Table of contents
Mortgage attribution does not break because marketers are careless. It breaks because the average loan takes about 45 days to fund, Meta’s default click window is 28 days, roughly 23.9% of finance conversions happen on a phone call, and only 70–82% of applications ever become revenue.
Key Takeaways
- The MBA baseline for application-to-fund cycle time is about 45 days, against automation-stack targets of 12–18 days.
- Pull-through at a typical mid-sized lender runs 70–82%, with best-in-class targets of 85–92%.
- Application completion rates of 55–70% mean a third of started applications never reach submission.
- Google Ads allows a maximum 90-day conversion window; Meta defaults to 28 days for clicks and 7 in some configurations.
- 23.9% of finance conversions happen over the phone rather than through a form.
- Only 56% of calls to businesses are answered by a person — 57% in financial services.
- 35% of answered financial-services calls from digital marketing are genuine leads, and 34% of those convert on the call.
- 64% of businesses never ask the caller to buy or book an appointment.
- Calls originating from ChatGPT show the highest lead rate of any channel at 49%.
- Multi-touch attribution adoption reached 47%, last-touch 41%, hybrid MTA+MMM 33% and marketing mix modelling 26% (from 9% in 2023).
- 7% of teams still run no formal attribution at all.
- Only 18% of multi-touch implementations are rated highly accurate by the teams that run them.
- Multi-touch programmes are associated with an 18% ROI lift, 22% better lead quality and 15% lower CAC.
- The dark-funnel gap averages 38% of pipeline across surveyed teams.
- Post-purchase surveys routinely surface 20–40% discrepancies against tracked attribution.
- The gap between GA4 and ad-platform conversion counts commonly runs 15–40%.
- Consent denial in the 30–50% range and cookie loss approaching 60% are now standard planning assumptions.
- Lead-to-funded rates range from 0.5–2% for shared aggregator leads to 50–70%+ for past-client referrals.
- Cost per funded loan runs $5,000–$10,000+ on shared aggregator leads versus $1,000–$3,000 on first-party campaigns.
- Call-to-funded conversion is 10–20% for paid leads and 25–35% for referral leads.
- Calling back within five minutes is reported to raise connection odds by up to 100x versus a 30-minute delay.
- Integrated LOS and CRM platforms have been associated with a $1,056 increase in gross profit per loan.
- 82.7% of brokers use an LOS or CRM daily, and 71% use online applications.
- IMBs earned a pretax net production profit of $727 per loan in Q1 2026 while production costs rose to $11,898.
- Attribution-capable teams spend 23% more on martech but report 1.6x larger marketing-sourced pipeline.
The Window Mismatch Nobody Configures Around
Start with the timing problem, because it silently corrupts every downstream report. The MBA industry baseline for application-to-fund cycle time is 45 days, pull-through at typical mid-sized lenders runs 70–82% against an 85–92% target, and application completion sits at 55–70%. Meanwhile Google Ads caps conversion tracking at 90 days and Facebook defaults to a 7-day window, with most platforms rarely extending beyond 30.
| Stage | Mortgage reality | Platform default | Consequence |
|---|---|---|---|
| Click to application | Days to weeks | Within window | Usually credited correctly |
| Application to funding | ~45 days (target 12–18) | Often outside window | Revenue event uncredited |
| Meta click window | — | 28 days (7 in some setups) | Long-cycle purchase files lost |
| Google click window | — | 90-day maximum | Workable if widened manually |
| Pull-through | 70–82% fund | Not modelled at all | Applications overstate revenue |
| Application completion | 55–70% submit | Counted as conversion at start | Inflated top-of-funnel |
Compound those two rows and the arithmetic is unforgiving: if 62.5% of started applications submit and 76% of submissions fund, a campaign reporting 100 “conversions” at application start is really delivering closer to 47 funded loans. Any budget decision made on the unadjusted number is off by roughly half.
