Choosing a Funding Model That Doesn't Reward the Already Connected

Every year, millions in grant money goes to the same organizations. The ones with polished websites, board members who know board members, and a track record that screams “safe bet.” But safe bets rarely change the world. They reward the already connected.

If you’re reading this, you probably suspect your funding model is part of the problem. Maybe you’ve seen the same faces show up in every RFP. Or you’ve heard the same complaint: “We don’t have the connections.” This piece is for you. We’ll look at why the usual approach fails, what a fairer model looks like, and how to actually pull it off without breaking your budget or your board’s trust.

Why the usual funding model is broken

The Matthew Effect in philanthropy

Money follows money. That sounds obvious until you watch a community-run health clinic in a rural county lose a grant because they couldn't afford a professional grant writer. Meanwhile, the well-endowed university hospital down the interstate—already sitting on a $40 million endowment—hires a former foundation officer to ghostwrite their proposal. Same application pool. Different starting lines. This is the Matthew Effect in philanthropy: to those who already have, more is given. And it's poison for equitable service access.

The problem isn't just that the already-connected win more often. It's that the funding model designs for that outcome. Grant committees prize polished narratives, established track records, and partnerships with known institutions—all proxies for 'safe investment.' But safe investment rarely reaches the neighbors who need service access most. Quick reality check—the organisations serving isolated communities are often too busy delivering care to curate a glossy annual report. They lose before they've applied.

'The same charity that fed my grandmother now can't get a laptop grant because they don't have a Twitter following.'

— field director at a tribal health coalition, overheard at a 2023 rural funding summit

How networks create invisible barriers

Most funding decisions happen inside closed loops. Board members recommend their friends. Program officers circulate drafts among previous grantees. Nobody is malicious—they're just busy. But the effect is a narrow pipeline: new applicants from unfamiliar geographies or underequipped networks rarely get a first look. I have seen a promising mobile clinic in eastern Kentucky rejected three times for a 'lack of organisational capacity' while the same funder renewed a city-based group whose overhead sucked 40% of every dollar. Capacity, in practice, meant 'who you know.'

The tricky part is that this bias feels invisible to the people inside it. A funder reviews a proposal from a well-connected urban nonprofit and sees professionalism. They review a proposal from a community land trust in the Mississippi Delta and see 'gaps.' Wrong frame. The gaps are in the system, not the applicant. What usually breaks first is trust—communities stop applying because they've learned the game is rigged. They're not wrong.

Fragments matter here. A single lost grant can shutter a food pantry for six months. A denied after-school programme means kids walk empty streets until dark. That hurts.

The cost of homogeneity

When every funded organisation looks alike—same zip codes, same board structures, same donor histories—the services they offer converge. Homogeneous funding produces homogeneous access. Rural dialysis slots shrink. Interpreter services for refugee families get cut first. The people most disconnected from the formal economy stay disconnected. I once watched a foundation reject a brilliant community doula programme because the lead applicant didn't have a graduate degree. She had delivered 400 babies. That counted for nothing in the eligibility matrix.

The rhetorical question nobody asks aloud: If your funding model consistently rewards the same zip codes, is it funding need or funding convenience? Most foundations don't have an answer. They have a repeat button.

One pitfall worth naming—this isn't a call to abolish experience requirements. Experience matters. But the current model conflates experience with access to elite networks. A farmer co-op that has fed a county for thirty years has deep experience; they just lack a grant-writing consultant. The difference between 'experienced' and 'connected' is a gap the usual funding model refuses to see. And until that gap is named, equitable service access remains a talking point, not a practice.

What a fairer funding model actually does

Core principles: anonymity, simplicity, transparency

A fairer funding model starts by stripping away the very things that make the existing system a club for insiders. No personal connections. No whispered endorsements. Instead, three pillars hold the whole thing up. Anonymity means the reviewer has no idea who submitted the proposal—no alma mater, no LinkedIn glow, no shared Slack history. Simplicity means the application fits on two screens, not a twenty-page deck that only a grant-writing consultant can parse. Transparency means every rejection comes with a specific, public reason, not a form letter.

The tricky part is that anonymity alone doesn't fix bias—it just moves the problem. If you strip names but keep a "tell us about your previous awards" field, you've rebuilt the same gatekeeping. I have seen a pilot where the team accidentally left a URL field in the form; reviewers immediately started clicking through to judge the application by the applicant's website design. That hurts. The discipline required is almost surgical: remove every cue that signals existing privilege.

Not every social checklist earns its ink.

Not every social checklist earns its ink.

