Answering the phone and knowing why it rang are two different jobs
There is a specific moment in a service business where these two things get confused. The owner installs an AI receptionist, the missed calls stop, the calendar starts filling, and the front office problem is genuinely solved. Then somebody asks which ad campaign is producing the work, and the honest answer is a shrug. The phone is being answered flawlessly. Nobody knows why it is ringing.
As of August 2026, Run with Jarvis treats these as two separate layers on purpose. The AI receptionist is the answering layer and it lives in Core at $500 a month. Call tracking, which is CallFlux, is the attribution layer and it switches on in Pro at $750 a month. This guide is about whether you actually need the second one, because for a meaningful share of service businesses the correct answer is no, and $250 a month is real money to leave on the table for a report nobody reads.
The short version of the decision rule is this. If you cannot name which campaign produced last month's booked jobs, and you are paying for ads, you are flying blind and attribution is the fix. If your work arrives through referrals, repeat customers, and organic search, there is nothing to attribute and you should stay on Core. Everything below is the long version, with the arithmetic.
What the answering layer already does on its own
It is worth being precise about how much is already handled before you spend another dollar, because the temptation with a feature list is to assume the higher tier does the important part. It does not. The important part is the call being answered at all.
The AI receptionist in Core answers around the clock in English and Spanish, qualifies the caller, quotes from your own price list, books the job onto the calendar, and sends the confirmation. Behind it sits the full operations layer, which is to say the CRM with customer history, the invoicing and payment links, technician dispatch with GPS and route optimization, automatic ETA and arrival texts, the point of sale, QuickBooks sync, and review automation. The platform's AI outbound follow-up calls are there too, working your own list of stalled quotes and unpaid invoices rather than cold prospects.
That is a complete front office. A business running entirely on Core has no gap in its ability to capture, book, run, and bill a job. If you have not yet worked out whether that layer belongs in your business at all, the primer on what an AI employee for service businesses actually is is the right starting point, and this article picks up after that decision is made.
What the answering layer cannot do is see backwards. Once a call is connected, the receptionist knows everything about the caller and nothing about the click. The caller does not announce that they came from a search ad, and asking them is close to worthless because people genuinely do not remember. That blind spot is not a shortcoming of the AI. It is a property of telephones.
What the attribution layer adds
Call tracking answers a different question entirely. Not who is calling and what do they need, but what did I pay to make this phone ring, and did that spend turn into revenue.
Here is the split, laid out plainly.
| Question you are asking | Answering layer, Core at $500 | Attribution layer, Pro at $750 |
|---|---|---|
| Was the call answered at 2 a.m.? | Yes, 24/7 in English and Spanish | Same |
| Was the job booked onto the calendar? | Yes, with confirmation sent | Same |
| Which ad, keyword, or channel produced this call? | Not visible | Dynamic number insertion plus gclid capture |
| Was this caller a real job or a wrong number? | Known from the booking outcome | Scored automatically from the transcript |
| What did the caller actually say and how did it go? | Outcome recorded, conversation not | Call recording, transcription, sentiment and intent |
| Which campaign should I cut this month? | Guesswork | Cost per booked job, by source |
| Does Google know which clicks became jobs? | No | Conversion upload back into Google Ads |
| Included AI call minutes and overage | 500 minutes, $0.45 per extra minute | 1,000 minutes, $0.40 per extra minute |
Read that table as two columns of one system rather than two products. Pro is cumulative, meaning it includes everything in Core and adds the right-hand behaviors. You are not choosing between answering and attribution. You are choosing whether to add attribution to answering.
Dynamic number insertion, in plain English
Dynamic number insertion is the mechanism that makes phone attribution possible at all, and it is much simpler than the acronym suggests.
Your website shows a phone number. Normally it is the same number for every visitor, which means every call arrives identical and anonymous. DNI changes the displayed number based on how the visitor got there. A visitor who arrived from a Google Ads click sees one tracked number. A visitor from organic search sees another. A visitor from a Meta campaign sees a third. Every one of those numbers forwards straight through to the same AI receptionist, so the caller experience is untouched, and the number they dialed becomes the label on the call.
