How lending floors should split fresh internet leads from aged stock: the routing, waterfalls, licensing filters and source reporting that make each one pay, from operators who run these campaigns daily.
How should fresh and aged mortgage leads be dialed?
Separately, always. Fresh internet leads need a call within minutes, a priority queue and an aggressive first-day cadence. Aged leads need a waterfall: priority tiers by age and source, gentler pacing, attempt caps and rest periods. Blended together, the pacing engine chases an average that matches neither, fresh leads queue behind cold records, and reporting can no longer show which source pays.
One campaign for everything is the most expensive tidy-looking decision on the floor.
Walk the lending floors we audit and you will usually find one campaign called something like Internet Leads. Inside it: this morning's exclusive web leads, a shared batch from three weeks ago, and forty thousand records bought for pennies last quarter. It looks tidy in the campaign list, and it quietly ruins all three.
Start with the economics. A fresh exclusive lead costs many multiples of an aged record, converts on a horizon of hours to days, and rewards speed above everything else. An aged record costs a fraction of the price, converts on a horizon of weeks if at all, and rewards patient, cheap capacity. Spending premium treatment on cheap stock is waste. Spending cheap treatment on premium stock is worse, because the lead you paid top dollar for goes stale in a queue behind someone's 2023 refi list.
Then the pacing math. A predictive dialer tunes its pacing to the answer rate it observes on the campaign. Blend a list that answers well with one that barely answers and the engine paces to an average that matches neither: it over-dials whenever the hot records surface and crawls through the cold ones. The over-dialing has a regulatory edge, because the federal safe harbor caps abandoned calls at three percent of calls answered live by a person, measured per campaign over each 30-day period, with a live agent required within two seconds of the completed greeting. Abandon rate is always abandoned calls divided by live answers, never by dials, and a blended campaign spends that three percent budget on your most valuable answers first.
Reporting rounds it off. A blended campaign's contact rate describes nothing that actually exists: not the fresh flow, not the aged buy, not either vendor. You cannot kill an underperforming source when its numbers are averaged into everyone else's. Separate campaigns are the cheapest reporting fix on the platform, and they also keep caller ID posture clean, since the numbers that grind through aged stock all day attract spam labels faster and should never be the same numbers carrying your fresh traffic.
Priority queues, sub-minute handoffs, and an honest look at the two routing patterns.
A fresh internet lead is a perishable good with a shelf life measured in minutes. The best public evidence remains Harvard Business Review's audit of 2,241 US companies: only 37 percent responded to a web-generated lead within an hour, 23 percent never responded at all, and firms that attempted contact within the hour were nearly seven times as likely to qualify the lead as those that tried even an hour later. That study is from 2011 and nobody has published anything better since, which tells you something about how this industry treats its own data. We keep a running review of what evidence does exist in our speed to lead evidence article, and our speed to lead build exists because borrowers fill in several forms in one sitting and the first competent call usually takes the application.
The mechanics are not complicated, just rarely done. The web form posts into a dedicated Five9 list through the API within seconds of submit: no spreadsheet, no hourly import, no human in the loop. That list lives in its own campaign at the top of the priority order, so a fresh record preempts every piece of aged work on the floor. The agents who take these calls sit in a dedicated skill with a short ring timeout, and the number to manage is lead-to-first-dial in seconds, on a wallboard, every day. Floors that measure it in minutes get minutes; floors that do not measure it get whatever the import schedule allows.
The lead is offered to several available originators at once and the first to accept takes it. Fastest possible response, no single point of failure, which is why high-velocity consumer-direct floors like it. The costs: a claim race that rewards quick fingers over the best-matched originator, duplicate work when the CRM is slow to lock the record, and friction when the same two people scoop up most of the flow. It suits comp plans that pay on speed and teams of roughly even skill.
Leads are assigned in strict rotation, next in line takes the next lead. Fair, predictable, easy to staff and easy to defend in a comp dispute. The cost is that one slow originator becomes the bottleneck for every lead that lands on them, so it only works with an SLA timer: no acceptance within a short window, sixty seconds is common, and the lead bounces to the next in line with the miss logged.