Which Models Teams Actually Run
Adoption is not accuracy. Only 18% of multi-touch implementations are rated highly accurate by their own marketing teams, even though programmes that work are associated with an 18% ROI lift, 22% better lead quality and 15% lower customer acquisition cost.
| Model | Adoption | What it answers well | Mortgage caveat |
|---|---|---|---|
| Multi-touch (MTA) | 47% | Which campaign touched this file | Blind to referral and phone paths |
| Last-touch | 41% | Nothing much, cheaply | Credits branded search for referrals |
| Hybrid MTA + MMM | 33% | Tactical and strategic together | Needs 2+ years of spend history |
| Marketing mix modelling | 26% | Marginal return by channel | Rate cycles confound the model |
| First-touch | 19% | Source of record for demand | Understates nurture and database |
| Self-reported survey | Complementary | Referral and dark-social credit | The only view of the other 38% |
The Phone Is Where Mortgage Converts
Form-based attribution misses the moment that matters. Ruler Analytics tracking data shows 23.9% of finance conversions happen via phone call. Invoca’s analysis of 70 million calls found only 56% of calls to businesses are answered by a person (57% in financial services), 35% of answered financial-services calls from digital marketing are genuine leads, 34% of those convert on the call, 64% of businesses never ask for the sale or appointment, and calls originating from ChatGPT show the highest lead rate of any channel at 49%.
| Call metric | Financial services | Cross-industry | Attribution impact |
|---|---|---|---|
| Calls answered by a person | 57% | 56% | Unanswered calls credited to nothing |
| Answered calls that are leads | 35% | 38% | CPL overstated without call data |
| Phone leads converting on the call | 34% | 42% | Fastest close path, least tracked |
| Businesses never asking for the booking | — | 64% | Lost revenue looks like weak traffic |
| Highest-lead-rate source | ChatGPT calls, 49% | — | AI referrals need their own tagging |
Handling quality then moves the numbers again. Call-to-application conversion runs 15–25% for paid leads and 30–45% for referral leads, call-to-funded 10–20% and 25–35% respectively, and calling back within five minutes is reported to raise connection odds by up to 100x versus a 30-minute delay. Without dynamic number insertion, none of that variance is attributable to a campaign. Our mortgage tracking and analytics statistics cover the instrumentation layer.
Attribute To Funded Loans, Not Leads
This is the single highest-leverage change available to most mortgage teams. Shared aggregator leads convert at 0.5–2% at $30–$100 each for a $5,000–$10,000+ cost per funded loan; premium aggregator leads run 1–3% for average operations and 8–10% for well-run ones; first-party paid campaigns convert at 2–5% at $15–$60 for a $1,000–$3,000 cost per funded loan; organic leads 5–12%; database reactivation 10–20%; agent referrals 40–60%; past-client referrals 50–70%+.
| Lead source | Lead-to-funded | Typical CPL | Cost per funded loan |
|---|---|---|---|
| Shared aggregator | 0.5–2% | $30–$100 | $5,000–$10,000+ |
| Premium aggregator (average ops) | 1–3% | $120–$250 | $4,000–$25,000+ |
| Premium aggregator (well-run) | 8–10% | $120–$250 | $1,750–$2,200 |
| First-party paid campaigns | 2–5% | $15–$60 | $1,000–$3,000 |
| Organic and SEO | 5–12% | ~$0 media | Dramatically lower |
| Database reactivation | 10–20% | ~$0 media | Very low |
| Agent referral | 40–60% | Relationship cost | Lowest |
| Past-client referral | 50–70%+ | Relationship cost | Lowest |
A dashboard reporting cost per lead treats the first and last rows as comparable. They differ by a factor of roughly forty on cost per funded loan. Once funded outcomes are imported back into the ad platforms, bidding starts optimising toward the sources that actually close — the point of our data intelligence engagements.
