The CTV and Streaming Direct-Response Playbook for Insurance and HVAC Leads

Quick answers: Why streaming is direct response now · Concepts and lengths · QR and the CTA · Produce it for a TV · Measure it and do the math · The first test · What it will not do

Key numbers, as of September 2026:

• Per Nielsen, streaming reached 48.6% of US TV watch-time in May 2026, more than broadcast and cable combined (Nielsen The Gauge).
• Per Nielsen, ad-supported viewing is about 73% of all US TV, and streaming is a record 46.6% of that ad-supported time (Nielsen Q1 2026 Ad-Supported Gauge).
• Per Vibe, fifteen-second CTV spots drove 44% more site visits and a higher return than thirty-second spots, though about 6% fewer direct purchases (Vibe Creative Performance Report, August 2026).
• Per the IAB, US CTV ad spend is reported to grow about 13.8% in 2026 (IAB 2026 Outlook, via StreamTV Insider).
• Per NCSolutions and Nielsen, creative drives an estimated 49% of advertising’s impact on sales, more than audience, reach, and recency combined (the Vibe report).

If you buy leads and calls, streaming is worth your attention for one reason: the audience moved and the ad format finally lets you ask for a response. This is the production playbook we use when we build these spots, aimed at operators running under-65 insurance and home-services offers. It sits next to our connected-TV advertising service, and it deliberately does not re-argue whether connected TV can actually generate leads or how to track connected TV like you track search. Those are settled on their own pages. This one is about what to make and how to make it.

Why streaming is a direct-response channel now

For years, TV was a reach buy. You paid for eyeballs and hoped some of them called later. That changed for two reasons. The audience is on streaming, and the platforms added ways to respond on screen.

The audience part is done. Streaming passed broadcast and cable combined in 2026 (Nielsen), and most of that inventory is ad-supported (Nielsen). The money followed, with US CTV spend reported growing double digits into 2026 (StreamTV Insider).

Here is the constraint that shapes every decision after this. A standard streaming ad cannot be clicked. As MNTN puts it, a television commercial cannot be clicked, so it rarely shows up as the source of a conversion (MNTN). So direct response on CTV is captured a different way: a QR code, a “send to phone” prompt, a second-screen action, or a household-IP match to a later visit. The creative has to carry the response, because the remote will not.

Each platform hands you a different response mechanism, and that decides what your creative asks for. This is the landscape as each platform documents it, current as of September 2026. Read it as your menu of on-screen actions, not as a ranking.

Google adding an auto-generated QR and send-to-phone on CTV is the update that matters most for our lane, because it turns Demand Gen and Performance Max into a CTV response path they did not have before. If you already run those, this is the CTV wrinkle in running an upper-funnel channel as direct response.

Start with three concepts in two lengths

Do not make one perfect ad. Make three different arguments for the same offer, each in two lengths, and let the results tell you which one people respond to.

Your first three tests, and the styles worth testing

There is no honest public ranking that says one video style wins for every advertiser. MNTN’s analysis of beauty creatives found user-generated footage and voiceovers among the leaders, in both fifteen and thirty seconds (MNTN), but that is 2024 data in a different category. It supports testing those styles. It does not name a winner for insurance or HVAC. So test, and start with three concepts that persuade in genuinely different ways.

Six ways to make the same offer. Pick three that persuade differently and test them.

Run offer-with-voiceover, expert-spokesperson, and process-demonstration first. They are three materially different ways to earn a response, which is exactly what a first test should compare. For the raw craft of a spot that survives the first few seconds, our notes on video creative that survives the three-second hook test carry over from short-form.

Fifteen seconds or thirty

Length is a real lever, and the honest reading is more useful than the headline. Across roughly 17,000 ads, Vibe found fifteen-second spots drove 44% more site visits and a higher return than thirty-second spots, but about 6% fewer direct purchases (Vibe). That is an efficiency edge on visits, not proof that short persuades better. When Vibe held the advertiser constant, the longer cut actually converted slightly better, worth roughly 7% in conversion odds, and its own ML lead called that directional (Vibe). Some of the short-ad advantage is which businesses reach for fifteen seconds in the first place.