Two practical notes. First, this only works for calls that originate from your website, so a Local Services Ad that a customer taps directly in the Google interface needs the separate LSA lead handling covered in the Local Services Ads guide. Second, DNI is what makes a source attribution possible, not a keyword attribution. Knowing that a call came from paid search is useful. Knowing which search term produced it is far more useful, and that requires the next piece.
The gclid, and why it is the part that matters most
When someone clicks a Google Ad, Google appends a unique identifier to the landing page URL called a gclid. It is a long opaque string, and it is effectively the receipt for that specific click, tied to the campaign, ad group, and keyword that produced it.
Capturing the gclid at the moment the visitor lands, holding it while they browse, and attaching it to the phone call they place minutes later is the difference between coarse and precise attribution. Without it you learn that paid search produced eleven calls last month. With it you learn that a particular keyword produced eight of those calls and one booked job while another keyword produced three calls and three booked jobs, at which point you know exactly where the next dollar should go.
This is also the piece that most homegrown setups miss. Businesses often track form submissions properly and phone calls not at all, which for a trade is backwards, because the phone is where the revenue arrives. The mechanics of stitching a click to a call are covered in depth in the phone call attribution for Google Ads guide.
Recording, transcription, and lead scoring
Attribution that stops at counting calls will mislead you. A campaign that produces forty calls looks like a winner next to one that produces twelve, right up until you discover that thirty of the forty were price shoppers, wrong numbers, and people outside your service radius.
The Pro layer records and transcribes calls, then scores them and reads sentiment and intent from what was actually said. That converts a raw call count into a qualified lead count, which is the only version of the number worth making decisions on. It also gives you something the outcome data alone never will, namely the exact language of the objection. When six callers in a month balk at the same thing, you will hear it in the transcripts long before it shows up in a booking rate.
Two adjacent guides are useful here. Lead scoring for service businesses explains what a score should and should not weigh, and cost per lead walks the arithmetic of turning spend and qualified-lead counts into a number you can compare across channels. If you are going to record calls, read the call recording consent rules as well, because the requirements vary by state and this is not a corner to improvise around.
Conversion upload, or closing the loop
The last piece is the one people skip and then wonder why their campaigns never improve. Google Ads optimizes toward the outcomes you report to it. If the only thing you report is clicks and form fills, it will get very good at producing clicks and form fills from people who were never going to book.
Conversion upload sends the booked job back into Google Ads and attaches it to the original click. Now the bidding system is optimizing toward booked revenue instead of toward phone-shaped noise. This is a slow, compounding benefit rather than a dramatic one, and it is the reason attribution is worth more in month six than in month one. The broader framework for reading all of this sits in the call tracking and attribution guide.
The decision rule, tied to ad spend
Here is the rule stated as plainly as it can be. If you pay for advertising and you cannot name, from memory or from a report, which campaign produced last month's booked jobs, you are flying blind and the attribution layer is the fix. If you do not pay for advertising, there is nothing to attribute and you should stay on Core.
Everything else is a variation on those two sentences. A business with a single always-on Local Services Ads campaign and no other spend is a borderline case, because there is only one source and the attribution mostly confirms what you already assume. A business running search ads plus Meta plus an LSA budget is not a borderline case at all, because the whole question is how the budget should be split between them, and you cannot answer that without source-level data.
The $250 gap, worked honestly
The step from Core to Pro is $250 a month. The honest way to evaluate it is not to ask what percentage of ad budgets are typically wasted, because nobody can tell you that about your account and any number quoted at you is decoration. The useful question is inverted. How much waste would $250 have to represent before finding it pays for the upgrade?
That is simple division against your own spend.
