Headway, Alma or Grow Therapy — or hold your own contracts

The question, and the fact that answers it for half of you

You want insurance clients, and you have found four ways to get them. Three are companies that hold the payer contract for you — Headway, Alma, Grow Therapy. The fourth is doing it yourself: applying to each payer, signing each contract, filing each claim, and chasing each denial. The companies are easier. The question is what easier costs, and nobody publishes the answer, so this page works it out.

Before any of that, the fact that ends the page for a large share of the people reading it. All three networks require full, independent licensure. Headway’s published list of accepted credentials for California is CNS, LCSW, LMFT, LPCC, MD/DO and NP — there is no AMFT, no ASW, no APCC and no registered psychological associate on it.^[1] Alma states it plainly: “At this time, Alma is only available to clinicians who are fully licensed.”^[3] Grow Therapy’s California list is LCSW, LPCC, LMFT, licensed clinical psychologist, PMHNP and psychiatrist,^[4] and its provider page requires “an active, unrestricted license in your state”.^[5]

There is one exception and it does not help you. Headway runs a supervisory-billing pilot that lets a fully licensed clinician bill for a provisionally licensed one. As of 21 July 2026 it covers group practices in New York and Texas only, with a short list of payers, and solo providers are not eligible.^[2] California is not in it.

So if you are an associate, none of this is available to you, and the honest advice is to stop reading and go and count hours instead — the 3,000-hour calculator will tell you which gate is actually holding you up. Your route to insurance income before licensure runs through an employer or supervisor who already holds the contracts and bills under them. That is a real route and it is the only one.

Before the comparison, what this site earns

This site carries an affiliate link to Headway. If you join through it, the site is paid a referral fee; the amount is not disclosed to me by the program, and it costs you nothing, because Headway takes its margin from the session rate whether you arrived through a link or typed the address in. The site earns nothing at all from Alma, nothing from Grow Therapy, and obviously nothing from you holding your own contracts. Every arrangement is listed at the affiliate disclosure, which is generated from a single file so a paid link cannot exist on the site without appearing there.

The research below was done, and its conclusions reached, before any of that mattered — and you can check that from the results rather than taking my word for it. On the sharpest number on this page, Alma wins. On the biggest one, holding your own contracts wins. Headway wins one row of the decision table and loses several. A comparison that landed on the one option the author is paid for would be worth nothing to you, and you would be right to assume the worst.

The four routes, and what each one takes

Start with the structure, because the feature lists all look the same and the structures do not. What differs is where the money comes out.

Headway charges you nothing. “Sign up for free with no membership fees,” says its provider page.^[6] It holds the payer contract, and it takes its margin inside the rate: “We keep a small percentage of session payments to support our operations — things like credentialing, claims, and provider support. This amount varies across health plans and billing codes.”^[8] How small is never said. Headway’s own explainer adds that “the amount we’re able to keep depends on the strength of the rates we’re able to negotiate”.^[7] You see your rate per payer and per CPT code only after credentialing, inside the provider portal.^[8]

Alma charges a flat membership fee, and it is the only one of the four with a price you can read before you commit: $95 a month billed annually, $1,140 a year.^[9] A third-party review puts the month-to-month price at $125.^[10] Alma is explicit that it takes nothing from your cash-pay work — asked whether it takes a cut of visits with cash-pay clients, its FAQ answers “No… The income from your cash-pay clients stays with you.”^[3] On insurance work it says less. You are credentialed and paid under Alma’s tax ID, which grants “access to enhanced payback rates”,^[11] and nowhere does Alma state whether it keeps any part of what the payer pays. Treat the $1,140 as a cost you can see sitting on top of one you cannot, not as a substitute for it.

Grow Therapy is the one where the common description is wrong, including the one I started with. It is widely said to take a percentage of each session. It publishes no percentage anywhere. Its own guide says only that “providers on Grow are not employees and do not pay subscription fees”,^[13] its payout FAQ describes timing and says nothing about a fee,^[16] and the fullest third-party account puts it this way: “Grow Therapy is completely free for practitioners to join… Rather than charge practices a fee to join, Grow Therapy makes its money through its negotiations with insurance payors.”^[12] That is the same shape as Headway, not a different one. An undisclosed spread, described as free.

Your own contracts take nothing at all. You apply to each payer, you sign the contract, the payer pays you the contracted amount, and no third party is in the payment. What you carry instead is the work: credentialing, re-attestation, claim submission, and every denial. The panels page sets out that process in order, with the statutory clock California puts on it.