Shifting focus from pedigree to potential

What replaces the old signals? Raw proposal quality, measured by clarity and feasibility rather than reputation. The model asks: does this solve a real problem, and can this person actually execute it within the budget? Not: did they go to a top-tier accelerator? Not: do they know our board member?

'We stopped asking for references. We stopped asking for press mentions. The only thing that mattered was the plan and the track record of delivering similar work—verified independently.'

— Operations lead, anonymous community fund, 2024

That sounds fine until you realize the edge case: a first-time applicant with zero track record. The catch is that "potential" can become just as subjective as "pedigree" if you're not careful. We fixed this by requiring a mini-prototype or a letter from a verified end-user—not a reference from a powerful person. A grandmother who needs the service counts as much as a university dean.

Examples of criteria that reduce bias

Concrete criteria matter more than philosophical statements. What actually works? Randomized review order—applications are shuffled so the tenth submission gets the same attention as the first. Cap on past funding—any applicant who has already received more than $50k in grants is scored down on a "novelty" axis. Demographic-blind scoring rubrics—reviewers tick boxes for technical soundness, community impact, and budget realism, and nothing else. Wrong order? You bet. If you score "brand recognition" first, you bleed bias into every subsequent category.

Most teams skip this: they announce a fair funding model but keep the old review committee. That's the seam that blows out. The committee itself must be rotated regularly and drawn from the community the funding serves—not from the usual donor class. One concrete anecdote: a local arts fund replaced five tenured professors on its panel with three bar owners, two retired nurses, and a teenager who runs a poetry zine. Returns spiked in year two—not because the bar owners were generous, but because they spotted projects the professors had waved past for a decade.

How it works under the hood

Designing a blind review process

The trickiest part is removing the very information that usually biases funding decisions. We built a two-stage gate: at entry, applicants submit a stripped project description — no team bios, no institutional logos, no past grant history. Only the community problem, the proposed solution, and a cost breakdown. A separate system holds the identity data under encryption, released only after the score is locked. I have seen this fail in one pilot where a reviewer recognized a writing style from a previous submission — so we now require all text to pass through a style-neutralizer that strips idiosyncratic phrasing. The catch? You lose context. A project from a well-known nonprofit with a shaky track record gets the same blind look as a first-time applicant from a remote village. That's the point — but it means the review panel sometimes funds naive proposals that lack operational reality.

Weighting criteria to favor reach over reputation

Most grant rubrics reward prestige: "Has the team published in peer-reviewed journals?" or "How many years of experience?" Our model flips that — we assign 60% weight to *potential reach per dollar* and 20% to *geographic or demographic marginalization*. Reputation gets zero points. Quick reality check — this creates a perverse incentive. Some applicants inflate their user numbers because they know we check reach. So we built a verification layer: instead of self-reported metrics, we require a pre-registration of target communities with local partner organizations. That adds friction, but it stops the gaming. The trade-off is slower onboarding — we lose about 15% of applicants who can't produce third-party validation.

Building in feedback loops for continuous improvement

What usually breaks first is the criteria themselves. After three funding cycles, we noticed our weighting system favored projects in densely populated urban slums over dispersed rural areas — simply because "reach per dollar" looks better when people are stacked on top of each other. We fixed this by adding a density-adjusted multiplier. That's not elegant, but it works. The feedback loop is ugly: every rejected applicant can request a de-identified copy of their scores and the panel’s comments. We then run quarterly audits checking if certain demographics are systematically underfunded. One audit showed that female-led projects scored higher on marginalization but lower on perceived feasibility — so we retrained reviewers on implicit bias. Does it eliminate bias entirely? No. But the loop means we catch the worst distortions within six months, not six years.

“Blind review doesn't guarantee fairness — it guarantees that the same old biases can't hide behind a familiar name.”

— operational lead from a community foundation that adopted a similar model

That said, the feedback loop has a dark side. Over-correction. After the rural density fix, we started seeing proposals that artificially dispersed their target population just to game the multiplier. We now run random spot-check audits on 10% of funded projects, and if we catch misrepresentation, the entire cohort’s criteria get recalibrated. Wrong order? Maybe. But the alternative — trusting a static rubric — is how the already connected stay connected.

A walkthrough: The Community Roots Grant

Setting up the grant: eligibility and outreach

The Community Roots Grant starts with a painful admission: most grants reward people who already know how to write grants. We fixed this by scrapping the application portal entirely. Instead, we sent three field workers—paid, not volunteers—into two census tracts where internet adoption hovered below 40%. Their job wasn't to sell anything. They knocked on doors, asked what people needed to get online, and listened. The tricky part was trust: one resident assumed the whole thing was a data-collection scam. We handed her a paper flyer in her first language, no QR code required. Eligibility was dead simple: household income under 200% of the federal poverty line, no prior fiber subscription. No essay. No references.

