The loan officer tech stack: CRM, dialer, local presence

Buying leads is a plumbing problem for about a week. After that it is an operating problem, and the operating problem is tooling.

Three jobs sit between a lead landing and a borrower answering. Every tool sold into this market does one of them.

• Hold the lead. Take it in without a human, name an owner, start a clock.
• Dial it. Carry six attempts per lead without eating a producer's day.
• Get the call answered. Show a number that rings rather than one a carrier has flagged.

A desk missing any one of the three loses leads no matter how good the other two are.

Where each tool sits

What each one is for

The CRM: the only one that has to exist

For lead flow, a CRM has four jobs. A desk can check all four in an afternoon, and the feature list is irrelevant to every one of them.

• Accept a lead over a webhook, with nobody pasting anything anywhere.
• Name an owner at the moment it lands, not at the next stand-up.
• Start a clock on the first attempt that somebody can see.
• Log every attempt against the lead, so attempt four knows about attempt three.

The common failure is not a missing feature. It is a lead arriving in a shared inbox, where it belongs to everybody and therefore to nobody.

Delivery speed is not usually the bottleneck

These two get blamed for each other, so they are worth separating. On our side, delivery into a client CRM has a median under one second, nine of ten leads are there inside five minutes, and 94.8 percent of landing-page leads are delivered.

Source: our own production delivery timing, measured 2026. First-party measurement, stated as such.

If the plumbing runs that fast and the first call still goes out hours later, the gap is not in the software. It is in who owns the lead during the first ten minutes.

The only mortgage-specific field test we can find on that gap is old and it was run by a vendor. Internet leads were submitted to lenders and the phone was timed.

Among the lenders that did call back inside a day, the median wait was just under three hours and the average was seven.

Source: 130-lender mystery shop, published by National Mortgage Professional on 11 January 2010, run by a lead qualification vendor. More than 350 internet leads submitted to more than 130 lenders, weekdays, before 5pm. nationalmortgageprofessional.com · Vendor research, not a controlled study, and sixteen years old. The same article also asserts a large conversion uplift for calling first, with no citation anywhere in it, so that part is not reprinted here.

The dialer: attempts per hour is the spec

The workload is set by the data rather than by taste. In a Velocify study of 3.5 million leads published around 2013, 93 percent of the leads that eventually converted had been reached by the sixth attempt.

Source: Velocify study of 3.5 million leads, published around 2013. The original paper is no longer hosted and every copy in circulation is a secondary citation, so quote it with the year attached.

We take that figure apart in how many calls it takes to reach a mortgage lead. For a tooling decision, only the load matters: six attempts per lead, spread over days, at different hours.

What to look at, in this order:

• Attempts per hour per seat. Multi-line dialing exists for this number and nothing else.
• Dialing straight from the list. If a caller copies a number between two windows, the cadence dies quietly.
• Attempt logging back into the CRM. A dialer that does not write back turns attempt four into a guess.
• Calling-window enforcement. The borrower's local time, per lead, not the caller's clock.
• Number reputation handling. Covered below, and the part vendors talk about least.

A CRM with a phone system already built in is usually enough for a desk of one or two callers. A separate dialer earns its keep when attempts per hour becomes the binding constraint, and that is a measurable moment rather than a feeling.

The outbound number, and the question attached to it

Local presence is the industry term for a dialer showing an outbound number that matches the area code it is calling. It exists because an unfamiliar area code gets answered less.

Two vocabulary notes first, because the wrong word here is expensive.

• "Neighbor spoofing" is the FCC's word for the scam version. It is not a synonym for local presence and it should never be used about your own workflow.
• Caller ID substitution and caller ID reputation are different problems. A number can be entirely yours and still get labelled by a carrier's analytics engine.

Source: FCC consumer guide, Caller ID Spoofing, which is where the word neighbor spoofing is defined. fcc.gov

What the two texts say

We are not going to tell you whether your setup is compliant. That is your counsel's call and this page is not legal advice. What we can do is put the texts in front of you.

The federal caller ID statute turns on intent. That is 47 U.S.C. 227(e)(1), the Truth in Caller ID Act of 2009.

