Section 06
Who Flips Your Loan
How this works
The SBA guarantee isn’t just insurance. It turns your loan into a product.
A Treasury-backed income stream sells at a premium on the secondary market — roughly $109 for every $100 of guaranteed principal. On a $1M loan with an 85% guarantee, that’s a $76,500 gain the week your loan funds. The originating lender books it as revenue. You never see it.
Some of this is fine. A liquid market turns ten-year illiquid paper into a security, and that lowers cost of capital across the whole program. If your bank sells your loan to recycle capital, you won’t know and it won’t matter.
What matters is whether your lender is actually a bank.
Real banks fund 7(a) out of deposits. They can hold the loan for a decade because their cost of capital is someone’s checking account. Selling is optional.
Other lenders don’t have that option. No deposits. No balance sheet. Capital from warehouse lines and the next premium they can book. These firms aren’t banks. They’re loan factories for the secondary market. Originate, flip the guaranteed portion at closing, pocket the 9 points, move to the next deal. The lending is the cost of making inventory. The flip is the business.
The FY25 league table ranks them both on volume. It won’t tell you which one just wrote your loan.
Section 06 — the hero
The case
“Sixty-nine percent of whether a 7(a) loan gets sold is explained by which lender wrote it.”
The univariate R² of the sold-flag across eight candidate factors comes out at 0.689 for lender. Subprogram explains 0.282. Rate structure, 0.077. State, 0.057. Term, 0.097. Size band, 0.018. Lender explains the sold-flag by a factor of seven or more against every borrower-side input tested. (7(a), FY25, n=65,529, univariate R² on sold-flag, 8 candidate factors)
Same deal, seven lenders, seven answers to the question “does this loan end up on a balance sheet or in a pool certificate?” The lender decided before you saw the term sheet. The decision set your price for the life of the loan.
Chart 6A
Volume rank says nothing about whether your lender keeps the loan.
Huntington funded 6,524 7(a) loans in FY2025 and sold 148 out of 16,406 three-year variable originations. Lendistry funded 1,774 and sold 2,840 of 2,843. Both appear on the same league table. One is a portfolio lender. The other originates loans to sell. The row order won’t tell you which is which.
The sorting mechanism
Holders keep the loan. They collect coupon income for up to ten years — the 7(a) cap on most working-capital and acquisition paper. Flippers book a one-time gain-on-sale the week the loan funds, service a retained strip for four basis points, and move to the next one. Two revenue models. One league table.
The reason is balance-sheet structure. Nonbanks have no deposits. They can’t warehouse a decade of loans on a funding base they don’t have. So they sell. Banks with deposit funding get to choose, and the money-center franchises mostly choose to hold. The bifurcation you see in Chart 6A is what happens when a charter decision meets a funding constraint.
Inside every size band, the sold cohort pays 51 to 77 basis points more than the held cohort on the same deal. (7(a), FY25, n=56,741, variable funded; 6 size bands) The gap isn’t credit. Collateral presence shifts the rate by two points and the spread by 201 basis points. Pricing reads collateral. Secondary-market appetite reads coupon. Sell-channel loans carry the coupon that funds a premium bid. Hold-channel loans get priced to the lender’s own cost of funds. The borrower signs one term sheet.
FY25 up-front premium captured by originating lenders on 7(a) guaranteed-portion sales: ~$1.15 billion. Roughly 46 percentof FY25 Year-1 industry lender revenue on the full 7(a) book. For Newtek, Readycap, Lendistry, Northeast, and BayFirst, the premium share runs past 70%. These firms aren’t lenders. They’re originators for a secondary market. The lending part is a feeder. Gain-on-sale isn’t a side business. It’s the business.
Chart 6B
Same size. Same year. Sold loans cost more.
The gap survives every size control. It survives credit controls too — collateral presence shifts the rate by two points but shifts the spread by 201 basis points. Pool investors treat the SBA guarantee as covering collateral-shortfall risk. They bid on coupon. A Prime+3.00% variable 7(a) is simply a more valuable cash-flow stream than an identical Prime+1.50%. It commands a higher premium. Sell-channel lenders book that premium at closing. Hold-channel lenders monetize the coupon. Both talk to the same borrower.
| Size band | n sold | n held | Sold vs held spread | Sold | Held | Gap (bps) |
|---|---|---|---|---|---|---|
| $0-$150K | 9,590 | 18,329 | P+4.03% | P+3.37% | +65 | |
| $150K-$350K | 4,571 | 6,201 | P+3.16% | P+2.40% | +77 | |
| $350K-$500K | 3,035 | 2,973 | P+2.62% | P+1.96% | +66 | |
| $500K-$1M | 2,785 | 2,569 | P+2.24% | P+1.73% | +51 | |
| $1M-$2M | 1,884 | 1,528 | P+2.09% | P+1.49% | +60 | |
| $2M-$5M | 1,838 | 1,438 | P+1.89% | P+1.14% | +75 |
The revealed preference
Five findings, stitched together.