Vibe runs the largest public self-serve CTV creative dataset, so it anchors much of this playbook, but the length trade-off shows up in independent data too. MNTN’s separate length study points the same way: 15-second ads drove 46% more site visits, while 30-second ads drove 24% more conversions (MNTN). Two datasets, one takeaway: short wins the visit, long wins the conversion. Neither is a verdict for your account, so test it.

Word counts are starting points. Phone numbers and disclosures eat seconds, so time a real read.

The practical call: test fifteen for efficiency and thirty for anything that needs explaining, which for under-65 insurance is most of it. Word counts are starting points. A phone number, a URL, and a required disclosure all eat time, so time an actual read before you lock the edit.

The opening, and the thirty-second structure

The first three seconds decide whether the rest plays to anyone. A practical opening answers at least one question fast: is this for someone like me, is this a problem I have, is there a benefit worth my attention. The platforms agree on the fundamentals: earn attention immediately, keep branding present throughout, focus the message, and prompt the action from the start (Google; Roku).

A thirty-second direct-response spot has four jobs and one constant: the response never leaves the screen.

Keep the URL and response information on screen through the whole story. Vibe’s compliance guidance is explicit that the company URL stays on screen throughout the ad (Vibe).

Here is a thirty-second insurance spot you can hand to an editor, written the way you would type it into a storyboard:

The HVAC version runs the same discipline in fifteen seconds, on a humor concept:

When you cut one concept into many placements and lengths, the discipline is the same one behind building one ad into ten crop-resistant assets: protect the offer and the response, let the rest flex.

Capture the response: QR codes and one clear CTA

The QR code is the workhorse of CTV direct response, and most failures are boring: the code is too small, it is not on screen long enough, or the label gives no reason to scan.

Only one QR rule is hard. DENSO Wave, which developed the format, specifies a quiet zone four modules wide on every side (DENSO Wave). Everything else is a recommendation you validate on a real television at couch distance: aim for roughly 240 to 320 pixels square in a 1920 by 1080 frame including the white margin, keep it visible throughout a fifteen-second ad and for at least ten to fifteen continuous seconds in a thirty, hold it still, and put it in a dedicated area clear of disclosures and player overlays.

Build to the one hard rule, then test everything else on an actual screen.

Label the code with the benefit. “Scan for your free quote” tells the viewer why to act; “Scan me” does not. Give each creative a distinct tracking redirect so you can tell which spot drove the scan. One case worth reading correctly: Vibe reports that a customer’s QR-driven traffic converted at 20% (Vibe). That is the conversion rate of the people who already scanned, not a scan rate across everyone who saw the ad.

Pick one primary response and make it the obvious action. A QR and a spoken phone number can be two routes to the same thing, but do not give a scan, a call, a follow, and a visit equal weight on screen. For under-65 insurance and home services, the inbound call is usually the higher-intent path, so if the phone is your primary action, show the number clearly and long enough to dial, read it at a natural pace, and route to a form or scheduler when the line is not staffed.

Choose one primary response. The others can support it, never compete with it.

One platform limit to respect: standard, non-interactive streaming inventory expects “scan, call, or visit” language, not “click below” or “press OK.” Amazon explicitly restricts click-style prompts in non-interactive streaming ads (Amazon). If you want a true on-screen button, you need a supported interactive format. Whatever the response, the call-tracking behind it is what tells the platform which responses were real, which is the job of the call-tracking stack that decides which calls the platform learns from.

Produce it for a TV across the room

A CTV ad is watched from ten feet away, often with the sound low. Produce for that room, not for a phone held at arm’s length.

Compose natively in 16:9. If you are reusing vertical footage, place it inside a deliberate horizontal layout with the offer and a persistent response panel, rather than letting black bars swallow the frame. Use large, high-contrast text, roughly four to eight words per main message, on screen long enough to read. Keep dialogue clean with a close mic and music under the voice, and make the message land through both sound and picture, because half of viewing happens sound-off (Vibe). Keep the response area stable while other footage changes, so a scanner is not chasing the code around the frame.

