PastBehavior research note

How Companies Replaced a Broken Acquisition Channel

Six companies occupied the same structural position. Three lost a dominant discovery channel. Three built embedded distribution before they needed it. What happened reveals where replacement distribution actually works.

6 companies · 20 distribution surfaces · 4-part screen

The strongest replacement channels didn't find another audience. They found someone who could see the need first.

We study what happened before to make better decisions about what happens next.

01 · The structural position

The same problem keeps appearing.

A company whose customer acquisition ran predominantly through a discovery channel it did not own, where the channel operator changed the rules, and which then had to build replacement distribution.

The names and industries change.

The underlying position does not.

The channel broke

Demand Media / eHow
Google changed the economics of organic discovery.
Chegg
AI answers began absorbing question-shaped search demand.
Angi
Intermediated demand became increasingly expensive and low quality.

Embedded before the crisis

Affirm
Embedded financing directly into commerce infrastructure.
Trupanion
Built distribution inside veterinary workflows.
BILL
Distributed through accountants, financial institutions and software partners.

The three successful embedded channels were not built as emergency replacements. They were built before the companies needed them.

02 · Six cases

What each company actually did.

Case 01

Demand Media / eHow

Google changed organic discovery.

eHow represented approximately 30–31% of company revenue before Google's Panda changes materially affected search visibility.

Outcome — The channel never recovered to its former position.

Mechanism

There was no relationship underneath the traffic.

Evidence · High confidenceView evidence +

Position

eHow depended heavily on Google organic discovery.

eHow represented approximately 30–31% of company revenue in 2012–2013.

What happened

Google introduced the Panda algorithm change in 2011.

Third-party measurement services reported major declines in traffic and search visibility, including a 66% decline in eHow's Sistrix visibility index.

Response

Demand initially disputed the magnitude of the external measurements.

The company subsequently worked to improve content quality while increasingly relying on its separate domain registrar business.

Outcome

The stock fell approximately 38% over two weeks and never returned to its former position.

The registrar business was eventually separated and the remaining company became Leaf Group.

Mechanism

Every visit was intermediated. Every visitor was effectively anonymous. When the intermediary changed the rules, there was no second surface where the same demand naturally appeared.

Evidence base · High confidence

Dated algorithm releases, third-party traffic measurement, company statements and revenue concentration derived from filings.

Case 02

Chegg

AI answers began absorbing question-shaped search demand.

Non-subscriber traffic eventually fell approximately 49% year over year as Google's AI Overviews expanded.

Outcome — Revenue and subscribers fell roughly 30%, and Chegg began pursuing a different customer.

Mechanism

Chegg replaced the channel by changing the customer.

Evidence · High confidenceView evidence +

Position

Chegg's subscription study-help business acquired significant demand through Google organic results for question-shaped queries.

What happened

As Google's AI Overviews expanded, Chegg's global non-subscriber traffic fell progressively: −8% → −19% → −37% → approximately −49% YoY by January 2025.

Response

Launched CheggMate.

Sued Google.

Reduced headcount substantially.

Moved toward institutional and B2B skills products.

Outcome

Q1 2025 revenue fell roughly 30%. Subscribers also fell roughly 31%.

The company explored strategic alternatives.

Mechanism

Selling courses to institutions introduced a different buyer, a different sales motion, a different cycle length and different unit economics.

That may create a new business. It could not replace the old distribution engine on the timeline the existing business required.

Losing discovery is especially dangerous when the intermediary can also substitute for the underlying product.

Evidence base · High confidence

Quarterly traffic disclosures, earnings calls, financial results and restructuring announcements.

Case 03

Angi

Intermediated demand became increasingly expensive and low quality.

Angi deliberately reduced lower-quality demand and improved marketplace economics.

Outcome — Revenue continued declining while unit economics materially improved.

Mechanism

Angi improved the demand it had without yet creating a fundamentally different source of demand.

Evidence · High confidence on metrics · Medium confidence on interpretationView evidence +

Position

Angi acquired homeowner demand through paid search, SEO and third-party affiliate networks.

What happened

Angi deliberately began reducing what management called “empty-calorie” demand while search behavior itself continued changing.

Response

Shifted toward proprietary demand and reduced network demand.