Two things fall out of that table, and both are worth more than a feature comparison. The first is that there are not three fee models here. There are two. Three of the four routes are the same shape — an intermediary holds the contract, pays you a rate it sets, and keeps an amount it does not disclose — and only one of the three also charges you a fee you can read. The second is that the rate column is empty. No platform in this comparison publishes what it pays a California therapist, for any code, at any license level. That is the finding rather than a gap in the research: the only way to learn your California rate on any of the three is to finish credentialing and read it off the inside of the portal.^[8][5]

One number gets quoted as if it filled that gap, and it does not. Alma’s provider page carries an interactive return-on-investment widget whose “average payout per session” control runs from $50 to $250.^[9] The $50 is the bottom of a slider in a marketing calculator. It is not a rate, it is not California, and it is not a promise. If you see it cited as Alma’s rate, the person citing it has not opened the page. Grow’s widely repeated “$21 a session” is a client cost, not a payout.

The arithmetic, at ten, twenty and thirty sessions a week

Since no network publishes a rate, the comparison cannot be run forwards. It has to be run backwards: work out what holding your own contract leaves you, then ask what each network would have to pay you per session to match it. That number you can take into a conversation, because it is the only figure in this whole area you can establish before you sign anything.

Here are the assumptions, stated openly, all of them arguable and all of them yours to change. The contracted rate is $134.47. That is the 2026 Medicare allowed amount for CPT 90837, the 53-minute individual session, for an LMFT, LCSW or LPCC in the Los Angeles locality — relative value units times geographic indices times the $33.4009 conversion factor,^[26] then multiplied by the 75% these professions are paid under statute.^[27] It is used here not because Medicare is the panel you would join but because it is the only California in-network rate that is published to the cent and checkable by you; commercial behavioral health contracts are commonly written as a percentage of it. The rates page shows the whole derivation and the same figure for five other codes and three localities. Substitute your own offer when you have one.

Forty-six working weeks. Your own time is worth $200 an hour, which is mid-range for a licensed LMFT or LCSW in private practice in Los Angeles — see the rate research for where that range comes from. That figure is doing real work here: the unbilled hours of a self-managed panel are not free, they are hours you could have billed, and pricing them at zero is how every other comparison of this decision reaches the wrong answer.

The unbilled hours. The panels page builds these from scratch and reaches about 90 hours in year one at 20 sessions a week: six payer applications at five hours each is 30 hours, three CAQH re-attestations on the 120-day cycle is 2 hours, and claims work is the rest. The claims part scales with your caseload and the other 32 hours do not, so it has to be rebuilt at each volume. At the 19% in-network denial rate KFF measured across HealthCare.gov issuers in 2024,^[25] and 20 minutes to work each denial, 460 sessions produce 29 hours of claims work, 920 produce 58, and 1,380 produce 87. Add the fixed 32 and you get 61, 90 and 119 hours. Joining a network is not zero either — call it 4 hours of onboarding, the same estimate the panels page uses.

Read the last three rows and nothing else, if you read nothing else. At 20 sessions a week, $115.77 is the number a network has to put in front of you before it is worth more than doing it yourself — and against a $134.47 contract, that means the network can keep at most 13.9% of what the payer pays. If it keeps more than that, you are subsidising it. If Headway or Grow offers you $110 a session, the answer is no, however good the software is.

Now the crossover between the two fee structures, which is sharper than anyone admits. Alma’s $1,140 divided across 920 sessions is $1.24 a session. That is the entire cost difference between the flat-fee route and the free-to-join route, before either one’s undisclosed cut. Against a $134.47 rate, $1.24 is 0.92%. So the flat fee is the cheaper structure the moment a no-fee network keeps more than about nine-tenths of one percent of your rate. Headway describes what it keeps as “a small percentage”.^[8] A small percentage that comes to less than 0.92% would be a remarkable thing, and it is not what an intermediary funding credentialing, claims and support out of the spread can plausibly be running on.

The same crossover at the other two volumes: $2.48 a session, or 1.84%, at ten a week; $0.83, or 0.61%, at thirty. The flat fee gets cheaper the busier you are, which is the ordinary behavior of a fixed cost and the reason a high-volume insurance practice should look hard at Alma before it looks at anything free.

There is a second, less obvious result in that table, and it runs against the way this decision is usually framed. The break-even rises with your caseload. A network has to beat $109.69 to be worth it at ten sessions a week and $117.80 at thirty, because the 32 fixed hours of credentialing get spread across more sessions as you get busier while the gross grows in a straight line. The busier you are, the better self-management looks, and the harder a network is to justify. If you are starting from an empty week, a network is at its most defensible; if you are full, it is at its least.

One more thing this arithmetic lets you do, which is to test a marketing claim. Headway’s provider page says you will “earn an average of $27K more per year compared to credentialing on your own”, with no footnote, no methodology, no sample and no date.^[6] Put it through the table. At 920 sessions, moving from 90 unbilled hours to 4 saves $17,200 of your time. For the total gap to be $27,000, Headway would have to pay you the remaining $9,800 in rate — which is $10.65 a session more than the contracted amount it is splitting with itself, whatever that amount happens to be, while also keeping a percentage of it. The claim survives only if your unbilled hour is worth $314 rather than $200 and the network keeps nothing. It is not a lie so much as a number with no stated baseline, and a number with no stated baseline is not evidence.

