Marketing Measurement ·

How can I estimate incremental reach and lift from an upfront CTV buy?

Learn how to estimate incremental reach and sales lift from an upfront CTV buy, why clicks and many MMMs undercount CTV, and how to forecast before you commit.

Linnea Zielinski · 12 min read

Listen
0:00 / 0:00
AI-generated audio
How can I estimate incremental reach and lift from an upfront CTV buy?

An author books a 20-city book tour a year before her novel comes out. The publisher pays for flights, hotels, and bookstore events months in advance. On tour, she counts the copies she signs at each stop, and the numbers look modest. But the tour's real impact rarely shows up at the signing table. It shows up weeks later, when people who heard her on local radio buy the book on Amazon, when a reader who came to one event tells her book club, and when indie bookstores in cities she never visited start reordering. If the publisher judged the tour by signed copies alone, they'd never book another one.

Upfront connected TV (CTV) buys put marketers in the same position. You commit a big budget months ahead to lock in premium streaming inventory, and then most of what that inventory does happens out of view: no click, a purchase days or weeks later, and often in a different place than the ad ran, like branded search, Amazon, or a retail store. With CTV ad spend forecast to grow 13.8% in 2026, according to the IAB, more of the budget is riding on this question every year. Brands that can estimate what an upfront actually added can commit with confidence and negotiate on value, while brands that can't end up either overpaying for impressions or pulling back from a channel that's filling their funnel.

Key takeaways

  • Incremental reach and incremental sales lift are separate questions, since reaching new households doesn't guarantee new sales and extra frequency on households you already reach can still drive revenue.
  • Reach and overlap estimates come from cross-media measurement, panels, and TV viewing data, while sales lift needs a model or a test that looks at your actual revenue.
  • Click-based attribution undercounts CTV because most viewers never click, and the sales CTV drives often land later and in other channels, like branded search, direct traffic, and Amazon.
  • Many traditional and open-source MMMs also undercount CTV because they treat it like a lower-funnel channel, apply short decay patterns, and let branded search take credit for the demand CTV created.
  • Prescient's marketing mix model measures CTV as a demand-creating campaign, including its halo effects, delays that vary by campaign, and response shapes that aren't forced to flatten during big flights.
  • Geo tests and holdouts can add a useful data point, but they describe one window and one spend level, so they shouldn't be the only basis for a full-year commitment.
  • Before signing, forecast both an added-budget scenario and a budget-neutral one, translate expected lift into profit, and size the commitment to how certain you are.

How CTV upfronts work

Upfronts are advance commitments where advertisers negotiate with media companies to buy inventory for an upcoming period, often several months to a year out. They started with broadcast networks selling the fall TV season, but today they cover streaming services, CTV platforms, and cross-platform packages that bundle linear and streaming inventory.

A typical upfront deal can include:

  • Guaranteed impressions or audience delivery across a set flighting window.
  • Negotiated CPMs, often lower in exchange for larger commitments.
  • Access to premium content, like live sports or high-demand programming.
  • Audience targeting guarantees for specific demographics or segments.
  • Cancellation options and makegoods, which give some flexibility if delivery falls short or plans change.

The tradeoff is that you're committing before you know how performance, audiences, or your own business will shift. That's why the measurement question matters so much more for upfronts than for buys you can adjust week to week.

What upfronts offerThe catch
Lower negotiated CPMsA lower CPM doesn't mean lower cost to acquire a customer
Guaranteed impressionsGuaranteed impressions aren't guaranteed new households
Premium inventoryPremium audiences may overlap heavily with your other buys
Predictable spendingLess flexibility to move money if results disappoint

Incremental reach and incremental sales lift are different questions

The phrase "incremental lift" often gets used to cover two very different things. Separating them is the first step to measuring an upfront well, because each one needs a different kind of measurement.