Introduced Homeowner Choice.

Adopted an AI-first operating model.

Increased investment in television.

Outcome

Revenue continued declining, but unit economics materially improved.

Homeowner NPS moved from negative 32 into positive territory. Pro win rate increased 10%. New-pro acquisition declined 41%. Lifetime value per new pro increased 50%.

Mechanism

Angi improved the quality of the demand it had without yet creating a fundamentally different source of demand.

Television is still discovery. It is discovery purchased differently.

Counter-evidence

Angi is not a failure case. Its marketplace economics improved materially and management has guided toward renewed growth. The outcome remains unresolved.

Evidence base · High confidence on metrics · Medium confidence on interpretation

IAC and Angi earnings calls, shareholder letters, the 2025 10-K and subsequent results disclosures.

Case 04

Affirm

Closest positive precedent

Affirm embedded itself into checkout rather than trying to become a destination.

Active merchants ultimately grew from approximately 6,500 to 419,000.

Outcome — Platform distribution produced orders-of-magnitude expansion in merchant reach.

Mechanism

Sell once to the platform. Let the platform distribute.

Evidence · High confidenceView evidence +

Position

At IPO, Affirm disclosed that approximately 28% of revenue came from a single merchant: Peloton.

The concentration was different from search concentration, but structurally similar: someone else owned the customer moment.

Response

Affirm did not primarily try to become a destination. It attached itself to platforms already controlling checkout: Shopify, Amazon, Stripe and Worldpay.

Outcome

Active merchants grew from approximately 6,500 → 102,000 → 168,000 → 207,000 → 419,000.

By the quarter ended September 30, 2025, Affirm reported approximately $10.8B GMV and $933.3M revenue.

Mechanism

One integration could reach hundreds of thousands of merchants.

The partner also had its own operational reason to drive adoption, because installments could improve conversion and average order value.

Counter-evidence

Platform distribution creates its own dependency. Klarna displaced Affirm as Walmart's exclusive BNPL provider in 2025.

Penetration inside a won platform can also take years. Large merchant counts do not automatically mean high attach rates.

Evidence base · High confidence

SEC filings, earnings releases and company product documentation.

Case 05

Trupanion

Trupanion built distribution inside the veterinary workflow.

The company ultimately developed relationships with more than 8,500 veterinary hospitals.

Outcome — Distribution became attached to the moment the customer encountered the need.

Mechanism

The veterinarian sees the trigger before anyone else.

Evidence · High confidence on structure · Medium confidence on some performance claimsView evidence +

Position

Pet insurance operates in a category with low penetration and expensive direct consumer acquisition.

Response

Trupanion built distribution inside veterinary hospitals. By 2025 it had relationships with more than 8,500 veterinary hospitals, supported by more than 120 Territory Partners.

It also built Vet Portal, allowing hospitals to receive payment directly.

Mechanism

The valuable moment is not when someone sees an insurance advertisement. It is when the pet owner receives a treatment estimate.

Audience access ≠ trigger access. The veterinary clinic does not simply have an audience of pet owners; it observes the moment the financial need becomes real.

The hospital also benefits operationally: easier payment, reduced friction and potentially greater treatment acceptance.

Counter-evidence

Embedded distribution can inherit the regulatory surface of the partner. Trupanion's model attracted scrutiny around whether veterinarians and Territory Partners were acting as unlicensed insurance producers.

Evidence base · High confidence on structure · Medium confidence on some performance claims

SEC risk-factor disclosures, veterinary-facing material, Territory Partner documentation, a Darryl Rawlings interview and VIN News coverage of regulatory scrutiny.

Case 06

BILL

BILL distributed through institutions SMBs already trusted.

Its indirect motion included accounting firms, financial institutions and software companies.

Outcome — Partner distribution became an important component of a durable customer base.

Mechanism

Embed through institutions the customer already trusts.

Channel-sourced customers can behave differently from directly acquired customers.

Evidence · High confidenceView evidence +

Position

BILL sells SMB accounts payable and receivable software into a market that can be expensive to reach directly.

Response

BILL ran direct acquisition alongside indirect distribution through accounting firms, financial institutions and software companies — organizations already trusted by the SMB.