Native 16:9. Offer and response panel where the eye lands, money elements off the edges.

Use AI where it saves time and keep it off the parts that have to be exact. It is good for spinning up concept variations, B-roll, and alternate spokesperson reads fast. It is not where you set the final text, the brand lockup, the phone number, or the QR code: assemble those by hand in the editor, so a hallucinated digit or a warped code never ships. If you test an AI spokesperson, use an authorized likeness and run it head-to-head against a real-person cut. And a testimonial has to be a real customer, not a generated one.

Now the specs. Clean Full HD is enough for a platform like Vibe, and its documented upload spec is the live one, from August 2026.

Publisher delivery specs can be stricter than a buying platform’s basic upload, and loudness is where that bites. A file that uploads fine can still be non-compliant for broadcast-grade delivery. The US standard, set by the CALM Act and ATSC A/85, targets an average loudness of -24 LKFS within about 2 LU, and delivery pipes like Extreme Reach normalize to it (Extreme Reach; FCC). Meter your mix against the spec that applies to where it is running.

Measure by qualified leads, then do the CPL math

Judge CTV by leads, calls, and customers, not by how many visits a platform matched to it. Watch three views at once, and know what each can and cannot claim.

First, direct response: QR visits, calls to a dedicated number, submitted forms, and booked appointments. This is the cleanest signal because the viewer took an action you can name. Second, attributed response: the visits and conversions a platform matches to exposed households. Vibe uses household-IP matching on a 30-day default window (Vibe). Useful, but a match after exposure is a correlation, not proof the ad caused it, as Vibe itself says (Vibe). Third, incremental impact: run a holdout or a controlled geo test and compare exposed against unexposed. That is the only view that establishes causation, and vendors concede the point, MNTN treating a 25% gap between its matched visits and analytics sessions as expected and healthy (MNTN).

Three views, one honest hierarchy: a direct action beats a match, and only a holdout proves cause.

That larger measurement discipline, and how it fits the rest of your reporting, is the whole subject of tracking connected TV the way you track search.

Now the economics, which is just arithmetic. Cost per lead is your cost per session divided by your session-to-lead rate, so the same traffic gets cheaper per lead as your landing page converts better. The chart below plots that relationship on an illustrative basis, with the numbers living in the image.

Cost per lead falls fast as the landing page converts better. The numbers are illustrative, not a benchmark.

The market context, not a rate card: CTV inventory tends to run in wide ranges, roughly a mid-teens to thirty-dollar CPM on open exchange and higher for premium direct placements (Vibe), and it varies by supply, season, and targeting. There is no defensible public cost-per-lead benchmark specific to under-65 insurance or HVAC on CTV, so do not let anyone sell you one. Use the method, watch your own qualified-call rate, and remember that cheaper visits can still buy more expensive customers.

Run a disciplined first test

Treat the first flight as a test with rules, not a launch. The rules are what let you read the result.

Produce the three concepts in both lengths. If budget is tight, launch fewer concepts and test alternate lengths once one shows promise. Hold the offer, the audience, the landing page, and the attribution settings steady while you compare, so length is the only thing changing. Give each creative its own QR tracking, and separate prospecting from remarketing so a warm audience does not flatter a cold concept.

Budget the test before you split it. Three concepts in two lengths is six cells, and every cell has to clear two bars at once: enough impressions to leave Vibe’s roughly 14-day learning period (Vibe), and enough responses that a gap between two cells reads as signal, not noise. Impressions come from spend against the CPM, and CTV runs a mid-teens to thirty-dollar open-exchange CPM (Vibe), so price out what one cell costs to run for the full two weeks before you count cells. The fall from impressions to leads is where a thin cell dies: only a fraction scan or call, and only a fraction of those convert, and a fixed budget splits every way you divide it. So set the floor per cell first, then count how many cells that floor buys. If your budget cannot power all six cells to it, cut cells, not per-cell spend. Two concepts tested properly beat six concepts each starved into noise.