None of this is your arithmetic until you put your own figures in it. Change the rate to the one you have actually been offered, the weeks to the ones you actually work, and the $200 to your own fee, in the practice simulator. And note what the table does not contain: private pay. If the real problem is empty hours rather than a low rate, every column here is the wrong conversation.

What you are actually buying, since it is not the rate

A network sells four things. Each is worth something. None of them is worth an unlimited share of your rate, and it is worth pricing them one at a time.

Credentialing speed, which is the one you should discount hardest. The marketing numbers are Grow at “5–7 days on average”,^[5] Headway at “as few as 30 days”,^[6] Alma at “45 days or less”.^[9] Each softens in the company’s own detail pages. Headway’s help center says it “can take anywhere from 3 weeks to 4 months”.^[14] Alma’s says credentialing “can take up to 45 days” only once the completed onboarding form is in.^[11] And Grow’s 5–7 days is its own internal step, not payer enrollment; the fullest independent account reports Grow stating 4–6 months on average to be fully credentialed and billable with any given payer.^[12] Against that, California law already gives you a 60-day statutory clock on a completed mental health credentialing application, with a seven-business-day completeness acknowledgment,^[24] and Blue Shield of California publishes 45 to 60 days for a complete packet.^[28] The gap between the network route and the direct route on speed is real but it is weeks, not years.

Payment whether or not the payer pays. This one is genuine and it is the strongest thing any of them offer. Headway: “No matter what the negotiation yields, we still pay Headway providers on time every two weeks — whether or not we’ve been reimbursed by the insurance company.”^[8] Grow guarantees insurance payouts even where collection fails.^[16] Read the boundary, though. Headway’s marketing says “we’ll protect you from clawbacks… when you get paid, you stay paid”;^[6] its audit article says it aims to protect you from clawbacks and reserves the right to adjust payment where a provider does not respond to documentation requests.^[19] A guarantee with a condition on your responsiveness is a good deal and it is not an unconditional one.

Denial risk, transferred. This is the same thing priced differently, and it is the largest line in the table above. At a 19% in-network denial rate,^[25] 920 sessions generate roughly 175 claims a year that somebody has to work, which is 58 hours, which is $11,600 of your time. That is what you are handing over, and it is a real transfer. It is also why the break-even is $115.77 and not $134.47.

Referral flow, which is the only one you cannot buy anywhere else. A directory listing on a platform that clients search is worth something a payer directory is not, and it arrives faster than any marketing you could do. Headway’s California page lists 8,544 providers in the state and gives its payer mix — Aetna 98%, Carelon 79%, Cigna 76%, Anthem Blue Cross 75%, Blue Shield of California 65%.^[29] If your week is empty, that flow is the entire case for a network, and it is a good case. If your week is full, you are paying for it anyway.

And cash flow, which people forget to compare. Grow pays out every Friday, arriving in one to three business days.^[16] Alma offers a weekly or biweekly schedule.^[9] Headway markets “consistent, biweekly payments”^[6] but actually pays on the 15th and the last day of the month,^[15] which is semi-monthly — 24 payments a year, not 26. Small, but it is the difference between a claim and a fact, and this is a page about which is which.

The risks, and the dates they landed on

Your rate can be changed without your agreement, and it has been. On 24 and 31 October 2024, Headway and Alma notified providers of Optum-driven cuts effective 1 January and 1 December respectively. Reported reductions ran from a few dollars to $43 a visit, or 30%, concentrated on 90834 and 90837; one psychologist’s rate fell from $144.27 to $103.00, and one clinician reported a $28,000 annual loss. Headway said fewer than 340 of its 40,000-plus providers were affected, and told them their new rates “reflect exactly what we’re paid from Optum directly, meaning Headway will make $0 on your Optum appointments”.^[20] That last sentence is the most informative thing Headway has ever said about its own margin: it is $0 only in the exceptional case, which means it is not $0 in the ordinary one.

The Aetna change that hit Alma on 15 August 2026. On 20 May 2026 Aetna announced that for clinicians on Alma it would pay 90837 at the 90834 rate, pay 99215 at the 99214 rate, and eliminate the pay differential between master’s-level clinicians and psychologists. The original date was 15 July.^[21] Alma negotiated, and on 8 July secured a partial rollback: 90837 and 90834 would not be consolidated, 99214 and 99215 would not be consolidated, and reimbursement would continue to reflect degree type. The cuts still landed, on 15 August 2026. Published state-level figures show 90791 down 9.7% and 90837 down 10.6% in New York, and 90791 down 16.8% and 90837 down 15.2% in New Jersey, with 90832 unchanged in both.^[22] No California figures have been published, by Aetna or by Alma. If Aetna is a material part of your caseload, that is a question to put in writing before you pay for a year of membership.