Outcome

For fiscal 2022, more than 82% of subscription and transaction revenue came from customers acquired before that fiscal year.

Mechanism

BILL disclosed customer retention excluding customers of financial-institution partners.

Channel-sourced customers can behave differently from directly acquired customers. A partnership channel can solve acquisition without producing identical retention or lifetime-value economics.

Evidence base · High confidence

Form 10-K for fiscal year ended June 30, 2022.

03 · What repeats

What repeats

Six companies. Five patterns.

01

Replacing discovery with different discovery does not work fast enough.

Demand Media remained dependent on being found. Chegg built its own AI product and eventually changed customers. Angi is investing more heavily in brand television. All three responses remain, in some form, dependent on generating discovery.

Angi's unit economics genuinely improved. The conclusion is not that discovery investment is useless. It is that different discovery is not the same thing as replacement distribution.

02

The channels that worked attached to a moment the partner could observe.

Affirm

Checkout

Trupanion

Treatment estimate

BILL

Accountant / bank workflow

Audience access ≠ Trigger access

A partner with millions of relevant customers gives you audience access. A partner whose own workflow detects that a customer has a specific unresolved need gives you trigger access.

Those are fundamentally different assets.

03

The partner's incentive was operational, not just financial.

The strongest partners had reasons to make the integration work beyond referral revenue. Affirm could improve commerce conversion and average order value. Trupanion could improve treatment and payment economics. BILL improved an existing SMB workflow.

The strongest distribution partners would still want the integration at zero referral revenue.

04

Speed came from integration leverage, not more deals.

One relationship → one audience

One integration → many customers

Both can produce real distribution. They scale on fundamentally different curves. Affirm's merchant count moved by orders of magnitude when platforms became the distribution layer.

05

Nobody in the cohort rebuilt a broken channel in four quarters.

Affirm signed Shopify in 2020 and was still building penetration years later. Trupanion spent years developing its veterinary network. BILL built indirect distribution alongside direct distribution from early in its development.

The three successful embedded channels were built before they were urgently needed.

A replacement channel may need to be judged initially on trigger quality, integration leverage and penetration — not simply near-term revenue contribution.

04 · The screen

The historical cases suggest a way to distinguish real replacement distribution from partnerships that simply provide more reach.

A four-part screen for replacement distribution

12 possible points. The goal isn't the biggest audience. It's the strongest combination of trigger, workflow, economics and leverage.

T1 · Trigger ownership

0–3

Does the partner observe the moment the need actually appears?

Highest score

The partner's own process stalls because the need is unmet.

T2 · Workflow adjacency

0–3

Is the product a natural step in something the customer is already completing?

Highest score

Declining the product leaves the customer's current task unfinished.

T3 · Partner economics

0–3

Does solving the problem improve the partner's own economics or operations?

Possible benefits: revenue, conversion, cost avoidance, reduced friction, greater completion.

Highest score

The partner would want the integration even at zero referral revenue.

T4 · Integration leverage

0–3

Can one integration reach many customers?

Highest score

Infrastructure, platforms or marketplaces with many downstream customers on one technical surface.

High T1 + Low T4 → Real trigger. Slow scale.

High T4 + Low T1 → Huge reach. Weak intent.

Most partnership decks optimize for reach. The historical winners optimized for presence at the moment of need.

05 · Applying the screen

Applying the screen: SMB legal & compliance

Where does someone already have the SMB when the legal need becomes real?

A framework is only useful if it changes where you look.

We applied the screen to SMB legal and compliance, a category where customer acquisition has historically depended heavily on capturing intent: someone searches for help forming a business, registering an entity, filing a trademark or resolving a compliance issue.

But many of those needs become visible somewhere else first.

A bank sees the missing formation documents. A lender discovers the entity isn't in good standing. A payroll platform sees the first out-of-state hire. A marketplace requires a trademark. A KYB provider finds the entity that cannot pass verification.

So we mapped 20 places where the legal need can appear before the customer goes looking for legal help.

20 SMB legal and compliance distribution surfaces scored against the PastBehavior screen. This is a category-level application of the historical framework, not a company-specific map.