Having run both, our observation is that the comp plan decides this more than the telephony does. Pay on speed and shotgun works; pay on conversion quality and round-robin with a hard SLA is the calmer floor.
Our free 17-question Five9 Health Check scores your campaign structure, pacing and compliance filters against the setups described here, and tells you what to fix first.
Tiers, rest periods and exit rules turn cheap stock into cheap funded loans.
Aged internet leads carry a reputation they only half deserve. Dialed flat, oldest to newest, they burn agent hours and number reputation for scraps. Worked as a waterfall, they are routinely the cheapest funded loans on the floor, and the discipline is not clever: it is priority tiers, worked in order, with the movement rules written down.
| Tier | What goes in | Treatment | Exit rule |
|---|---|---|---|
| 1 | Recent misses: leads under 30 days old that were never reached, plus fresh-day leftovers | Worked daily in the best calling hours, power or predictive pacing | Rest after the attempt cap, re-enter at tier 2 |
| 2 | 30 to 90 days old, plus tier 1 rest-outs | Worked every second or third day, rotating hours | Rest 30 to 60 days, re-enter at tier 3 |
| 3 | 90 days to a year, bulk purchases | Fills spare capacity and off-peak hours only | Rest 60 days, one more pass, then archive |
| 4 | Over a year old | Better suited to a re-permission email pass than a cold dial; dial only what re-engages | No response to re-permission: archive |
Two mechanisms make the table work. The first is attempt caps: a record earns its rest after a set number of attempts, and the floors we run mostly settle between six and ten before the first rest, though the right figure is wherever your own attempt curve goes flat rather than anything we could print here. The second is terminal exits: a wrong number, a do-not-call request, or a clear no twice over takes the record out of the waterfall for good, and the suppression file it lands in outlives every campaign.
Source matters as much as age. A 60-day-old record from a vendor whose leads answer is worth more than a 20-day-old record from one whose leads never did, so let source-level results (covered below) promote and demote whole batches between tiers. And ask every aged vendor two questions before the file loads: when was consent captured, and to what wording. Records gathered under permissive 2019-era language may not meet the standard you want to rely on now, and vendors rarely raise the subject unprompted.
Front-load attempts, then taper: the curve flattens whether you plan for it or not.
Every attempt-curve report we have ever pulled has the same shape: the first few attempts do most of the work and each attempt after that returns less. You will find charts online assigning precise percentages to attempt five versus attempt six; we have never been able to trace one back to a source worth citing, so we will not repeat their numbers. The shape, though, is real, and your own campaign reports will draw it for you this afternoon.
The shape implies front-loading. On day one a fresh lead gets several attempts spread across different hours, because a miss at 10am usually means work and a miss at 7pm usually means dinner, and only a different hour fixes either. Days two to five, one or two attempts a day, rotating the hour each time. Then a taper: every few days, then weekly, then the record joins the aged waterfall. Late attempts are not worthless, they are just cheap, so schedule them where capacity would otherwise sit idle.
On the best time to call, the honest answer is that your own connect-by-hour report, cut by lead source, beats any published chart. On the mortgage floors we run, lunch and early evening generally out-pull mid-morning for consumer borrowers, but the pattern moves with source, state and season, which is exactly why it should come from your data rather than ours.
The legal envelope sits around all of it. The federal rule confines telephone solicitation to 8am to 9pm local time at the called party's location, and states layer their own restrictions on top: Florida, for one, caps solicitation calls at three per 24-hour period on the same subject matter. State telemarketing statutes have been busy since 2021, so the cadence engine needs a state-by-state overlay for hours and attempt frequency, and that overlay should be checked against current statute text rather than a blog post from two legislatures ago.
Layered list filters, not good intentions, keep unlicensed states out of the hopper.
