into search and retargeting because their numbers look better. For a while nothing seems to break. Then the pool of people who already know the brand starts to shrink, branded search volume flattens, retargeting audiences get smaller, and cost per acquisition rises across the board. By the time it shows up, the cause is months in the past and the connection is hard to prove.

Les Binet and Peter Field documented this pattern in The Long and the Short of It, their 2013 effectiveness study for the IPA (Institute of Practitioners in Advertising). Sales activation produces quick, visible spikes, while brand building grows sales more slowly but for much longer. Their work is the source of the widely quoted 60:40 split between brand building and activation for consumer brands, with a smaller brand share for most B2B companies. The exact ratio matters less than the lesson: a budget judged only on next week’s conversions will keep cutting the activity that fills next year’s pipeline.

Upper-Funnel Metrics That Actually Signal Impact

Not every awareness metric deserves a place in a report. A useful way to sort them is by the question each one answers: did we reach the right people, did they pay attention, did they remember us, did demand grow, and did the business grow? Track at least one metric from each layer, and treat the first two layers as checks rather than results.

Layer	Metric	What it tells you	Where it misleads
Exposure	Reach and frequency	How many people saw the campaign, and how often	High reach in the wrong audience is waste, and very high frequency annoys more than it builds memory
Attention	Viewability and video completion rate	Whether the ad had a real chance to be seen	Autoplay, muted and small placements can inflate it
Memory	Brand lift (ad recall, awareness, consideration)	Whether exposed people remember the brand better than an unexposed control group	Needs enough reach to be reliable, and recall does not always turn into purchase
Demand	Branded search volume and share of search	Whether more people are actively looking for you	PR, seasonality and competitor news move it too, so compare against a baseline
Demand	Direct traffic and new-visitor share	Whether new people are arriving on their own	Direct traffic also collects untagged links and app traffic
Platform	View-through and engaged-view conversions	A rough signal of who converted after seeing an ad	Each platform counts generously, and the platforms overlap
Business	New customers and incremental revenue	Whether the campaign actually grew the business	Only trustworthy when measured against a control group or a model

Share of search deserves a special mention. At the IPA’s EffWorks Global conference in 2020, Les Binet showed that a brand’s share of branded search volume in its category tracked its market share in categories such as cars, energy and mobile phones. Changes in search share were often followed by similar changes in market share, with a lag of up to a year for cars. It is not a perfect predictor, but it is one of the cheapest leading indicators available. You can build it from Google Trends or keyword tool data for your brand and your three to five closest competitors.

How to Measure What Upper-Funnel Campaigns Do

No single tool can see the whole effect of an awareness campaign. The six approaches below each cover a different blind spot, and they are strongest in combination. Start with the ones your data and budget allow, then add the others as spend grows.

Collect Impression Data Where You Can

For display and programmatic campaigns, impression-level data can be collected through the buying platforms and connected to later site visits and conversions, without relying on third-party cookies. This brings exposure into the customer journey instead of leaving it outside.

It only works if the rest of your tracking is clean. Consistent UTM naming, key events in Google Analytics 4 and lead source fields that travel into your CRM are what let an impression be connected to a sale weeks later. If a recent site change broke any of that, our checklist on keeping lead source attribution intact after a website redesign shows what to verify.

Model the Exposure You Can’t Observe

Walled gardens like Meta, TikTok and YouTube don’t share user-level impression data, so direct collection isn’t possible. The alternative is to estimate which journeys were likely exposed, based on campaign reach, targeting and timing, and add those modeled touchpoints to the journey. This approach to measuring the impact of upper-funnel campaigns lets awareness spend be judged alongside clicks within the same attribution model, rather than through each platform’s own self-reported view-through numbers.

Modeled touchpoints are estimates, so treat them that way. Ask how the model assigns exposure, how often it is recalibrated, and whether its results line up with a geo test or lift study. A model that has never been checked against an experiment is a hypothesis, not a measurement.

Editor’s note: Visualmodo has no commercial relationship with Roivenue. We link to its upper-funnel tracking page as one example of the modeled-exposure approach described above, and any such model should be validated against an experiment before it drives budget decisions.

Watch the Leading Indicators

Branded search volume, direct traffic and new-visitor share often react to upper-funnel activity before sales do. Tracking them week by week against campaign flighting gives an early signal of whether awareness spend is working.

Put flight dates and spend on the same chart as the indicators, and record at least eight weeks of baseline before a new campaign starts. Google Search Console gives you branded query impressions and clicks for free: filter for queries that contain your brand name and export them weekly. A lift that begins after a flight starts and fades some weeks after it ends is a pattern worth showing your finance team. A lift that appears everywhere at once is more likely seasonality or press coverage.