Outreach ran for six weeks. We posted in laundromats, on community bulletin boards, and—oddly effective—at the counter of a local bodega where the owner vouched for us. The grant pool was $180,000, which covered hardware, installation, and twelve months of service for roughly 200 households. That sounds fine until you do the math—$900 per household, which means no fancy routers. We used refurbished units from a university surplus sale. Ugly? Yes. Functional? Mostly.

Application and review: step-by-step

No online form. The field workers carried paper intake sheets—two sides, ten questions. Name, address, number of children in school, current internet situation. One question asked: What would change for your family with reliable internet? We didn't care about grammar. One man wrote kids stop crying about homework. That got approved. The review committee was three people: a local librarian, a retired teacher from the neighborhood, and a grants manager who lived two miles away. They met in the library basement, sorted applications into three piles—yes, no, maybe—in two hours. The whole review cost $1,200 in stipends. Not fancy. Fast.

Flag this for social: shortcuts cost a day.

Flag this for social: shortcuts cost a day.

What usually breaks first is the maybe pile. We gave each maybe a second look: did the applicant have a child in remote school? Was someone in the household job-hunting? If yes, they moved to approved. That heuristic felt crude, but it kept the process under a week. One family slipped through the maybe-to-yes crack—turns out the father was an overnight janitor who needed Wi-Fi to schedule shifts. The committee didn't catch it. A neighbor mentioned it at a block meeting. We called them the next morning. Good instincts beat perfect rules.

Results and lessons learned

After six months, 187 of the 200 households were still using the service. Thirteen dropped out: two moved, three couldn't afford the electricity to power the router (a pitfall we missed), and eight found the speed too slow for video calls. That last one hurt. We had promised enough for school and work, but the refurbished routers maxed out at 25 Mbps. For a family with three kids and a parent on Zoom? Not enough. We should have tested the hardware in multi-device homes—one laptop in a quiet room is not the real world.

The biggest lesson, though, was about referrals. We didn't ask for them. But once the grant was public, neighbors started walking people over to the field workers. By week four, half of the applications came from word-of-mouth, not our outreach. My cousin got it, now I want it. That's the kind of distribution money can't buy—but it only works if the initial pool of grantees actually trusts the program. We almost broke that trust with slow routers. Next time, I'll pay more for hardware and less for logo-branded flyers.

‘We spent two hours debating whether to buy nicer routers or nicer flyers. The routers lost. That was a mistake.’

— Field coordinator, post-grant debrief

So what's the takeaway for someone building their own version? Keep eligibility binary. Keep review fast. And for heaven's sake, test the gear in a real house with three devices streaming at once. The already-connected don't have that problem—they already know what works. Everyone else is stuck guessing. The Community Roots Grant guessed wrong on hardware but right on trust. That trade-off, ugly as it's, still got 187 homes online.

When it doesn’t work: edge cases

High-trust vs. low-trust environments

Blind allocation works beautifully where people already believe the system is fair—but drop it into a place where trust has eroded, and the whole thing backfires. I once watched a community fund in a mid-sized city collapse because two factions each assumed the other was gaming the "random" selector. They weren't. But the perception of rigging was enough. In low-trust settings, connection-agnostic models need a transparent audit trail—public draw seeds, verifiable timestamps, a witness from each stakeholder group. Without that, you don't get equity; you get conspiracy theories instead of grant applications.

The contrast is stark. A cooperative in rural Vermont ran the same blind draw for three years, and nobody blinked. Same mechanic, different social fabric. That sounds fine until you realize we can't rebuild trust just by choosing a clever algorithm. The funding model can't fix what the community broke. What it can do is refuse to make things worse—by not handing extra shots to people already holding the microphone.

Sectors where connections are essential

Not every field can afford to ignore networks. In clinical trial recruitment, for example, "who you know" is often a proxy for "who is alive and willing to enroll." A blind lottery for a rare disease grant could accidentally exclude the very patients a clinic has spent years building rapport with. The trick is distinguishing between access (which we want to equalize) and contextual knowledge (which we should preserve). One mitigation: cap the referral advantage rather than ban it entirely. Let a connected applicant get a modest bump—say, +10% weight—but not a guaranteed win. Otherwise, the model becomes ideologically pure but practically useless. I have seen a brilliant cancer screening project lose funding to an obscure applicant who had zero domain experience. Technically fair. Ethically messy.