The Untracked Share, Sized Honestly
Some of the mortgage funnel will never be tracked, and pretending otherwise produces worse decisions than admitting it. 38% of pipeline sits in the dark funnel across surveyed teams, and post-purchase surveys routinely find 20–40% discrepancies against tracked sources. In mortgage the untracked layer is structurally larger, because nearly 90% of borrowers arrive through a referral or existing relationship rather than a click.
| Untracked path | Why tracking misses it | Best available signal |
|---|---|---|
| Realtor referral | Verbal handoff, no click | Partner source field in CRM |
| Past-client referral | Phone or text introduction | Self-reported form field |
| Builder or CPA referral | Offline relationship | Named-partner tagging |
| Social DMs and group chats | No pixel, no UTM | Self-reported survey |
| AI assistant recommendation | No referrer on many calls | Call-source tagging |
| Branded search after word of mouth | Credited to search | First-touch survey question |
The workable structure is a required “how did you first hear about us” field on every high-intent form, a partner field that survives the CRM-to-LOS handoff, and an explicit acceptance that the residual belongs to marketing-mix analysis rather than click attribution.
Modeled Conversions And The GA4 Trust Gap
Consent changes have turned platform reports into estimates. GA4 fills unconsented behaviour with machine-learning estimates and blends them into reported figures, with consent-denial rates in the 30–50% range and cookie loss approaching 60% commonly cited. The practical result is a persistent 15–40% gap between GA4 and each ad platform’s own conversion count — platforms use server-side signals and view-through logic GA4 does not.
| Data source | What it is good for | What it should never decide |
|---|---|---|
| GA4 | On-site behaviour and page paths | Channel budget or commissions |
| Ad platform conversions | In-platform bidding signals | Board-level ROI claims |
| Call tracking | Phone-sourced revenue | Digital-only funnel views |
| CRM applications | Pipeline health, daily ops | Final revenue attribution |
| LOS funded units | Revenue truth | Real-time optimisation |
| Self-reported survey | Referral and dark-funnel credit | Precise channel percentages |
The hierarchy matters more than the tooling. GA4 describes behaviour, the platforms optimise delivery, and the loan origination system is the only place revenue is real.
The Plumbing Pays Before The Reporting Does
Attribution in mortgage is mostly an integration project. A MarketWise Advisors study found lenders running integrated origination and CRM platforms saw a $1,056 increase in gross profit per loan, yet most operations still run the two systems in silos. Adoption of the underlying tools is already high: broker survey data shows 87.3% using e-signature, 82.7% an LOS or CRM, 80.2% pricing calculators and 71% online applications — the systems exist, they just do not talk.
The margin context explains the urgency. MBA reporting shows independent mortgage banks earned a pretax net production profit of $727 per loan in Q1 2026 while total production expenses rose to $11,898 per loan, with 76% of lenders profitable overall. On a $727 margin, a mis-allocated marketing budget is not an analytics inconvenience; it is the difference between a profitable and unprofitable quarter.
| Component | What it fixes | Effort | Priority |
|---|---|---|---|
| Call tracking with DNI | The 23.9% phone blind spot | Low | First |
| Self-reported source field | Dark-funnel and referral credit | Low | First |
| Offline conversion imports | Optimising to funded loans | Medium | Second |
| CRM-to-LOS source persistence | Lead-to-funded joins | Medium | Second |
| 90-day lookback windows | 45-day cycle mismatch | Low | Second |
| Monthly three-way reconciliation | Trust in the numbers | Medium | Ongoing |
| Marketing mix modelling | Brand and offline effects | High | Later |
Note the sequencing: the two cheapest items recover the largest share of missing credit. Teams that build this capability spend about 23% more on martech and report 1.6x larger marketing-sourced pipeline — a premium that reads very differently once cost per funded loan is the reporting unit.
A Reporting Standard Worth Adopting
- Report cost per funded loan as the headline metric; keep cost per application as the daily signal.
- Widen lookback windows to the platform maximum and note the 45-day cycle in every deck.
- Track calls as conversions — 23.9% of finance conversions never touch a form.
- Import funded outcomes to Google, Meta and LinkedIn monthly so bidding learns from revenue.
- Publish the pull-through rate next to marketing metrics so a 70–82% haircut is visible.