First.The market saturates at 60%. Inside a single bucket ($500K–$1M variable 7(a) Guaranty, FY25, n=5,210), sold-rate climbs from 3% at P+<0.5% to 60% at P+1.5–2.0%, then flattens across the P+1.5% to P+3.5% band. Above P+3.5% it doesn’t rise further. Investors price prepayment: a P+3.5%+ borrower is refinancing the first day rates fall (52% of P+2.75%+ loans pay off before half their term). The premium bid caps out. The sale incentive caps out with it. Whoever originates above the ceiling eats their own duration bet.
Second. Live Oak runs two books inside one charter. Sold and held cohorts price 65 to 106 basis points apart in every size band. (7(a), FY23–25, n=4,506, Live Oak funded variable; 6 size bands)
| Size band | Sold spread | Held spread | Within-lender gap |
|---|---|---|---|
| $0-$150K | P+3.17% | P+2.10% | +107 bps |
| $150K-$350K | P+2.86% | P+1.90% | +96 bps |
| $350K-$500K | P+2.36% | P+1.67% | +69 bps |
| $500K-$1M | P+1.45% | P+0.61% | +84 bps |
| $1M-$2M | P+1.27% | P+0.51% | +76 bps |
| $2M-$5M | P+1.31% | P+0.23% | +108 bps |
Live Oak is priced to sell on the sold side. The premium bid from the secondary market sets the floor, so the quote a sold Live Oak borrower sees tracks what pool investors will pay, not what Live Oak’s deposits cost. Live Oak is the canary in this report: a deposit-funded bank that’s been routing its book further down the flipper spectrum year after year. The 65- to 106-basis-point gap is the visible edge of that drift.
Third. Newtek Bank NA prices to the ceiling. Sold loans clear at exactly P+3.00%median across all six size bands — to the basis point. That’s the secondary-market cap. Above P+3.5% the bid caps out and the incentive to sell collapses; Newtek sits right below that line and doesn’t try to push it. What ends up in the held book is the residual — sub-$150K loans at P+5.14% that pool investors wouldn’t bid at 110+. Live Oak runs two priced books. Newtek runs one flat price and lets the market decide what clears. Two fingerprints, same reason: the secondary market sets the ceiling.
Fourth.FA$TRK sits outside this market. FY25 variable 7(a) sale rate: Guaranty 61.9% (n=37,904), FA$TRK 0.7% (n=18,382). FA$TRK loans run smaller (~$150K vs ~$500K+) and concentrate in two deposit-funded banks — Huntington (19,765 FY23–25 FA$TRK variable, 0.8% sold) and M&T (5,223, 0.2% sold). Pool investors don’t want tiny-loan single-obligor tranches. Guaranty is the secondary market. FA$TRK is a separate economy using the same program stamp. Under the FA$TRK threshold, the premium-pricing framework doesn’t apply.
Fifth. The aggregate flip rate climbed from 37.4% in FY23 to 47.7% in FY25. Ten points in two years, concentrated in the sub-$150K band where nonbank SBLCs scaled (Readycap, Lendistry, BayFirst, Celtic, Northeast all running 95%+ sub-$150K flip rates). The rise is nonbank share. If Lendistry, Readycap, or BayFirst charter up as banks and get access to deposits, the bifurcation compresses and the flip rate ticks down. If they don’t, it keeps climbing. (7(a), FY23 and FY25, n=91,847, funded variable; FY23 n=42,275 / FY25 n=49,572)
Read together: the market pays for coupon, caps at 60%, and absorbs whatever nonbank SBLCs push into it. Portfolio lenders hold the rest. Live Oak, Newtek, and Huntington all operate inside the same guarantee and end up in three different positions because their balance sheets are shaped differently. One borrower per term sheet.
Chart 6C
The flip rate climbed ten points in two years.