'We swapped one kind of gatekeeping for another: randomness replaced relationships, but nobody asked whether the randomness was kind.'

— Lead evaluator, after a failed blind allocation trial in medical research

Coping with fraud and low-quality applications

Here is the most uncomfortable edge case: when you remove the network filter, the noise floods in. A connected ecosystem self-policed bad actors partly through reputation—but blind systems lack that immune system. We fixed this by adding a lightweight pre-screen: three yes/no eligibility checks, no essay required. Not a barrier, just a sieve. That stopped the automated spam while letting genuine but unconnected applicants through. The catch? Fraudsters adapt. One group submitted 47 identical applications under different names after they realized the draw was truly random. We responded by hashing applicant identities and capping submissions per IP range—but that penalized rural co-ops sharing a single internet connection. Wrong order. We later shifted to a proof-of-humanity check (a simple CAPTCHA variant) combined with a one-week cooling period. Fraud dropped; rural applications recovered. The lesson: blind models need thick boundaries, not open borders. Pure randomness without friction invites exploitation, and exploitation erodes the very fairness the model was built to deliver.

Limits of the approach

Scalability challenges

The honest truth: this model is harder to scale than a flat fee. I have watched teams try to roll it out across a hundred distinct communities—and the administrative load multiplies fast. Each local council, each microloan officer, each community liaison needs training. Not just “here’s the form” training; they need judgment calls about what counts as “already connected.” That judgment is expensive. And slow. A city-wide program that processes 5,000 applications in a week becomes a 90-day bottleneck when every case demands a human chat about actual need. The friction is real—and if your organisation lacks the staff or trust networks, the model stalls.

There is also a quiet problem of perception. Even when distribution is fairer, the appearance of complexity can scare off early adopters. Quick reality check—one pilot I observed lost half its potential applicants in the first month because the intake form asked for “proof of social bandwidth” alongside income data. People dropped out, suspicious of the extra step. The trade-off is brutal: you gain equity but you lose reach, at least in the short term.

Resistance from established players

Incumbents don't love this approach—and they have good reason not to. If your current business model rewards the already-connected (referral bonuses, legacy discounts, network-tier pricing), a model that deliberately starves those same channels looks like an attack. It's. One telecom executive told me, “You’re basically asking me to stop feeding my most profitable customer segment.” He wasn’t wrong. The catch is that those profitable segments often got their advantage not through merit but through existing infrastructure—better credit scores from richer parents, earlier broadband access, geographic luck. Asking them to subsidise newcomers feels like a tax. And they will lobby, delay, or simply refuse to participate.

That resistance creates a second-order limit: voluntary adoption rarely reaches critical mass. Unless a regulator or a major funder mandates the shift, the model becomes a niche experiment. I have seen promising pilots die because the anchor institution that controlled the purse strings pulled out after pressure from legacy partners. Not a conspiracy—just inertia dressed as efficiency.

Reality check: name the services owner or stop.

Reality check: name the services owner or stop.

Trade-offs between speed and equity

Here is the raw edge: fast deployment almost always unevenly benefits the already-connected. We fixed this by slowing down intake—adding community interviews, cross-referencing social graphs, verifying offline ties. That slowdown hurt. A program that could have wired 200 homes in two weeks instead wired 60 homes in three weeks. The equity gain was measurable—those 60 homes had zero prior access, versus the 200 homes where 40% would have been partial upgrades. But tell that to a funder expecting quarterly metrics. “You delivered 30% of the target.” That sentence kills budgets.

The trade-off worsens when emergency response is needed. During a disaster, waiting to map who is “already connected” when you could just drop free hotspots feels indefensible. Wrong order. Yet the same emergency-drip approach that works in a crisis calcifies into permanent inequality when applied to routine access. The model has a blind spot: it can't pivot fast without sacrificing its core promise. A rhetorical question, then: should a framework that fails under extreme duress be used at all in calmer times? I think yes—but only if you admit upfront that speed is the second priority.

‘Equity without speed is a luxury; speed without equity is a repeat of the original problem.’

— Field note from a community organiser, rural broadband co-op, 2023

That quote sticks with me because it names the limit without excusing it. If you adopt this funding model, own the slowness. Build buffer time into your grant cycle. Teach your board that “fewer homes, deeper impact” is a defensible metric. Otherwise the trade-off will decide for you—and it usually chooses the already-connected.

Reader FAQ

How do I convince my board to try this?