Sort by
1KYB and business verification providers12

Middesk · Alloy · Persona · Baselayer

Verification fails because the entity is delinquent or unregistered.

Trigger 3 · Workflow 3 · Economics 3 · Leverage 3 strongest structural fit

2SMB lending and capital underwriting11

Bluevine · Fundbox · OnDeck · Shopify Capital · Amex Business Blueprint

Loan application blocked on good standing.

Trigger 3 · Workflow 3 · Economics 3 · Leverage 2 strongest structural fit

3Marketplace brand registries11

Amazon Brand Registry · Etsy · Walmart

Registry enrollment requires a registered trademark.

Trigger 3 · Workflow 3 · Economics 2 · Leverage 3 strongest structural fit

4E-commerce seller onboarding11

Shopify · Etsy · Amazon Seller Central · eBay

Seller signup collects entity and tax identity.

Trigger 3 · Workflow 3 · Economics 2 · Leverage 3 strongest structural fit

5Payments and merchant onboarding KYB11

Stripe · Square · Adyen

Merchant underwriting requires verified business identity.

Trigger 3 · Workflow 3 · Economics 2 · Leverage 3 strongest structural fit

6Franchisors and franchise brokers10

Franchise agreement mandates entity formation per unit.

Trigger 3 · Workflow 3 · Economics 3 · Leverage 1 strong

7Business banking onboarding10

Mercury · Novo · Relay · Bluevine · Found · Lili

Account opening blocked without EIN and formation documents.

Trigger 3 · Workflow 3 · Economics 2 · Leverage 2 strong

8Payroll and multi-state registration10

Gusto · Rippling · ADP · Paychex · OnPay

First hire in a new state triggers registration and foreign qualification.

Trigger 3 · Workflow 3 · Economics 2 · Leverage 2 strong

9Domain and website builders10

Wix · Squarespace · Hostinger · Namecheap

Name search and business launch.

Trigger 2 · Workflow 3 · Economics 2 · Leverage 3 strong

10Accounting and bookkeeping networks9

Intuit ProAdvisor · Xero advisors · Bench · Pilot · Padgett

The advisor sees bad standing before the owner does.

Trigger 3 · Workflow 2 · Economics 3 · Leverage 1 strong

11Business insurance underwriting9

Next · Hiscox · Coalition · Thimble

Underwriting checks entity status and finds a problem.

Trigger 3 · Workflow 2 · Economics 2 · Leverage 2 strong

12PEO and EOR9

TriNet · Justworks · Deel · Remote

Multi-state employment setup.

Trigger 3 · Workflow 2 · Economics 2 · Leverage 2 strong

13Vertical SaaS with licensing exposure9

Toast · ServiceTitan · Jobber · Mindbody · Housecall Pro

Trade licenses, DBAs and permits become relevant during onboarding or expansion.

Trigger 2 · Workflow 3 · Economics 2 · Leverage 2 strong

14Tax filing at the Schedule C moment8

TurboTax Self-Employed · H&R Block · Keeper

Self-employment tax prompts the entity or S-corp question.

Trigger 3 · Workflow 2 · Economics 1 · Leverage 2 moderate

15Gig and freelance platform onboarding8

Uber · DoorDash · Instacart · Upwork · Fiverr

Worker crosses from side income toward operating as a business.

Trigger 2 · Workflow 2 · Economics 1 · Leverage 3 moderate

16AI assistants and agent surfaces8

ChatGPT · Claude · Copilot · Perplexity

The question is asked, but the underlying action has not yet been completed.

Trigger 2 · Workflow 2 · Economics 1 · Leverage 3 moderate

17Business-for-sale marketplaces7

BizBuySell · Flippa · Acquire.com

Entity transfer or dissolution becomes necessary at close.

Trigger 3 · Workflow 2 · Economics 1 · Leverage 1 moderate

18Commercial leasing and flex space6

WeWork · Industrious · Regus · commercial brokers

Lease requires an entity or certificate of good standing.

Trigger 2 · Workflow 2 · Economics 1 · Leverage 1 audience-heavy

19Professional and job networks6

LinkedIn

Profile activity may signal intent, but no transaction is necessarily present.