It makes it unlawful to knowingly transmit misleading or inaccurate caller identification information "with the intent to defraud, cause harm, or wrongfully obtain anything of value".

Source: 47 U.S.C. 227(e)(1), Truth in Caller ID Act of 2009. law.cornell.edu

The operating rule for outbound sales is the FTC's Telemarketing Sales Rule. 16 C.F.R. 310.4(a)(8) makes it a violation to fail to transmit a number to a caller identification service.

It then carries a proviso: substituting the seller's own customer service telephone number is permitted, "which is answered during regular business hours".

Source: 16 C.F.R. 310.4(a)(8), FTC Telemarketing Sales Rule, current text. ecfr.gov

Those two sentences generate four questions, and all four are operating questions rather than legal ones.

• Is every number in the pool one your own account controls?
• Does a callback to it reach you?
• Is it answered during regular business hours?
• Does the caller name themselves and the company promptly on the call?

A rotating pool of numbers that rings nowhere fails those four before anybody opens a statute.

Reputation is a separate job, and it never ends

Caller ID authentication signs a call at the originating carrier, and the level of attestation depends on that carrier being able to verify you are authorised to use the number you present.

Source: FCC, Combating Spoofed Robocalls with Caller ID Authentication, the STIR/SHAKEN framework. fcc.gov

Two consequences a desk feels directly:

• Numbers held inside your own dialer account authenticate better than numbers you do not control, so the careful version is also the version that connects.
• Authentication does not stop a carrier's analytics engine labelling a number with a bad history. Getting a flagged number cleared is a standing task, not a setup step, and it is worth asking any dialer vendor who does that work and how long it takes them.

A second channel, and the thing nobody sells you

Text alongside the dials is worth having for one reason: the number ringing back arrives attached to a name the borrower recognises.

A category of service now puts blue-bubble messaging inside a CRM, so the message lands in the handset's own format rather than as a green SMS. Two things belong on the purchase order next to it.

• There is no public sending interface for that format. The handset maker's own business messaging product is customer-initiated and does not render the same way, so these services run infrastructure outside the sanctioned channel. That is an availability risk to plan around, not a compliance one.
• Skipping a carrier registration is not a consent exemption. A message is a message. Prior express written consent, honoured opt-outs and calling hours apply to a blue one exactly as they apply to a green one.

Buy in this order

1. Fix ownership. A named person, a named window, written down. Costs nothing and changes the most.
2. Then CRM intake. Webhook in, owner assigned on arrival, clock visible to the floor.
3. Then dialing capacity, once you can see how many attempts a seat actually gets through in a day.
4. Then the numbers. Local presence and reputation work, with the four questions above answered before the first dial.
5. Then a second channel, last, because it is the only one of the five that does nothing on its own.

What none of it fixes

• No tool creates consent. It is a property of what the consumer agreed to. It arrives with the lead or it does not.
• No tool assigns a cadence owner. That is a rota with a name on it.
• No tool changes the loan. Once a file is in an origination pipeline the outcome is underwriting, and the public data there is the good part of this market: 79.8 percent on purchase and 77.5 percent on refinance.

Source: ICE Mortgage Technology Origination Insight Report, December 2021. The denominator is applications that reached a loan origination pipeline, which is already a filtered population. The report is no longer publicly hosted, so it is cited here without a link.

The credibility tier for that figure and every other number quoted at you is in what the data supports between a lead and a funded loan.

Which leaves the only arithmetic that should decide a tool purchase. At a 5 percent average close rate, twenty leads sit behind one funded loan.

Take the annual cost of any tool, divide it by the loans it plausibly adds at that ratio, and most of these decisions answer themselves.

The short version

• A CRM that takes the lead in by itself, names an owner and logs the attempts.
• Dialing capacity sized to six attempts per lead, bought when a producer's hour is what is being spent on them.
• Numbers you control, answered in business hours, with reputation treated as a standing job.
• A second channel only if the callback needs a name on it, on the same consent rules as the calls.
• A written rule for who runs attempts one to six. Free, and it decides more than the other four.

The note on the trigger lead ban covers who else is calling your applicant while your team is on attempt three, and buying questions rather than operating ones are on our FAQ.