Boards hate risk — especially when the current model keeps the lights on. The trick is to frame the shift not as charity but as portfolio diversification. Show them the grant that went to a suburban rec centre that already had three basketball courts, then ask: "What did that buy us?" Nothing. Equitable models buy reach. I have seen one board flip after I mapped their last twelve awards onto a zip-code heatmap — nine went to the same two postal codes. You don't need to overhaul everything; propose a pilot ring-fenced fund, maybe ten percent of your annual allocation, with a separate evaluation rubric. The catch is you must let the pilot fail gracefully — boards panic if you promise miracles. Frame the downside: worse case, you learn where your application pipeline actually leaks.

'We worried about 'lowering standards' until we realised the old standard was just geography and who could afford a grant writer.'

— Director of a regional arts council, after their first equitable round

What if we get too many applications?

That's a good problem — but it can crush your reviewers. The usual response (raise the bar) punishes the very groups you want. What usually breaks first is the eligibility screener. We fixed this by replacing the essay-length 'tell us your impact theory' with a two-minute video or a voicemail. Screenshots of a community WhatsApp group count as proof of reach. Yes, it feels messy. The trade-off is you trade polish for volume; plan for a triage phase where a paid community reader — not a program officer — flags applications from organisations under 50k annual budget. Wrong order: don't filter by format first. Let everyone in, then use a lightweight lottery for the final thirty slots. That sounds fine until a donor asks 'but is it fair?'. Fairer than the old system where only three orgs knew the password to the portal.

Can this work for large-scale government grants?

Harder, but not impossible. Government procurement rules often mandate lowest-price or most-points scoring, both of which favour incumbents with glossy budgets. The seam where you can insert equity is in the weighting of evaluation criteria. One state agency I advised moved 'prior community investment' from 5% to 20% of the score — suddenly a rural health co-op that had run a teen clinic for twelve years outranked a hospital chain that had never set foot in the county. The pitfall: auditors will ask for objective definitions of 'community investment.' You need verifiable proxies — years of service, number of local hires, not a vibe check. That said, large grants also attract political scrutiny. If a legislator complains, the fallback is to show the model's cost-neutrality. We're not spending more; we're spending differently. Not every edge case survives this test — when the grant is tied to federal highway funds, state law may lock the scoring formula. So start with discretionary pools, prove the data works, then push for rule changes.

Your next move: pull last year's award list and colour-code it by the distance from the applicant's address to the wealthiest census tract in your city. Show that map to one board member over coffee — not at a formal meeting. That conversation will tell you if the model has a shot.

Practical takeaways

Three steps to audit your current model

Grab your last funding round’s rubric. Not the mission statement—the actual scoring sheet. Step one: isolate every criterion that rewards network size, follower count, or existing infrastructure. Circle them. That number alone tells you whether your model is a gatekeeper dressed as a grant. Step two: run a mock round with ten anonymized applicants—strip names, social links, past grant history. Compare the results. If the finalist list flips entirely, your current process was selecting for brand, not potential. Step three: publish a one-page summary of who got funded and why—use categories like ‘technical feasibility’ and ‘underserved need,’ not ‘strong team track record.’ The catch is transparency shrinks the gap between stated values and actual outcomes. It also invites pushback. That’s the point.

One quick win: the anonymized first round

Implement this by next Tuesday. Strip identifiers before any human reads a proposal. I have seen a small foundation in rural Kenya use this move and see their applicant pool triple—four of six finalists were first-time applicants who had never passed the initial screen. The mechanism is dead simple: assign a random ID, block the evaluator’s view of the applicant’s name, LinkedIn, and past grants. Yes, reviewers grumble about losing context. That’s a feature. Context is how bias slips in. The trade-off? You might fund a weaker operational plan from an unknown team and miss a solid one from a known group—but the data says the pool’s average quality holds steady while equity spikes.

“We stopped funding the same five orgs three years running. Turned out the sixth was doing better work. We just never saw it.”

— program officer, anonymous foundation, 2023 field visit

Resources for going deeper

Most teams skip this: grab the Grantmaking Equity Toolkit from the Fund for Shared Insight—it’s free, it’s ugly, and it works. Pair it with one afternoon mapping your funding flows on paper. Draw circles for each type of org you fund, size them by dollar amount, then draw lines to the zip codes and communities they serve. Wrong order. The seam blows out when you realize 80% of your money lands in three postal codes where the median income is double the national average. One rhetorical question to leave you with: would your funding survive a blind test? If not, fix the round before the next deadline. That hurts. Do it anyway.