It's 11pm on a Tuesday. You've just provisioned your first tracking numbers, wired up a basic IVR, and turned on traffic from a network you found last week. Three hours later you've spent $340 on calls that went to voicemail because you forgot your buyer closes at 6pm Pacific.
That was me in 2022. And honestly, most operators have a version of this story.
The first week of a pay-per-call campaign is where budgets die quiet deaths. Not because the channel doesn't work — it works extremely well when set up right — but because there's a specific sequence of setup tasks that, when skipped or rushed, burn real money before you've learned anything useful.
This is the checklist we'd hand someone starting their first campaign today. Day by day, in order, with the mistakes we've watched operators make (and made ourselves).
Day 1: Account Setup and Carrier Selection
You can't route calls without numbers. You can't get numbers without a carrier. Start here.
Pick your carrier stack. VeloCalls supports Twilio, Telnyx, and Bandwidth out of the box. If you're on Ringba or CallRail, you're probably defaulting to their in-house telephony. The tradeoff: in-house is simpler, BYOC (bring your own carrier) is cheaper at scale. My take? For a first campaign under 5,000 minutes/month, simplicity wins. Don't optimize carrier costs until you have volume worth optimizing. Operators who start by negotiating 0.3¢/min rate differences before they've proven the unit economics work are solving the wrong problem.
Provision your tracking numbers. You need at least one number per traffic source you're testing. Testing three publishers? Three numbers minimum. The mistake here: provisioning one number and thinking you'll sort out attribution later. You won't. Attribution is table stakes — if you can't see which source sent which call, you can't cut the bad ones.
Set up your account hierarchy. This varies by platform, but you want: one campaign per vertical, one offer per buyer relationship, one number pool per traffic source. We've onboarded operators who dumped everything into a single campaign because "it's just a test" — then spent two weeks untangling the data when they wanted to scale the one source that actually worked.
Configure your payment method and billing alerts. Set a daily spend cap. Set an alert at 50% of that cap. The number of operators who've blown through a week's budget in 12 hours because they had no alerts is embarrassing. I'm including myself in that count — once, in 2023, on an insurance campaign that looked healthy until it wasn't. Woke up to a $1,200 spend notification. That was supposed to be my monthly test budget.
Day 2: IVR and Routing Configuration
This is where most first-time operators underinvest. They get the calls to ring. They don't get them to convert.
Build your IVR flow. At minimum, your IVR needs: a greeting, a service-type prompt (if you're buying for multiple verticals or job types), and a routing decision based on the response. Keep it short — every additional IVR menu step loses 8-12% of callers. Two prompts max before you connect to a human.
Set up geographic routing. Caller's area code and IP geolocation should both match the buyer's service area. About 8-12% of calls in the industry come from outside the buyer's actual coverage — if you route them anyway, you're paying for hangups.
Configure time-of-day routing. This is the one that got me on that Tuesday night in 2022. Build a schedule that matches each buyer's actual availability. Include holidays. Include lunch hours if your buyer's a small shop that doesn't answer noon to 1pm.
Set qualification thresholds. Most networks pay on a 90-second or 120-second minimum duration. Set your own internal threshold slightly higher — 100 seconds for a 90-second network payout, for example. The buffer catches calls that technically qualify but are clearly price shoppers who won't convert. We've seen this move effective conversion rates 3-5 points in the first month.
Test the flow yourself. Call every number from a real mobile phone. Walk through the IVR. Verify the whisper message plays. Verify the recording captures. If anything feels clunky to you, it will feel worse to a caller who's actually stressed about their burst pipe. Fix it before you turn on traffic. For deeper routing setup, we covered this in our guide on why pay-per-call campaigns bleed money. Our IVR abandonment rate study shows every additional menu level costs 6% of callers — keep it tight.
Day 3: TCPA Compliance and Consent Records
Skipping this step is how operators wake up to five-figure legal bills. The FCC's 2024 one-to-one consent rule is enforced in 2026. This is not optional.
Request consent records from every traffic source. If you're buying calls from a publisher running a lead-gen form or comparison site, you need the consent record — with timestamp, IP, and the exact disclosure language — before you pay for anything. We've audited operators paying six figures a month who had never seen their consent records. That's TCPA exposure sitting in a drawer. (I'll be honest: I ran campaigns for eight months in 2021 before I even asked about consent records. Got lucky. Don't be me.)
Verify the consent language names you specifically. Under the 2024 rule, blanket "we and our partners may contact you" disclosures don't count. The consent has to name your business specifically, or the consent isn't legally yours. This sounds paranoid until you see the settlement numbers — home services TCPA class actions have hit $5-12M. For a deeper breakdown, read our TCPA one-to-one consent guide.