The FY22 → FY23 trough sits under the Fed’s fastest tightening. Investor premium bids compressed as spread-to-alternatives narrowed, and deposit-funded banks held more of what they wrote. FY24–25 is the reverse: nonbank SBLCs scaled origination, and every loan they wrote had to sell to fund the next one. The leading explanation for the ten-point rise is supply-side — nonbank share grew — with a demand-side assist as investor appetite recovered. The 2027 edition gets the second data point that tells us whether the climb continues.
The revenue model
$1.15 billion. Forty-six percent of Year-1 revenue.
FY25 up-front premium captured by originating lenders on 7(a) guaranteed-portion sales: ~$1.15 billion. Premium as a share of FY25 Year-1 industry lender revenue on the full 7(a) book: ~46 percent. Not a footnote. Not a side business. Roughly half of what the 7(a) lending industry earned in Year 1 of FY25 was gain-on-sale. (7(a), FY25, n=26,166, $12.46B sold guaranteed at vol-wtd ~109.2 per SBA 7(a) Premium Tables PDF)
For Newtek, Readycap, Lendistry, Northeast, BayFirst, and Celtic, gain-on-sale isn’t a revenue line. It’s the whole company. Premium runs past 70% of Year-1 revenue at all six. Their business model is originate-and-distribute: write the loan, sell it the week it funds, service a thin retained strip, book the gain, repeat. The lending part is a feeder. They are lead generators for a secondary market with a banking charter attached. Newtek prints this out loud in its 10-Q. The loan-level data adds the shape of the book the revenue comes from.
FY25 sector — Year-1 7(a) lender revenue
Newtek — FY2024 worked example
| FY | Sold guaranteed | Loans sold | Vol-wtd premium | Est. up-front premium |
|---|---|---|---|---|
| FY2021 | $10.82B | 13,101 | ~113.0 | ~$1.41B |
| FY2022 | $12.47B | 16,713 | ~111.5 | ~$1.43B |
| FY2023 | $9.25B | 15,815 | ~108.0 | ~$0.74B |
| FY2024 | $10.73B | 23,188 | ~108.3 | ~$0.89B |
| FY2025 | $12.46B | 26,166 | ~109.2 | ~$1.15B |
Chart 6E — rotating anomaly
In the second half of FY25, the flip rate on $1M+ loans fell twelve points in a single half.
Prior three years held H1/H2 parity within four points. FY25 breaks it: H1 62.0%, H2 49.9%, a 12.1 pp gap on $1M+ variable 7(a). Two candidate mechanisms, weighted differently. Rate-cycle inflection is the leading one — the Fed cut 100 bps from Sep24 to Dec24, and high-coupon paper is worth more on balance sheet if the investor bid is falling. Data recency (public release as of 2025-12-31, late-FY25 originations have had less time to sell) explains 2–3 points based on prior-year H1/H2 differentials. The remaining nine points is a behavioral shift. If FY26 H1 closes back toward parity, recency won. If the flip rate stays compressed, lenders decided to retain high-coupon paper through the cut cycle.
FY2022
Gap -3.1 pp
FY2023
Gap -3.6 pp
FY2024
Gap +3.2 pp
FY2025
Gap +12.1 pp
Top-12 flippers — FY23–25 funded variable 7(a)
Ranked by flip rate with n ≥ 50 floor. Spread and size are raw cohort averages across the FY23–25 funded-variable pool. (7(a), FY23–25, n=41,512, funded variable; n ≥ 50)
| # | Lender | Type | n | Flip % | Avg spread | Avg size |
|---|---|---|---|---|---|---|
| 1 | Lendistry SBLC, LLC | nonbank | 2,843 | 99.9% | P+4.44% | $205K |
| 2 | Northeast Bank | bank | 5,180 | 99.1% | P+2.97% | $137K |
| 3 | Readycap Lending, LLC | nonbank | 6,704 | 98.8% | P+4.48% | $366K |
| 4 | BayFirst National Bank | bank | 6,947 | 98.6% | P+4.29% | $158K |
| 5 | Newtek Small Business Finance, Inc.† | nonbank | 549 | 98.5% | P+3.28% | $582K |
| 6 | Harvest Small Business Finance, LLC† | nonbank | 961 | 95.7% | P+2.51% | $500K |