Trigger 1 · Workflow 1 · Economics 1 · Leverage 3 audience-heavy

20Membership and affinity organizations5

AAA · chambers · NFIB · SBDCs and similar organizations

Member benefit placement rather than an observed transaction.

Trigger 1 · Workflow 1 · Economics 2 · Leverage 1 audience-heavy

Scores represent outside-in judgment, not measured conversion. T1 and T2 are closer to observable. T3 and T4 require more judgment. The ranking is more defensible than any individual point estimate.

Individual company onboarding behavior should be verified before acting on the map.

Where we'd look first

Four territories rise to the top.

01 ·Business verification

12/12

The partner detects the legal problem as part of its core product.

02 ·Lending

11/12

The legal or compliance problem can block access to capital.

03 ·Commerce

11/12

Entity or trademark status can block seller activation or marketplace participation. This territory combines marketplace brand registries and e-commerce seller onboarding.

04 ·Payments

11/12

Business verification is already embedded in merchant underwriting and onboarding.

These are not four companies to contact. They are four structural territories that rise to the top of the framework.

06 · What the map shows

The biggest audiences aren't at the top.

The highest-scoring surfaces are not necessarily the companies with the largest SMB audiences. They are the companies most likely to know why a business needs help right now.

Audience access

“We have small businesses you can market to.”

Trigger access

“We know which small businesses need you right now.”

The legal need often appears inside someone else's workflow first.

The distribution opportunity is to be there when it does.

That difference is the central finding of the map.

Lending, verification, payments, payroll and banking score highly because the legal/compliance problem can prevent the SMB from completing something it is already trying to do.

Professional networks and affinity organizations can provide enormous reach without necessarily observing the moment the need becomes real.

07 · Highest-scoring surface

Highest-scoring surface · 12/12

Business verification infrastructure

KYB and business-verification providers such as Middesk, Alloy, Persona and Baselayer routinely verify businesses for banks, marketplaces, fintechs and payments companies.

As part of that process, verification can expose situations where an entity is delinquent, dissolved, unregistered, missing required documentation or otherwise unable to pass verification.

Detecting the legal problem is already part of the partner's core product.

The SMB is not browsing for legal or compliance assistance. It is trying to open an account, receive financing, onboard to a marketplace, accept payments or complete another business transaction, and cannot proceed.

The legal need appears inside another company's workflow.

This is a strategic hypothesis produced by the screen, not proven conversion performance.

08 · Implications

What this suggests

01

Separate partnership types

Not all “partner-sourced” demand is the same. Platform integrations, workflow integrations, referral relationships and affinity distribution operate on fundamentally different scaling curves.

Combining them into one partnership metric can obscure which model is actually producing durable replacement distribution.

02

Optimize for trigger, not audience

Audience size is easy to measure. Trigger quality is more important.

The strongest distribution surfaces are those where another company's workflow exposes the customer's unresolved need.

03

Give the partner its own reason to care

Referral economics alone are weak alignment. The historical winners created operational value for the distribution partner itself.

Would the partner still want this integration if the referral fee were zero? If not, the distribution may remain permanently sub-optimized.

04

Measure what compounds

The cohort provides no evidence that a dominant discovery channel can be replaced through embedded distribution in four quarters.

Early channel measurement should therefore include integrations launched, downstream customer reach, trigger quality, activation inside each integration, attach rate, expansion potential and retention by acquisition source — not simply current-period revenue contribution.

09 · The decision

The replacement isn't another place to be discovered.

The historical failures remained dependent on attracting attention.

The historical successes inserted themselves into transactions and workflows that were already happening.

The question is not: Who has our audience?

Who knows the moment our customer needs us?

10 · Limits

What this does not prove

Selection bias
The cohort is weighted toward documented failures. Companies that quietly rebuilt distribution may be underrepresented.
The successful cases were not emergency replacements
Affirm, Trupanion and BILL built embedded distribution before a crisis. They demonstrate that embedded distribution can work, not that it can replace a broken channel quickly.
Angi remains unresolved
Its revenue declined while its unit economics materially improved. Calling the strategy a failure would be premature.
Platform dependency does not disappear
Embedded distribution can replace one concentration risk with another. Affirm's Walmart relationship illustrates renewal risk.
Regulatory exposure can travel through the channel
Trupanion demonstrates that embedding a regulated product in another party's workflow can create new regulatory questions.
Channel customers may behave differently
BILL's retention disclosure suggests partner-sourced customers should not automatically be modeled using direct-acquisition assumptions.
The scores are judgment
The 20-surface map is a structured outside-in assessment, not measured conversion data.