Set up recording disclosures for two-party states. If you're routing calls to or from California, Florida, Illinois, or about a dozen other states, you need explicit recording disclosure at call start. Most publisher whisper messages don't qualify. Build it into your IVR. One California judgment for a single recorded call: $40K.
Document everything. When (not if) you get a demand letter or litigation threat, the first thing your attorney will ask for is consent records and recording disclosures. If you can't produce them in 24 hours, your defense gets a lot more expensive. We've covered the specifics in our home services pay-per-call playbook.
Day 4: Test Traffic and Fraud Filters
You've built the routing. Now you stress-test it before real money flows through.
Run 20-50 test calls across your number pool. Use different devices, different numbers, different times of day. Walk through the full flow. Verify that geographic routing works — if you have access to numbers in different area codes, test whether out-of-area calls get rejected or flagged properly.
Set up basic fraud filters. At minimum: duplicate number detection (flag any number that calls more than twice in 24 hours), VoIP detection (flag calls originating from known VoIP ranges), and duration pattern analysis (flag calls that consistently end within 5 seconds of the qualification threshold). These won't catch everything, but they catch the obvious stuff that bleeds budgets in week one.
Configure call velocity alerts. Any single number generating more than 5 calls in a week should trigger a manual review. Legitimate callers don't call that often. Fraud rings do.
Verify your analytics wiring. Calls should flow into your reporting with source attribution, duration, recording URL, and disposition. If any of those fields are missing, you can't optimize — you're just guessing. We wire call data into JustAnalytics to see which sources drive qualifying calls vs. which drive garbage. The number of "premium traffic sources" that turn out to be 60% recycled leads is frustrating. But you won't know until the data's actually flowing.
Day 5: First Live Traffic (Small Budget)
Now you turn on real spend. But small.
Start with $50-100/day per traffic source. The goal is not volume — the goal is data. You're learning whether this source sends callers who convert, not trying to scale yet. If a source can't prove itself on $50/day, it won't prove itself on $500/day either.
Monitor in real time for the first few hours. Watch the call volume dashboard. Watch the after-hours routing. Watch the qualification rate. If something looks off — way too many short calls, weird geographic distribution, calls at 3am when your buyer is closed — pause traffic and investigate before more money walks out the door.
Pull 10 call recordings manually. Don't wait until day 7. Listen to 10 calls. Are they real service inquiries? Is the caller in the right area? Does the conversation sound natural, or is it a bot playing background noise? Ten recordings will tell you more than a week of dashboard stats.
Document anomalies. Write down anything that feels wrong — even if you can't prove it yet. "Caller 7 sounded scripted" or "4 calls from the same area code in 20 minutes" might be nothing, might be a pattern. You'll want those notes when you're deciding whether to dispute or scale.
Day 6: Recording Review and Quality Assessment
This is the day most new operators skip.
Don't.
Pull a full sample of yesterday's calls. If you ran 50 calls, listen to at least 15. If you ran 100, listen to 25. Yes, this takes time. It also tells you whether you're paying for real demand or paying for noise.
Score each call on three dimensions. First: intent. Was the caller looking for service, or just shopping quotes for a project they haven't decided to do? Second: service area match. Was the caller actually in the buyer's coverage zone? Third: outcome. Did the buyer schedule an appointment, give a quote, or just take a message?
Flag disputeable calls. If a call is clearly out of service area, under duration threshold, or obviously fraudulent (robotic speech, background loops, caller says they've already called before), flag it now. Most networks have a 7-14 day dispute window. Don't let legitimate disputes age out because you didn't review in time.
Identify routing improvements. Maybe callers asking about "gas line repair" are getting routed to a plumber who doesn't do gas work. Maybe 6pm calls are hitting voicemail when a West Coast buyer is still open. These are fixable. Fix them before day 7.
Day 7: Dispute, Adjust, and Go Live
You've tested. You've listened. Now you clean up and decide whether to scale or cut.
Submit disputes within the window. Most networks require specific timestamps, call recordings, and a reason code. Don't dispute "I didn't like the quality" — dispute "call from area code 602 at 14:32:07 on 7/14, caller explicitly stated they were located in Nevada, outside buyer's Arizona service area, attached recording at timestamp 0:47." The more specific, the higher your approval rate.
Adjust routing based on day 6 findings. Time-of-day tweaks, skill-based routing changes, tighter geographic filters — implement them before you scale spend. Every day you run traffic on broken routing is money you're donating to your network.