| 7 | Celtic Bank Corporation | bank | 2,917 | 92.7% | P+2.84% | $290K |
| 8 | SouthState Bank, N.A. | bank | 1,063 | 84.7% | P+2.59% | $1.07M |
| 9 | Newtek Bank, National Association† | bank | 5,685 | 74.8% | P+3.01% | $420K |
| 10 | Byline Bank | bank | 1,161 | 71.7% | P+2.67% | $793K |
| 11 | Live Oak Banking Company† | bank | 4,506 | 64.6% | P+1.92% | $1.37M |
| 12 | KeyBank, N.A. | bank | 2,106 | 30.7% | P+2.29% | $648K |
Top-12 holders — FY23–25 funded variable 7(a)
Every row below sold zero loans or sub-1% of its FY23–25 funded-variable pool. Money-center franchises run small-ticket Express programs as acquisition tools. Regionals run larger-ticket 7(a) as a relationship product. Same 0%-flip bucket, two strategies. (7(a), FY23–25, n=31,306, funded variable; flip ≤ 0.2%; n ≥ 50)
| Lender | HQ | n | Flip % | Avg rate | Avg size |
|---|---|---|---|---|---|
| The Huntington National Bank† | OH | 16,406 | 0.9% | 10.22% | $205K |
| TD Bank, National Association | NY | 7,975 | 0.0% | 11.61% | $99K |
| Manufacturers and Traders Trust Company | NY | 4,499 | 0.2% | 11.89% | $128K |
| JPMorgan Chase Bank, National Association | CA | 3,885 | 0.0% | 12.00% | $205K |
| Wells Fargo Bank National Association | CA | 3,485 | 0.0% | 13.42% | $28K |
| U.S. Bank, National Association | CA | 2,896 | 0.0% | 10.68% | $208K |
| Zions Bank, A Division of | UT | 1,780 | 0.0% | 10.71% | $221K |
| First Bank of the Lake† | CA | 1,419 | 4.4% | 11.13% | $681K |
| Bank of America, National Association | CA | 1,018 | 0.0% | 9.76% | $338K |
| PNC Bank, National Association | TX | 962 | 0.0% | 12.22% | $64K |
| Banco Popular de Puerto Rico | PR | 846 | 0.0% | 12.03% | $130K |
| Comerica Bank | TX | 488 | 0.0% | 9.17% | $586K |
Cohort summary — bimodal distribution
Lenders with n ≥ 50 funded variable 7(a) in FY23–25, grouped by flip-rate cohort. Pure flippers price 64 bps higher than hold-to-maturity peers on average and originate at 2.2× the mean loan size. The middle is the thinnest cohort — the bimodal thesis holds.
| Cohort (flip %) | Lenders | Total loans | Mean rate | Mean spread | Mean avg size |
|---|---|---|---|---|---|
| 0-5% (hold-to-maturity) | 89 | 54,571 | 9.88% | P+1.908% | $456K |
| 5-50% (mixed) | 39 | 12,651 | 9.88% | P+1.930% | $626K |
| 50-90% (mostly flip) | 99 | 29,995 | 10.19% | P+2.244% | $941K |
| 90%+ (pure flipper) | 44 | 34,967 | 10.46% | P+2.547% | $1.02M |
The two-book structure
Newtek, Live Oak, Columbia — three routing fingerprints.
Newtek Bank NA’s 74.8% aggregate flip rate is the volume-weighted average of two book shapes. Sold cohort prices P+3.00% flat across every band. Unsold cohort holds the sub-$150K tail — 1,419 loans at an average P+5.14% that the secondary market would not bid at 110+.
| Newtek size band (FY23–25, n=5,685) | Flip % | Sold spread | Unsold spread | Rate gap (pp) |
|---|---|---|---|---|
| $0-$150K | 60.6% | P+3.01% | P+5.49% | +2.48 |
| $150K-$350K | 82.5% | P+3.01% | P+3.88% | +0.87 |
| $350K-$500K | 91.0% | P+3.01% | P+3.02% | +0.01 |
| $500K-$1M | 87.5% | P+3.00% | P+3.05% | +0.05 |
| $1M-$2M | 86.7% | P+3.01% | P+3.04% | +0.03 |
| $2M-$5M | 86.0% | P+3.01% | P+2.96% | -0.05 |
Newtek’s aggregate 74.8% is half an Express book (52.5% flip) and half an above-$350K standard book (75–85% flip). Same charter. Two books.
Columbia Bank shows the same shape at smaller scale: 12.8% overall flip (n=1,347), sold loans average $1.07M at 10.03%, retained loans average $69K at 12.51%. The sold book is a million-dollar-plus acquisition product; the retained book is a small-ticket working-capital product the pool market wouldn’t take. Same line item. Two products.