The SAFE Act requires anyone acting as a residential mortgage loan originator to be state-licensed or federally registered, with state licenses issued and tracked through NMLS. Where exactly a marketing call crosses into origination activity is a question for your compliance team, not a dialer setting. What operators own is blunter: the dialer should be physically unable to call a state you have not cleared, and that takes layered filters, because any single filter fails eventually.
The floors that get this right treat the license table as live reference data with a named owner, reviewed monthly, feeding every filter above from one place. The floors that get it wrong keep it in a spreadsheet titled FINAL_v7. Licensing scope and state calling rules both shift, so verify the current position for each state you dial and take proper advice where the answer decides a campaign.
Signed September 2025, effective March 2026: the trigger-lead market is not what it was.
A trigger lead is created the moment a lender pulls a borrower's credit for a mortgage: the credit bureaus would sell that event, within hours, to competing lenders who dialed the borrower while the original application was still in flight. For buyers it was cheap, high-intent data. For borrowers it was dozens of calls at exactly the wrong moment, and for the originating lender it was watching competitors work a client it had just paid to acquire.
That market is now largely closed. The Homebuyers Privacy Protection Act was signed on September 5, 2025 and took effect in early March 2026. It amends the Fair Credit Reporting Act so that a consumer reporting agency may furnish a mortgage trigger lead only in narrow cases: with the consumer's documented authorization, or to the consumer's current mortgage originator or servicer, or to an insured bank or credit union holding an active account for that consumer. A GAO study of trigger leads delivered by text message is due in September 2026, so the rules in this corner may not have finished moving.
What that means on a dialing floor in mid-2026 is fairly plain. If a vendor is still offering credit-event leads, ask which statutory exception the furnishing relies on and get the answer in writing before a single record reaches a list. If your shop is itself a permitted party, the current servicer for instance, the furnishing may be lawful while your outbound consent position remains a separate question with its own set of rules. The Act is young and enforcement practice around it is younger; read the current text and involve qualified counsel before building a campaign on any particular interpretation.
Cost per funded loan by source is the number the whole strategy answers to.
Lead buying should be a feedback loop. On most floors it is a habit, renewed monthly because the vendor relationship exists and nobody can prove it is failing. Proving it either way takes four numbers per source, and the last is the only one your finance director will read.
| Metric | Definition | What it decides |
|---|---|---|
| Cost per lead | Spend on the source divided by records delivered | Nothing on its own; cheap stock that never answers is expensive |
| Contact rate | Leads with at least one live conversation with the named borrower, divided by leads worked | Whether the data is real and the numbers are right |
| Application rate | Contacts that become submitted applications | Whether the intent is real and the pitch fits the source |
| Cost per funded loan | Source spend divided by loans funded from it | The buying decision, full stop |
The trap is attribution lag. Funded loans land 30 to 90 days after the lead was bought, so the source identity has to survive the whole journey: vendor to dialer list to CRM to LOS. On floors where the source field dies at a handoff, buying reverts to gut feel and the loudest vendor wins. Carry vendor, batch and tier on the record all the way to funding, and date the spend so the comparison is batch against batch, not this month's spend against last quarter's fundings.
The same numbers should also reach backward into the waterfall: a source whose aged records still contact well earns a higher tier for its whole back catalog, and a source that goes dead at 30 days gets rested early and bought thinner. When the segmentation, cadence and reporting above are all in place the movement is not subtle; across our client portfolio, rebuilt outbound setups average around a 40 percent higher connect rate than the configurations they replaced. That is a portfolio aggregate rather than a promise, but the direction has been consistent for years.
Screen pops with loan context, Velocify write-backs, and dedupe against your own pipeline.
None of the above survives contact with a floor where the dialer, the CRM and the LOS do not talk to each other. Three notes from stacks we have built and repaired.
Salesforce first. Five9's Salesforce integration runs the agent inside Salesforce and pops the matching record on connect. The detail that decides whether it earns its keep is what pops: the loan opportunity with property state, amount, stage and last-conversation notes, not a bare contact card. A screen pop without loan context saves the agent three seconds of searching; one with context changes the first sentence of the call, and the first sentence is where fresh-lead conversions live.