Test It With Geo Holdouts and Lift Studies

Geo holdouts are particularly well suited to upper-funnel channels. Run the campaign in some regions and not in comparable others, then compare sales, branded search and new customers over a long enough period to capture the delay. Google’s researchers describe the method in their paper on measuring ad effectiveness using geo experiments, where regions are randomly assigned to test and control groups so hidden differences between markets do not skew the result. A single well-designed test often does more to secure future budget than months of dashboards.

A geo holdout test is only as good as its design:

• Match the regions. Group cities or states with similar sales history and seasonality, then randomly assign them to test and control.
• Hold everything else steady. Keep search, retargeting and promotions the same in both groups, so the awareness campaign is the only difference.
• Run long enough. Most teams plan at least four to six weeks of activity plus a cool-down period, because the delayed effect is exactly what the test is meant to catch.
• Pick the success metric in advance. New customers, branded search or total revenue, chosen before launch, not whichever number looks best afterwards.

Ad platforms offer their own experiments too. Google Ads Brand Lift and Meta’s Brand Lift and Conversion Lift studies split audiences into exposed and holdout groups, then survey or measure both. A brand lift study is easier to run than a geo test, but it measures the platform’s own campaigns with the platform’s own data, so use it to compare creatives and audiences rather than to settle the budget question.

Geo tests also work where nothing can be clicked at all. Billboards, local events, print and storefront activity, the kinds of tactics covered in our guide to promoting your brand offline, can only be measured this way or through a model.

Use Marketing Mix Modeling for the Big Picture

Marketing mix modeling (MMM) works from the top down. Instead of following individual users, it uses weekly spend, sales and outside factors such as seasonality, pricing and promotions to estimate how much each channel contributed over time. Because it needs no cookies or user-level data, it is one of the few methods that treats Meta, YouTube, connected TV and offline media on equal terms.

Open-source tools have made MMM far more accessible. Google opened Meridian to all advertisers in January 2025, and Meta maintains Robyn. Both are free, and Meridian can take results from past experiments as inputs, which anchors the model to tested reality. The trade-offs are real, though. A reliable model typically needs two or more years of weekly data with enough variation in spend, someone comfortable with statistics, and patience: MMM answers quarterly and annual budget questions, not which ad to pause on Tuesday.

Extend the Lookback Window

Whatever method you use, give upper-funnel channels a lookback window that matches your real consideration cycle. Judging a video campaign on seven-day results is like judging a new store on its first morning.

In Google Analytics 4, the key event lookback window lives under Admin, Data display, Attribution settings. Acquisition key events allow 7 or 30 days, and all other key events allow 30, 60 or 90 days, with 90 as the default. Ad platforms set their own click-through and view-through windows separately, so check that the numbers you compare were counted over the same period. If your sales cycle runs longer than 90 days, no attribution window will capture it, which is one more reason to lean on experiments and MMM.

Attribution vs. Incrementality Testing vs. MMM: Which Method Answers Which Question

Each method answers a different question, at a different speed and cost. Google’s modern measurement playbook combines all three: attribution for day-to-day optimization, incrementality experiments to prove cause and effect, and marketing mix modeling for budget allocation, with each one checking the others.

Method	Best for	Data needed	Speed	Main weakness
Platform-reported view-through conversions	Quick in-platform optimization	Platform data only	Real time	Platforms overlap and grade their own work
Multi-touch attribution with impression or modeled exposure data	Comparing channels and campaigns week to week	Clean tracking plus impression or modeled exposure data	Daily to weekly	Shows correlation, not cause and effect
Brand lift or conversion lift study	Testing creative, audiences and messages	Enough reach on one platform	Two to six weeks	Limited to one platform’s campaigns
Geo holdout test	Proving whether a channel causes growth	Sales or leads by region	Six to twelve weeks, including cool-down	Needs enough comparable regions and discipline
Marketing mix modeling	Splitting the annual budget across all channels	Two or more years of weekly spend and sales	Quarterly	Slow to react and sensitive to modeling choices
Leading indicators	Early warning between bigger reads	Search Console, analytics, Google Trends	Weekly	Also moves with PR and seasonality

For most teams the practical order is simple. Fix tracking and watch leading indicators first, run one geo or lift test on the largest awareness channel, and add MMM once spend across channels is large enough that a few percentage points of reallocation justify the effort.

Worked Example: Reading a Geo Holdout Test

The figures below are illustrative, but the way to read them applies to any test. An online retailer runs YouTube and programmatic display in eight test regions and holds out eight matched control regions for six weeks, then keeps measuring through a four-week cool-down. Media spend in the test regions is $10,000. Each change is measured against the same period before the campaign.