Note what that episode actually demonstrates, because it cuts both ways. A network negotiated on behalf of thousands of clinicians and got a real concession that no solo therapist would have got alone. And a network was also the unit the payer chose to cut, all at once, on a date it set. Both are consequences of the same structure.

Clients you meet on the platform are not portable. This is the risk that costs the most and gets discussed the least. Headway’s own account article says clients who find you through Headway’s search “will continue to be seen on Headway”, and that it does not expect providers to build a caseload from Headway referrals and then migrate it off platform.^[17] Grow is blunter: “Referrals provided by Grow Therapy must be managed and billed through the Grow platform as outlined in your Provider Agreement,” and providers who deviate “will receive warnings, and multiple violations will terminate your partnership with Grow”.^[18] Alma publishes no equivalent rule that I could find, which is not the same as not having one — read your own agreement. Whichever you join, the clinical consequence is the same: a caseload built on platform referrals is a caseload you cannot take with you, and a therapeutic relationship should not be ended by a contract you signed before you met the person.

Your notes are on somebody else’s system. Headway tells providers to download progress notes before terminating an account, because it refers records requests back to the provider.^[17] Whatever you join, put a quarterly export in your calendar on the day you sign up, not on the day you leave.

Credentialing does not travel, and it can shadow what you already have. Headway states that an existing direct credentialing status with a payer “cannot be applied to or honored through Headway’s credentialing process” — you complete its process regardless.^[14] So a therapist on both routes with the same carrier holds two separate contracts at two different rates. That is worth knowing before you are surprised by it, and it is worth looking yourself up in each payer’s public directory afterwards to check which contact details are showing.

And the concentration risk, which is the one nobody prices. If every payer you bill sits behind one company, then that company’s rate decisions, its contract renegotiations, its policy changes and its corporate direction are your income. Headway began requiring government-ID photos and facial scans through a third-party vendor for clients and providers, announced 3 April 2026 and rolled out from June, with no opt-out short of leaving.^[30] Whatever you think of that on the merits, it is a decision made about your practice by somebody who is not you.

The largest survey of this question found the trade-off is genuinely mixed rather than one-sided. The Psychotherapy Action Network surveyed 667 mental health professionals in spring 2025: 52% reported higher income using these platforms, 18% about the same, and 32% somewhat or significantly less. 84% said they had not been made aware of the platforms’ fee-splitting arrangement, 71% of users did not know who owned or had invested in the platform they used, and 85% said they would not use one if they knew an insurer owned or part-owned it.^[23] Half of clinicians doing better is a real result and it belongs in this page. So does the 84%.

Which route, for which reader

The honest version of this decision is that it is decided by your situation and not by the products, and that four different therapists should give four different answers. Find your row.

Two rows there recommend a company this site earns nothing from, and three recommend doing it yourself, which earns nothing from anybody. That is what the arithmetic produced. If it had come out the other way I would have had a harder page to write and you would have had a worse one to read.

What to do on Monday

The decision does not have to be made in the abstract, and it should not be. Four steps, in order, and the first three cost you nothing.

• Work out your own break-even before you talk to anyone. Take the rate you would hold directly, multiply by the sessions you will actually book, subtract your unbilled hours at your own fee, add back four hours of platform onboarding, and divide. Mine came to $115.77 at 20 sessions a week on a $134.47 contract. Yours will differ, and it is the only number that lets you evaluate an offer.
• Apply, and get the rate in writing before you commit to anything. All three will credential you before you have seen a rate, and none of them will show you one before that. That is not a reason not to apply — applying is free and reversible — but it is a reason to treat the application as information-gathering rather than a decision. Ask for the full schedule by payer and by CPT code, not a headline number.
• Compare the offer against your break-even, not against your fee. A $118 offer against a $200 private-pay fee looks terrible and may be right. The same offer against a $115.77 break-even is the actual comparison, and it is close.
• Ask the three questions that decide the next five years, not the next quarter. What happens to clients I meet through your directory if I leave? What notice do I get of a rate change, in writing? What share of what the payer pays do you keep? You will get a clear answer to the first, a vague one to the second, and no answer at all to the third — and the shape of those three answers tells you more than any comparison table, this one included.

And keep the routes separate in your head, because they are not exclusive. You can hold two direct contracts and sit on one network. You can run a network for two years while you build a private-pay caseload and then leave. The thing to avoid is the position where one company sits between you and every payer you have, because that is the one arrangement with no exit that does not cost you clients. If the real problem is empty hours rather than a low rate, the referral arithmetic is the cheaper fix and it is worth trying first.