At $1,000 a month in ads, $250 is 25 percent of the budget. You would need to find and kill a quarter of your spend as dead weight for the upgrade to break even on that basis alone. Possible, but not a slam dunk.
At $2,500 a month, $250 is 10 percent. One underperforming ad group out of several clearing that bar is a realistic month's work.
At $5,000 a month, $250 is 5 percent. At that spend level, the odds that not one twentieth of your budget is going somewhere it should not are slim, and the upgrade is close to an obvious yes.
Now adjust for the part most people forget, which is that the $250 is not entirely a new cost. Pro doubles included minutes from 500 to 1,000 and drops the overage rate from $0.45 to $0.40 a minute. If you are already running past Core's pool, you are paying some of that gap today. At 800 minutes a month you are paying 300 overage minutes at $0.45, which is $135, so your real incremental cost to reach Pro is $115 rather than $250. At 1,000 minutes on Core you are paying $225 in overage, at which point the gap has almost entirely closed and you are getting attribution for what amounts to rounding. The full version of that ladder is in the plan-choosing guide.
So the real test is a two-part one. Take your true incremental cost after overage credit, then ask whether that number is small relative to what one bad campaign is costing you. For a business at $3,000 in monthly ad spend already running slightly over its minutes, the answer is usually yes and it is not close.
When you should stay on Core and keep the money
This deserves its own section because the sales-shaped version of this article would not include it.
Stay on Core if your work comes from referrals, repeat customers, and organic search. There is no marketing dollar to trace, so tracing it costs $250 and returns a report confirming that your customers came from your reputation. Stay on Core if you have no ad budget at all and no plan to start one this quarter. Stay on Core if you have exactly one marketing channel and no intention of adding a second, because attribution is fundamentally a tool for allocating between options and there is nothing to allocate.
Stay on Core, too, if you are honest with yourself that nobody is going to look at the data. Attribution is not passive. Somebody has to open the numbers monthly, compare cost per booked job across sources, and actually cut the loser. If that hour a month does not exist, the upgrade buys a dashboard rather than a decision, and a dashboard nobody opens is the most expensive kind. The habits that make it stick are laid out in the KPI dashboard guide, which is worth reading before you buy rather than after.
Because every plan is month to month with zero setup fees, none of this is a permanent choice. You can add attribution the month you launch a campaign and drop it the month you stop, and stepping back down to Core never touches your CRM, your customer history, your invoicing, or your dispatch. Those live in the entry tier and stay there regardless.
The first sixty days, if you do upgrade
Attribution is worth nothing until it is set up correctly and read consistently, so treat the first two months as a project rather than a switch.
Get DNI installed on every page a paid visitor can land on, not just the homepage, because a service page that still shows the static number will quietly launder its calls into the untracked bucket. Confirm gclid capture is surviving the hop from landing page to call by placing a test call from a real ad click. Turn on conversion upload early, since it needs weeks of data before the bidding changes anything. Then leave it alone for a full month before drawing conclusions, because service call volume is seasonal enough that a two-week sample will tell you a confident lie. The seasonal call volume guide is a useful check on that instinct.
At the sixty-day mark you should be able to sit down with one page showing spend, calls, qualified calls, booked jobs, and cost per booked job, broken out by source. If that page exists and you can point at the row you are cutting, the upgrade paid for itself. If it does not exist, the problem is process rather than product, and no tier fixes that.
Put it together
Answering and attribution are separate jobs and it is worth keeping them separate in your head. The AI receptionist makes sure no opportunity is lost after the phone rings. Call tracking makes sure you know what caused it to ring, and therefore where the next marketing dollar should go.
If you advertise, cannot name your best campaign, and have someone who will read the report, the attribution layer in Pro is the right call, and the gap after overage credit is usually smaller than the sticker suggests. If you do not advertise, keep the $250 and run a very good front office on Core. Compare the current numbers on the pricing page, and if you want the arithmetic run against your actual spend and call volume, get in touch and we will work it with your numbers rather than an example.