Metric	Test regions	Control regions	What it means
Branded search volume	+18%	+2%	About 16 points of extra branded demand
Direct traffic	+9%	+1%	More people arriving on their own
New customers	1,000 to 1,070 (+7%)	+1%	About 60 extra new customers
Revenue	$400,000 to $416,000 (+4%)	+0.5%	About $14,000 in extra revenue
Conversions credited to video and display by last-click attribution	3	Not applicable	Why the campaign looked like a failure in standard reports

Read it from the bottom up. Last-click attribution credited the campaign with three conversions, which would have been enough to cancel it. Against the control regions, though, the campaign produced roughly 60 additional new customers and $14,000 in additional revenue, or $1.40 in revenue for every dollar of media. At a 50% gross margin that is $7,000 in gross profit, so the campaign has not paid for itself within the test window.

This is where the new-customer number decides the case. The incremental cost per new customer is about $167. If a new customer brings in $250 in gross profit over their first year, those 60 customers are worth about $15,000, and the campaign pays back within twelve months. If they are worth $100, it does not, and the honest answer is to change the creative or targeting before spending more. One more caveat: a single test can be distorted by a local event, so repeat it before making a large budget change.

Measuring Upper-Funnel Impact on a Small Budget

Geo tests and marketing mix models assume regional media buying and years of data. Many small businesses have neither, but they can still get a useful read on whether awareness spend is working.

• Run an on and off test. Run the campaign for four weeks, pause it for four weeks, then run it again. Compare branded search and new customers in each period against the same weeks last year. It is weaker evidence than a geo test because seasonality and news can interfere, but repeated cycles that move in the same direction are hard to ignore.
• Keep one holdout region if you can. Even a single comparable city or state left without ads gives you a baseline that a simple before-and-after comparison cannot.
• Ask new customers how they found you. A required “How did you hear about us?” field on forms and checkouts catches the YouTube video or podcast mention that analytics never sees. Treat the answers as directional, because people misremember.
• Lean on free tools. Google Search Console shows branded queries, Google Trends shows branded interest over time, and GA4 tracks new-visitor share. On a WordPress site, Google’s Site Kit plugin connects Analytics and Search Console to your dashboard without editing theme files.

If analytics is not installed yet, start there. Our post on why every website needs analytics from day one explains what you lose by waiting, and every week without a baseline is a week you cannot compare against later.

How to Start Measuring Upper-Funnel Impact in 30 Days

1. Name the outcome the campaign should move, such as new customers, branded demand or qualified leads, and write it down before launch.
2. Pull at least eight weeks of baseline data for branded search, direct traffic, new-visitor share and new customers.
3. Audit tracking: UTM naming, GA4 key events, CRM source fields and lookback windows that match your sales cycle.
4. Choose one experiment for your largest awareness channel: a geo holdout if you sell across enough regions, a platform lift study or an on and off test if you do not.
5. Agree on the decision rule in advance, for example: keep the budget if the incremental cost per new customer stays below $200.
6. Report it all on one page: spend and flight dates, leading indicators, the test result and the decision taken.

Common Upper-Funnel Measurement Mistakes

• Judging awareness campaigns on seven-day ROAS. The delay is the whole point of the channel, so a short window will always make it look like a loss.
• Treating view-through conversions as proof. They show who converted after seeing an ad, not who converted because of it.
• Testing during a big promotion or launch. Large sales events swamp the signal and make test and control groups look alike.
• Changing several things at once. If the creative, the audience and the budget all change in the same month, no method can tell you which change worked.
• Stopping a test early. Flat results in the first two weeks are normal, because most of the effect arrives later.
• Reporting reach and completed views as results. They confirm the campaign ran properly. They do not show that it grew the business.

Making the Case Internally

The goal isn’t to argue that upper-funnel spend always works. Some campaigns don’t, and good measurement will show that too. The goal is to judge every channel on the same evidence. When impressions and clicks sit in one model, and when experiments back up the attribution, budget debates become about which campaigns earn their place rather than which part of the funnel has the most flattering report.

Speak the finance team’s language when you present the results. Incremental cost per new customer, payback period and incremental revenue carry more weight than reach or completed views. A decision rule and guardrail metrics agreed before the test also stop the meeting from turning into an argument about which numbers to trust. Our guide to data-driven decision making shows how to set a baseline and guardrails before a test, which is exactly the discipline an upper-funnel test needs.

That shift tends to change behavior quickly. Teams stop defending channels and start optimizing them, including the awareness campaigns that were previously first in line for cuts.

How This Guide Was Researched

This guide was last checked in October 2026. Sources: Jon Vaver and Jim Koehler, Measuring Ad Effectiveness Using Geo Experiments (Google, 2011); Google Analytics documentation on attribution settings and lookback windows; Google’s modern measurement playbook; Les Binet and Peter Field, The Long and the Short of It (IPA, 2013); Les Binet’s share of search research presented at IPA EffWorks Global (2020); and John Dawes’s 95:5 research for the LinkedIn B2B Institute (Ehrenberg-Bass Institute). The worked example uses illustrative figures, not client data.

Upper-Funnel Measurement Questions: Metrics, Lift Tests, MMM and Lookback Windows