Live Oak mirrors Newtek’s fingerprint. 72.9% of its 4,502 funded variable-rate loans sit below cluster median (per Section 4 Chart 6); 38% sit below cluster p25. Volume-weighted, Live Oak’s book is below the 7(a) median even with a middling 64.6% aggregate flip rate. Three shapes visible: Newtek’s flat ceiling, Live Oak’s left-skewed distribution, Columbia’s two-product split.
Section 2 handoff
The league table hides three business models.
Nonbank SBLCs sell 97% of variable-rate 7(a) (n=12,933). Banks sell 38% (n=129,869). CDCs operating 7(a) sell 52% (n=164). Same guarantee, three different balance-sheet constraints. See the H/F badge on Table 2B.
Section 4 handoff
Where the expensive half of the dispersion ends up.
Section 4 Panel A reports 220 bps of p10–p90 spread dispersion on comparable deals. 52.8% of FY25 variable 7(a) dollars ($14.7B) flowed through the secondary market, and the sold cohort prices 51–77 bps above the held cohort in every band. The dispersion maps to the sale channel.
Section 5 handoff
Add one question to the scorecard.
“Does this lender sell more than 90% of its variable-rate 7(a) book over the last three fiscal years?” If yes, your rate was priced against pool investors’ yield demand — not the lender’s cost of funds. Their “cost of capital” argument doesn’t apply to you. Shop a holder for a comparison quote. FA$TRK borrowers sit outside the framework.
Methodology & data
- Source data
- SBA’s public 7(a) lending disclosure, loan-level, through 2025-12-31. Secondary-market flag is binary Y / blank. No loan-level premium is published in any SBA release. Premium distributions come from the SBA 1086 Premium Heat Maps PDF (gross-loan bands × price bands × FY). Any premium-per-loan stat is a model output over the aggregate distribution, shipped with its estimation sidebar.
- Par sold on a 1086
- Equals
gross_approval × guaranty_pct.gross_amountalone overstates by 25–50% (the retained unguaranteed portion). Small Loans (85% guaranty) and the Feb–Sep FY21 90% window (Economic Aid Act, expired Oct 1 2021) change the ratio; we use the actualguaranty_pct. - PPP excluded
- Separate 7(a) authority, CARES Act 100% guaranty, never sold on 1086s. The heat-map dataset filters PPP at source.
- Settlement vs approval windows
- PDF counts FY25 1086 settlements at 26,166 loans / $12.46B par. Our loan-level approval-cohort counts variable-rate funded 7(a) with sold-flag set at 23,929 loans FY25 / ~$12.2B guaranteed. Both are correct under their windows — PDF = settlement date, loan-level = approval date. Expected gap 5–10%; ~9% observed.
- What the data does not show
- Sale date is not published at loan level — we can flag which loans sold but not when. Per-loan premium is not published. Broker-channel flag is not in the public release at loan level (Form 159 is filed but not published). Lender concentration of FY25 sold volume is not separately disclosed.
- Full methodology
- Methodology index carries the full rule set, regulatory citations (13 CFR §§ 120.600–660; § 120.612; § 120.223; § 120.425), and the SHARED-METHODOLOGY cross-reconciliation table.
Open questions for next year
Inaugural edition. Same five questions next year. The chart structures freeze. The answers update.
- Does the nonbank 97% flip rate compress as nonbanks build bank charters? Newtek Bank NA’s 74.8% is already lower than Newtek Small Business Finance’s 98.5%. If Lendistry, Readycap, BayFirst acquire bank charters in the next five years, the bank/nonbank flip split converges toward 70–80% and the bimodal distribution collapses. Chart 6A reprises annually. A visible compression is the story.
- Lender concentration of FY25 sold volume. Top-10 sell-everything lenders likely account for >60% of the $14.7B sold-variable flow. Candidate for a Section 2 league-table cut (channel vs. balance-sheet lender-type) and a Chart 6D expansion next edition.
- Does the sold-cohort spread gap (Chart 6B) widen or compress when rates fall? FY21–25 shows the gap widened from ~6 bps in FY21–22 to ~75 bps in FY24–25. If FY26 shows compression under easing rates, the 46% premium share of Year-1 revenue shrinks mechanically.
- FY25 H2 collapse causality. Rate-cycle inflection vs. recency artifact. FY26 H1’s number tells us which.
- Permanent ask of SBA, annual until resolved: publish Form 1086 settlement data at the line-item level, including premium price by originating lender. Form 1086 is already filed and settled through Guidehouse; publication is a disclosure-policy choice, not a data-collection constraint. Until then, the $1.15B sector figure carries its ±15% model-output caveat and lender-level premium stays at distribution-aggregate level.