- Run one self-reported question forever; it is the only view of the 38% dark funnel.
- Reconcile platform, CRM and LOS counts monthly and document the variance instead of hiding it.
Done consistently, attribution stops being a debate about models and becomes a margin tool. For teams comparing channel performance across the whole funnel, our mortgage digital marketing statistics and mortgage Google Ads statistics provide the benchmarks to reconcile against, and our growth marketing team builds the wiring.
Frequently Asked Questions
Why is marketing attribution harder in mortgage than in most industries?
Three structural reasons compound. First, timing: MBA data puts the industry baseline for application-to-fund cycle time near 45 days, while Meta's default click window is 28 days and Google's maximum is 90 — so the revenue event routinely lands outside the window that credited the click. Second, the phone: roughly 23.9% of finance conversions happen on a call, and calls are invisible to form-based tracking unless call tracking is wired in. Third, pull-through: only 70–82% of applications at a typical mid-sized lender actually fund, so a dashboard that stops at 'application' overstates marketing performance by up to 30% before anything else goes wrong.
What conversion event should a mortgage lender attribute to?
Funded loans, with application as the leading indicator. Cost per lead is almost meaningless in mortgage because lead-to-funded rates range from 0.5–2% for shared aggregator leads to 40–60% for agent referrals and 50–70% for past-client referrals. Two campaigns at an identical $80 cost per lead can differ by a factor of forty in cost per funded loan. The practical setup is to import the funded event from the LOS or CRM back into each ad platform as an offline conversion, keep application as the daily optimisation signal, and reconcile monthly.
How much of mortgage marketing impact is simply untrackable?
A material share, and the honest answer is to size it rather than deny it. Attribution research across 1,200+ teams puts the dark-funnel gap at about 38% of pipeline, post-purchase surveys routinely find 20–40% discrepancies against tracked sources, and in mortgage the untracked layer is unusually large because nearly 90% of borrowers arrive via referral or existing relationship. The mature response is a self-reported attribution field on every high-intent form plus a partner-level source field, then treating the residual as marketing-mix territory instead of forcing a click path onto it.
Do modeled conversions in GA4 make mortgage reporting unreliable?
They make it approximate, which is different from useless. With Consent Mode enabled, GA4 fills in unconsented behaviour with modeled estimates and reports them blended with observed data, and consent-denial rates in the 30–50% band with cookie loss approaching 60% are commonly cited. The gap between GA4 and the ad platforms' own conversion counts frequently runs 15–40%. For mortgage, the fix is not to abandon GA4 but to demote it: use it for on-site behaviour, use server-side and CRM data for revenue truth, and never let a rate-table page's GA4 conversion count drive commission or budget decisions.
What does good mortgage attribution infrastructure look like in 2026?
Six components. Call tracking with dynamic number insertion on every page and campaign. A single source field that survives the handoff from CRM to loan origination system. Offline conversion imports of funded loans into Google, Meta and LinkedIn. A self-reported 'how did you hear about us' field on high-intent forms. Lookback windows widened to 90 days where the platform allows it. And a monthly reconciliation between platform-reported conversions, CRM applications and funded units. Integrated LOS and CRM environments have been associated with a $1,056 increase in gross profit per loan, so the plumbing pays for itself before the reporting does.
Sources
Digital Applied — Marketing Attribution Statistics 2026 (140 data points)
MarqOps — Multi-Touch Attribution in 2026
Confer Solutions — 10 Mortgage LOS KPIs for Mid-Sized Lenders 2026
leadPops — Mortgage Lead Conversion Rates by Source
MortgageLeads.com — Mortgage Call Conversion Rates 2026
Invoca — Lead Conversion Benchmarks 2026 (70M calls)
Ruler Analytics — Finance Marketing Statistics 2026
Cometly — Long Sales Cycle Attribution Guide 2026
ACCS — Cookieless Attribution and Modeled Conversions in GA4
MortgageWorkSpace — Integrating LOS and CRM Platforms
HousingWire — IMB Profit and Costs, Q1 2026 (MBA)