The research identifies where the historical evidence points. It does not establish that those surfaces will convert.

11 · Methodology

How we did this

Six companies occupying the same structural position were established from dated public sources. We compared what happened, what each company did next, and the mechanism behind the outcome. Those mechanisms were converted into a four-part distribution screen and applied to 20 SMB legal and compliance distribution surfaces.

  1. 6 historical cases
  2. 5 repeated patterns
  3. 4-part screen
  4. 20 distribution surfaces
  5. Implications

12 · Sources

Sources

Demand Media / eHow6 sources · View +
  • Google Search Central blog·Finding more high-quality sites in search (Panda release)·February 24, 2011
  • Google Search Central blog·High-quality sites algorithm goes global·April 11, 2011
  • Sistrix·Visibility index measurement for ehow.com·2011
  • Search Engine Land·Demand Media statement on Panda traffic impact·2011
  • Variety·Retrospective on Demand Media and the formation of Leaf Group·2016
  • Demand Media SEC filings·Annual reports on Form 10-K; eHow revenue concentration·Fiscal years 2012–2013
Chegg8 sources · View +
  • Chegg, Inc.·Q3 2024 earnings call transcript·November 2024
  • Chegg, Inc.·Q4 and full-year 2024 results·February 2025
  • Chegg, Inc.·Q1 2025 results and shareholder letter·May 2025
  • Chegg, Inc.·Q2 2025 results·August 2025
  • Chegg, Inc.·Restructuring and workforce reduction announcements·May and October 2025
  • The Washington Post·Coverage of Chegg's traffic decline and AI Overviews·2025
  • Higher Ed Dive·Coverage of Chegg's subscriber decline and pivot to institutions·2025
  • Fox Business·Coverage of Chegg's lawsuit against Google·February 2025
Angi6 sources · View +
  • IAC Inc.·Q2 2023 earnings call transcript·August 2023
  • Angi Inc.·Q4 2024 earnings call and shareholder letter·February 2025
  • Angi Inc.·Q1 2025 earnings call and shareholder letter·May 2025
  • Angi Inc.·Annual report on Form 10-K·Fiscal year 2025
  • Angi Inc.·Restructuring disclosure·January 2026
  • Angi Inc.·Q4 2025 results·February 2026
Affirm6 sources · View +
  • Affirm Holdings, Inc.·Registration statement on Form S-1 and IPO disclosures·November 2020 – January 2021
  • Affirm Holdings, Inc.·Q1 FY2022 earnings release·November 2021
  • Affirm Holdings, Inc.·Q3 FY2022 earnings release·May 2022
  • Affirm Holdings, Inc.·Q1 FY2026 results and shareholder letter (quarter ended September 30, 2025)·November 2025
  • Shopify·Shop Pay Installments product documentation·Accessed 2026
  • Third-party analyst commentary·BNPL merchant attach-rate estimates·2024–2025
Trupanion5 sources · View +
  • Trupanion, Inc.·Annual report on Form 10-K, risk-factor disclosures·Fiscal year 2024
  • Trupanion, Inc.·Veterinary-facing material and Vet Portal documentation·Accessed 2026
  • Trupanion, Inc.·Territory Partner program documentation·Accessed 2026
  • Interview with Darryl Rawlings, founder and CEO·Discussion of veterinary-channel distribution strategy·2023
  • VIN News Service·Coverage of regulatory scrutiny of pet insurance sales in clinics·2023–2024
BILL1 source · View +
  • BILL Holdings, Inc.·Annual report on Form 10-K; retention and channel disclosures·Fiscal year ended June 30, 2022
Market context1 source · View +
  • Public market and industry sources·Filings, earnings disclosures and trade coverage underlying the surface screen·2011–2026

Facing a problem that has happened before?

PastBehavior uses evidence from what has already happened to help companies decide what to do next — and execute against it.

Talk to PastBehavior →