Calculate your effective margin. Payout per qualified call, minus cost per call, times qualification rate. If the margin is below $5 per call, either fix the routing or kill the source. We've written more on this calculation here.
Decide: scale, hold, or cut. Positive margin and real intent on recordings? Scale 2x and watch for another week. Marginal? Hold and keep optimizing. Negative or garbage quality? Cut it. Move on. Don't chase sunk cost. Most operators — myself included — hold onto bad sources way too long because "maybe next week will be better." It won't. The data's usually right by day 7.
Honorable Mentions
Call scoring beyond duration. Some platforms offer AI-powered intent scoring that goes beyond just "did the call last 90 seconds." VeloCalls has AI conversation intelligence — transcription, sentiment analysis, summarization — that flags low-intent calls before they qualify. On Ringba or CallRail, you'll build this logic yourself or sample manually. Worth exploring in week 2-3 once you have baseline data. We wrote a full breakdown on when AI voice qualification cuts CPL vs. adds it.
Multi-carrier failover. If your primary carrier has an outage (Twilio has had a few in the past two years), calls fail. Configuring failover to a secondary carrier — Telnyx and Bandwidth are the common backups — means calls still connect even when infrastructure hiccups. Not a day-one priority, but good insurance once you're past $10K/month in spend.
Webhook integrations to CRM. Getting call data into your buyer's CRM automatically (source, duration, recording URL, disposition) closes the attribution loop. Service Titan and Housecall Pro both accept webhook payloads — though I'll be blunt, getting Service Titan webhooks to fire reliably took our team three attempts. The agencies that win are the ones who know, within 24 hours, whether a call turned into a closed job.
Quick Verdict
If you only do one thing from this checklist: review recordings before you pay any invoice. Every new operator I've watched blow budget in week one skipped this step. They trusted duration metrics. They trusted network QA scores. They didn't listen to what the callers actually said.
Duration tells you how long someone stayed on the line. Recordings tell you whether they wanted the service. Those are different questions. The second one determines whether you make money.
The first week is not about volume. It's about building the system that will let you scale without bleeding money. Get the routing right. Get the compliance right. Get the fraud filters in place. Then scale.
For the infrastructure — smart call routing, real-time bidding, visual IVR builder, AI conversation intelligence, and TCPA compliance built-in — VeloCalls is what we built for exactly this use case. AI sales agents are on the roadmap (coming soon per the site, not shipping today). But honestly, this checklist applies regardless of platform. Pick your tools. Run the checklist. Don't skip the recordings.
And if you're also running paid search traffic into these campaigns, ClickzProtect handles click-fraud detection on that side — same waste economy, different surface.
Frequently Asked Questions
How long does it take to set up a pay-per-call campaign from scratch?
Plan on 5-7 business days for a basic campaign. Day 1-2 is account setup, number provisioning, and carrier config. Day 3-4 covers IVR flows and routing rules. Day 5-6 is test traffic. Day 7 you review recordings, fix what's broken, and go live with real spend. Operators who try to compress this into 48 hours almost always miss something — usually TCPA consent records or after-hours routing.
What's the most common mistake in week one of a pay-per-call campaign?
Paying for calls before reviewing recordings. New buyers turn on traffic, see qualified calls hitting duration thresholds, and assume everything is working. Then they pull recordings at day 7 and find 20-30% are garbage — wrong service area, price shoppers, or duration-stuffed fraud. Sample 10 calls manually before you pay any invoice.
How do I know if my tracking numbers are set up correctly?
Call each number from a real mobile phone (not VoIP) and walk through the IVR. Does it route to the right buyer queue? Does the whisper message play? Does the recording capture clearly? Check that geographic routing matches by calling from different area codes if possible. If any of this fails during your test, it will fail on live traffic.
When should I dispute a call with a network or publisher?
Dispute within the network's window — usually 7-14 days from the call date. Flag calls that are under duration threshold, clearly out of service area, obviously fraudulent (robotic speech, background noise loops), or where the caller explicitly says they've already called before. Pull the recording before you dispute. Networks reject disputes that don't include specific timestamps and reasoning.
Try VeloCalls for Your Vertical
AI calling + pay-per-call platform built for HVAC, plumbing, roofing, PI lawyers, Medicare brokers, and insurance. Smart routing, real-time bidding, visual IVR builder, AI conversation intelligence. Per-minute pricing — Managed starts at 4¢/min, BYOC at 2¢/min, both drop as you scale.