Incremental reachIncremental sales lift
What it asksHow many new households did this buy reach that the rest of the plan didn't?How much revenue did this buy add that wouldn't have happened without it?
What it measuresAudienceBusiness outcomes
Where the answer comes fromCross-media reach measurement, panels, and TV viewing dataA marketing mix model or a well-designed test
Why it mattersShows whether you're expanding your audience or paying to repeat itShows whether the buy is worth the money

Here's a simple example of why they diverge. Say a $1 million streaming buy reaches 8 million households, but 6 million of them already see your ads on linear TV, YouTube, or social video. The buy's incremental reach is just 2 million households.

That doesn't tell you its sales lift, though. Some of those 2 million new households may never be interested in your product. And some of the 6 million overlapping households may become more likely to buy after seeing you on a big screen. Extra frequency on an audience you already reach can still be valuable, as long as it isn't past the point where more exposure stops helping.

Compare buys on cost per incremental household, not just CPM

CPM tells you what impressions cost. It doesn't tell you how efficiently a buy expands your audience. Dividing spend by incremental households gives you a better comparison. Here's a hypothetical example:

Publisher APublisher B
Investment$1M$1M
CPM$25$40
Households reached8M5M
New households (incremental reach)2M3M
Cost per new household$0.50$0.33

Publisher B costs more per impression but reaches new households more cheaply. That doesn't automatically make it the better buy, since audience quality and sales response still matter, but it's a good reason not to negotiate upfronts mainly on CPM discounts.

Where reach and overlap numbers come from

Incremental reach has to be measured with audience data, because it's a question about people and households. The common sources each come with limits worth knowing.

SourceWhat it showsLimitation
Automatic content recognition (ACR)Viewing data from participating smart TVsOnly covers participating devices
Audience panelsProjected reach and demographicsBased on samples, so estimates carry uncertainty
Publisher first-party dataDelivery and audience within one publisherCan't see exposure on other services
Cross-media measurement providersDeduplicated reach across participating mediaCoverage varies by provider and partner

Ask for the full frequency distribution, not just average frequency. Two campaigns can both average three exposures per household while one spreads them evenly and the other hits a small group ten or more times.

Why click-based measurement misses most of what CTV does

Once you know who you reached, the harder question is what that reach did for revenue. Pixel-based and click-based tools struggle here for a few reasons that stack on top of each other:

  • Almost no one clicks a TV ad. Viewers are on the couch, not on a device with a checkout button, so click-based tools often label CTV inefficient.
  • The purchase happens later. Someone might see an ad Monday, search your brand Thursday, and buy two weeks after that.
  • The purchase happens somewhere else. That sale might land through branded search, a direct visit, Amazon, or a retail store, each with no link back to the ad.
  • Privacy changes make the connection even harder. Browser restrictions and Apple's App Tracking Transparency limit cross-device tracking from a TV screen to a phone or laptop.

The result is that CTV's sales impact gets credited to whatever channel happened to be present at the end, usually branded search or retargeting.

Why many MMMs undercount CTV too

Marketing mix modeling avoids the click problem because it works from aggregated spend and revenue instead of individual journeys. But the structure of many traditional and open-source MMMs creates a different set of blind spots, and upfront CTV runs straight into them.

CTV gets treated like a lower-funnel channel

Many MMMs slice revenue into independent pieces, one per channel, and treat every channel the same way. That ignores the fact that CTV mainly creates demand while branded search mainly captures it. When both sit in the model as independent slices, the model has no way to see that CTV is what's filling branded search in the first place.

Branded search takes the credit

When a CTV flight lifts brand interest, branded search queries, direct visits, and Amazon orders often rise over the following weeks. A model that treats branded search as an independent driver of sales will hand much of that revenue to branded search, leaving CTV looking like it barely paid off.

Short, fixed decay cuts CTV off too early

Many MMMs apply one fixed decay pattern that assumes an ad's impact fades quickly. CTV's effects often build over several weeks of a flight and fade gradually afterward. A model that writes off most of an ad's impact within days will miss a big share of what an upfront delivered.

Big flights get labeled saturated

Upfront buys tend to concentrate spend into heavy flighting windows—big premieres, live sports, holiday stretches—rather than spreading it evenly. Models that use curves built to flatten eventually can read those high-spend weeks as proof CTV is maxed out, even when the data doesn't show it. That can lead to recommendations that cap future CTV spend too early.