Velocify and Encompass next. Plenty of lending floors run Velocify, ICE Mortgage Technology's lead management product, for distribution and cadence logic, alongside Encompass as the LOS in the same ICE ecosystem. The division of labor that works: Velocify decides who gets called next and why, the dialer does the calling with the pacing, licensing filters and abandon management, and dispositions write back promptly so the cadence engine advances. Where the write-back is missing, both systems believe they own the cadence and the borrower hears from you twice before lunch, which is an efficient way to convert a warm lead into a complaint.
Last, dedupe against your own book. Aged purchases overlap with your past applicants and current pipeline more often than vendors admit, so suppress every load against the active Encompass pipeline and the funded-loan file before it dials. Power-dialing a borrower who funds with you next week is the cheapest own goal in the business, and most floors only need to do it once in front of the sales director to fund the integration work that stops it.
There is no official cutoff. Operationally, a lead is fresh while the borrower still remembers filling in the form, which in practice means minutes to a few days. Most floors treat anything under 24 to 72 hours as fresh and everything older as aged, then tier the aged stock by month. The useful test is behavioral: if speed still changes the outcome, it is fresh; if persistence matters more than speed, it is aged.
Within minutes, ideally under five. The strongest public evidence is Harvard Business Review's 2011 audit of 2,241 companies, which found firms attempting contact within an hour were nearly seven times as likely to qualify the lead as those that tried even an hour later. Borrowers submit to several sites in one sitting, and the first competent call usually wins. Automate the handoff from web form to dialer list; a human forwarding a spreadsheet is already too slow.
Cap attempts and read your own attempt-curve report: attempts per record against incremental contacts. The floors we run mostly settle between six and ten attempts before a first rest of 30 to 60 days, after which the record re-enters one tier down. The exact number matters less than having a cap at all, because uncapped campaigns keep spending agent time and caller ID reputation on records that stopped paying several attempts ago.
Your own connect-by-hour report, cut by lead source, is the only trustworthy answer; published best-time charts rarely survive contact with a specific floor. Patterns we commonly see are stronger connects at lunchtime and early evening for consumer borrowers, but this shifts by source, state and season. Whatever you find, the federal rule confines solicitation calls to 8am to 9pm at the called party's location, and several states set tighter hours.
The SAFE Act requires loan originators to be state-licensed or federally registered, and marketing calls sit close enough to origination that most lenders treat unlicensed-state dialing as off-limits. Operators typically enforce this with layered filters: a licensed-state scrub at list import, campaign-level state filters as a second gate, and routing so connects land with an originator licensed in that state. Where exactly the rules bite for your model is a question for your compliance team.
Mostly no, as a purchasable product. The Homebuyers Privacy Protection Act, signed in September 2025 and effective March 2026, amends the FCRA so credit bureaus may furnish mortgage trigger leads only with the consumer's documented authorization, or to the consumer's current originator or servicer, or to a bank or credit union holding an active account. If a vendor still offers credit-event leads, ask which exception applies and get the answer in writing.
Shotgun (offer to several originators, first to accept wins) is faster and suits floors where comp rewards speed; it breeds claim races and favors the quick over the well matched. Round-robin is fair and predictable but lets one slow originator become the bottleneck unless an SLA timer skips them after a short window. In our experience the comp plan decides this more than the telephony does. Either way, fresh leads must outrank all other outbound work.
US telemarketing rules cap abandoned calls at three percent of calls answered by a live person, measured per campaign over each 30-day period, with a live agent required within two seconds of the person's completed greeting. Abandon rate is always calculated against live answers, not dials. Mixing a high-answer fresh list into an aged campaign concentrates abandons on your most valuable answers, which is one more reason the two belong in separate campaigns.
We design, build and run Five9 floors for mortgage and lending teams in the US and UK: fresh-lead routing, aged waterfalls, licensing filters and the reporting behind the buying. A dedicated pod, live in weeks.
Talk to us