Time the flight to the vertical. Demand is not flat, so do not spend flat. For under-65 health, the spike is the open-enrollment season, which the federal Marketplace currently lists as November 1, 2026 to January 15, 2027 (HealthCare.gov); because a shorter window was floated for this cycle, re-check that primary date at launch and weight the budget into it. For HVAC, follow the equipment: cooling demand peaks in summer, heating in winter, with a shoulder in October, so launch cooling creative in March or April and heating in September or October. One production note that only insurance needs: the required on-screen disclosures eat roughly three to five seconds of a thirty-second spot, so budget that read into the script before you shoot.

Three concepts, two lengths, one change at a time. The discipline is what makes the read trustworthy.

Then wait it out. With conversion lag, two weeks of thin data can still be inconclusive. Monitor lead quality from day one, not just volume, because a cheap flood of junk is worse than a trickle that closes. Once a concept looks promising, change one variable at a time: the opening, the proof, the CTA, the response method, or the length. Change two and you learn nothing. This is the same operator restraint behind buying U65 health insurance leads and calls or reading how streaming video builds HVAC demand your last-click report misses: let the data name the winner.

What a CTV playbook will not do for you

CTV is a strong top-of-funnel channel that can finally ask for a response. It is not a lead-quality machine, and a few honest limits keep expectations straight.

It will not make a weak offer convert. A polished thirty-second spot pointed at a thin quote form just gets you to the same bad result with better production. It will not replace your higher-intent channels either; CTV is lean-back household reach that tends to make your search and call channels more efficient, which is why the inbound-call path carries the intent for these verticals, and why buyers still lean on things like the pay-per-call buyer’s guide.

And it will not tell you whether the response was a real person. A QR scan or a form fill is the start of the lead, not proof of one, and a household-IP match is weaker still. Verifying that a response is a live, interested human, before you pay for it, is a separate layer, and it is the reason we verify the response with a one-time passcode before you pay. Get the offer, the response method, and the verification right, and CTV becomes a channel you can scale. Skip them and it becomes an expensive brand video.

Frequently asked questions

Can you really generate leads from a TV ad nobody can click?

Yes, but not by clicking. A standard streaming ad cannot be clicked (MNTN), so the response runs through a QR code, a spoken phone number, a send-to-phone prompt, or a household-IP match to a later visit. The platforms now support these directly, from Roku Action Ads to Google’s auto-generated CTV QR codes (Google). The creative carries the offer and the response method, and you measure the actions people take.

Should an insurance or HVAC spot be 15 or 30 seconds?

Test both, and expect them to do different jobs. Fifteen seconds is more efficient at driving visits (Vibe), which suits reminders and remarketing. Thirty seconds buys the time to explain under-65 coverage or a big repair and to set up a qualified call, and within a single advertiser the longer cut converted slightly better in Vibe’s own modeling, an effect it called directional (Vibe). For trust-heavy offers, start at thirty and test fifteen for efficiency.

How big should the QR code be and how long should it stay on screen?

Only one rule is fixed: a four-module quiet zone on every side (DENSO Wave). The rest is a recommendation to validate on a real TV. As a starting point, aim for roughly 240 to 320 pixels square in a 1920 by 1080 frame including the margin, keep it up throughout a 15-second ad and for at least 10 to 15 continuous seconds in a 30, and hold it steady. Then scan it from the couch with a few different phones before you ship.

How do you know the CTV ad actually drove the lead?

You know for certain only with a holdout. Direct actions like a scanned QR or a dedicated-number call are strong signals. Platform-attributed conversions, matched by household IP on a default 30-day window (Vibe), are a correlation, not proof, as Vibe states plainly (Vibe). To prove causation, run a holdout or a controlled geo test and compare exposed households against unexposed ones.

What does CTV advertising cost?

It varies too much for one number. CTV CPMs run in wide ranges by supply, season, and targeting, roughly the mid-teens to thirty dollars on open exchange and higher for premium direct (Vibe). There is no reliable public cost-per-lead benchmark for under-65 insurance or HVAC on CTV, so judge the channel with the CPL method above and your own qualified-call rate, not with someone else’s headline figure.

If you would rather hand the build to a team that runs streaming lead-gen for a living, from the three concepts through the QR tracking and the holdout, our connected-TV advertising service is where to start.