One practical trap: modeling by contract date

Whatever model you use, make sure it reflects when the ads actually ran, not when the contract was signed or paid. An upfront paid in May for inventory running September through December shouldn't show up as if all that exposure happened in May.

How Prescient measures what an upfront CTV buy added

Prescient's marketing mix model was built to handle the kinds of campaigns that traditional models undercount. Instead of slicing revenue into independent pieces, our model treats demand as something marketing builds and sustains over time, which fits how CTV actually works.

Here's what that looks like for an upfront buy:

  • CTV is measured as demand creation. Our model credits the campaign that created demand, not just the one that captured it, so branded search doesn't automatically take credit for interest CTV built.
  • Halo effects are measured across destinations. CTV's impact on branded search, direct traffic, and Amazon sales shows up as that campaign's halo effects, so you can see its full contribution beyond direct conversions.
  • Delays and decay vary by campaign. A CTV campaign's effects can build and fade over weeks, and our model lets that pattern differ from a retargeting campaign's instead of applying one formula to everything.
  • Response shapes aren't forced to flatten. If CTV is still returning at a steady rate during a heavy flight, our model can show that, rather than declaring it saturated because spend was high.
  • Results are campaign level and updated daily. Separate CTV campaigns, like individual publisher buys, can be compared directly when they run as their own campaigns.

Because our model estimates each campaign's contribution, it can answer the question behind an upfront: how much revenue would you have earned without this CTV buy? That's the sales lift that matters for the next negotiation.

Testing CTV lift with experiments

Many brands want to check their model's CTV estimate against a test, especially before a big renewal. Experiments can add a useful data point, and there are a few common designs.

Test designHow it worksBest for
Geo testSome regions get the CTV buy and comparable regions don'tBrands with enough regional sales data, including Amazon and retail
Household holdoutThe publisher randomly withholds ads from a group of eligible householdsPlatforms that support true random assignment
Heavy-up testCTV spend increases in some regions while staying flat elsewhereDeciding whether to commit more
Go-dark testCTV spend pauses in some regionsEstimating what you'd lose by cutting back

Heavy-up and go-dark tests are often more useful for upfront planning, since the real question is usually whether to commit more or less, not whether TV works at all.

What tests can't tell you

A test run on one publisher, with one creative, at one spend level, during one season, describes exactly that. It doesn't tell you what a full-year commitment will do, and a strong result at $250,000 doesn't mean $2 million will return eight times as much. Geo tests also face real design problems, since no two regions are truly alike and regional events can skew results.

Prescient's view is that incrementality tests are locally accurate and globally inaccurate. That's why Prescient shows brands their model's accuracy with and without a test included. You make the final call, and if you trust a test, it can be built into the model as a starting assumption.

Estimating lift before you sign the upfront

Measurement after the fact matters, but the most valuable use of all this is before you commit. An upfront is a financial decision made in advance, so it deserves a forecast, not just a post-campaign report.

Forecast two scenarios, not one

There are two different questions hiding inside every upfront decision, and they can lead to very different answers:

  • Adding budget: What happens if we add $1 million in CTV while keeping the rest of the plan the same?
  • Moving budget: What happens if we shift $1 million from other channels into CTV?

A CTV buy can add revenue on its own and still be a poor reallocation if the channels losing budget would've added more. Our Media Forecasting feature forecasts what happens when you scale a campaign's spend or shift budget into it from other channels, using what the model has learned about your seasonality and halo effects.

Here's a hypothetical forecast comparing plans with and without an upfront:

Without the upfrontWith the upfront
Annual marketing spend$10M$11M
Annual revenue$50M$52.8M
Added revenue from the upfrontN/A$2.8M
Sales liftN/A5.6%
Return on the upfrontN/A2.8

Translate lift into profit

Added revenue isn't the same as added profit. If your contribution margin before advertising is 45%, the $2.8 million in added revenue above produces about $1.26 million in contribution, or about $260,000 after the $1 million media cost. That's positive, but much thinner than the revenue number suggests.

Bigger packages don't always pay off

Publishers often offer lower CPMs for bigger commitments. Those discounts can look great until you account for how returns change at higher spend levels. A hypothetical example at a 45% margin:

PackageSpendCPMAdded revenueProfit after media
Small$500K$30$1.8M$310K
Medium$1M$25$2.85M$282.5K
Large$1.5M$22$3.35M$7.5K

The large package has the lowest CPM and the most revenue, but almost no profit. A brand focused on growth may still choose it, but that should be a deliberate tradeoff, not a side effect of chasing a discount.

Size the commitment to your uncertainty

Every forecast comes with a range. A buy with a strong expected return can still lose money if results land at the low end. Look at downside and upside cases, not just the base case, and check how confident your model is in its CTV estimate. The less certain you are, the more valuable cancellation options and smaller initial commitments become.

Why your CTV numbers won't all match

Once an upfront is running, you'll likely see several different numbers describing the same buy. That's expected, since each source measures something different. Here's a hypothetical example:

SourceWhat it reportsWhat it actually measures
Publisher delivery report40M impressionsWhat was delivered
Cross-media reach measurement2M new householdsHow much the audience expanded
Publisher attribution$5M in attributed revenuePurchases by exposed households, including people who'd have bought anyway
Geo test$2.5M in added revenueOne test design in one window
Marketing mix model$2.8M in added revenueCTV's contribution within the full mix

The question isn't which number is right. It's which one fits the decision in front of you. Reach data helps explain why performance changes, while modeled contribution tells you whether those changes matter for revenue.

Common mistakes to avoid with upfront CTV measurement

A few habits come up again and again when brands evaluate upfronts. Avoiding them can make a big difference at renewal time.

  • Treating guaranteed impressions as guaranteed new households.
  • Negotiating mainly on CPM instead of cost per new household and expected added revenue.
  • Judging CTV by clicks or direct website conversions alone.
  • Ignoring Amazon and retail sales, which can make CTV look much weaker for omnichannel brands.
  • Using a model that applies short, fixed decay to every channel.
  • Comparing CTV's modeled return with another channel's platform-reported ROAS.
  • Planning a full-year commitment from one short test.

Traditional MMM vs. open-source MMM vs. Prescient for upfront CTV

How a model is built decides how much credit an upfront CTV buy can get. Here's how the main approaches compare on the issues that matter most for CTV.

Traditional MMMOpen-source MMMPrescient's MMM
How CTV is treatedOne independent slice, like any other channelOne independent slice, like any other channelA demand-creating campaign that builds interest over time
Credit for downstream searches and salesOften goes to branded search and direct trafficOften goes to branded search and direct trafficMeasured as CTV's halo effects
Delay and decayOne fixed pattern, often shortFixed decay formulasDelays and decay that vary by campaign
Heavy flighting weeksCurves built to flatten can label CTV saturatedCurves built to flatten can label CTV saturatedFlexible shapes that follow the data
Level of detailUsually channel levelUsually channel levelCampaign level, updated daily

A plain-English guide to the technical terms

Upfront CTV measurement tends to come with a lot of jargon. This quick reference can help when talking with publishers, analysts, or leadership.

Technical termWhat it meansWhy it matters for upfronts
Demand creationCampaigns that build new interest instead of collecting existing interestKeeps CTV from losing credit to branded search
Modeling demand over timeTreating brand interest as something that builds and fadesExplains how September CTV ads still drive October sales
Varying decayLetting each campaign's impact fade at its own paceKeeps a model from writing off TV's impact after a few days
Estimating revenue without the buyComparing your actual results with what the model expects you'd have earned without CTVGives you the sales lift to bring into the next negotiation

Where Prescient comes in

Prescient's marketing mix model measures upfront CTV the way it actually works: as a campaign that creates demand, pays off over weeks, and drives sales through branded search, direct traffic, and Amazon. Our model credits CTV for the halo effects it creates, lets its delays and decay differ from other campaigns, and doesn't force heavy flights into a saturation curve. Results are campaign level and updated daily.

That gives omnichannel brands a clear estimate of what each upfront added, plus our Media Forecasting feature to compare adding budget against shifting it before you sign. If you want to see what your CTV commitments are really driving, book a demo.