Section 1
The evidence that you are a tenant
Start with what the platforms actually do, because the numbers describe a rental market, not an ad market. On Angi, a standard lead is sold to several contractors at once. Reported ranges run from three to eight businesses for the same homeowner, and in competitive trades like roofing operators report the same lead going to as many as sixteen. Every one of those contractors pays full price. Shared-lead conversion rates sit around 13 to 20 percent, against 27 to 30 percent for exclusive leads, so you are paying full freight for a fractional chance (LeadTruffle, 2026). The math is not subtle. A lead shared five ways is a lottery ticket priced like a booking. Google Local Services Ads looks more legitimate because it wears Google's badge, and it is a better channel than most. But it is still rent. Across roughly 888 contractors and 126,650 leads tracked in early 2026, the average cost per lead landed near 53 dollars, with HVAC in major metros running 45 to 80 (Searchlight Digital / BlueGrid, 2026). Here is the landlord tell: in 2024 Google removed manual lead disputes and replaced them with an automated credit system that returns roughly 6 to 7 percent of spend. In 2025 it stopped issuing credits for "job type not serviced" and "geo not serviced," so you now pay for leads you cannot even serve if your settings drift. In October 2025 it folded Google Guaranteed into a single "Google Verified" badge and quietly dropped the money-back guarantee that used to sit behind it (BGCollective, 2026). None of those changes were negotiated with you. They were posted, like a notice taped to a door. Bark runs the same model with the meter more visible. Leads go to multiple professionals. You buy credits at about 2.35 dollars each to unlock the chance to quote, and a single lead can cost 5 to 20 credits, so answering one inquiry can run past 40 dollars whether or not the customer ever replies. From November 2025, newly purchased credits expire three months after purchase, and the shift to a Marketplace subscription in late 2025 triggered a wave of complaints about credits vanishing and automated charges (Bark Help Center; sidehustles.com, 2026). Credits that expire are rent you paid for a room you never entered. Checkatrade and the other directories charge a membership on top, which is the cleanest possible statement of the relationship. You pay a recurring fee to keep your name on their wall. Stop paying and your name comes down, along with every review you earned that lives on their domain rather than yours. Read those four together and the category resolves into one sentence. The platform owns the demand. You rent access to it, per job, at a price it controls, alongside competitors it invites, under rules it can rewrite without asking.
Section 2
The framework: four models pointed at one trap
One lens will mislead you here. Game theory alone makes it look like a pricing fight you can win with discipline. Network analysis alone makes it look hopeless. Run four models together and you get both the mechanism and the exit. Model 1: Colonel Blotto, on how you allocate spend Blotto is the game of allocating a fixed force across several battlefields against an opponent doing the same. Whoever commits more to a given front wins that front. The counterintuitive result is that spreading evenly is usually a losing strategy. Concentration on chosen fronts beats thin presence everywhere. Your acquisition budget is the force. The battlefields are channels: rented aggregators, owned search, referrals, repeat customers, local reputation. Most operators spread their spend across every platform that will take their card, which is exactly the even allocation Blotto punishes. Worse, on a shared-lead front you are not fighting the platform, you are fighting the four to fifteen other tenants who bought the same lead, and the platform wins that battle no matter who among you takes the job. The winning Blotto move is to pull force off the fronts you cannot own and concentrate it on the one or two fronts where committing more actually compounds in your favor. Colonel Blotto (game theory): the allocation lens • Assumes: a roughly fixed acquisition budget you can move between channels. • Fits because: you have conflicting claimants competing for the same finite spend. • Breaks when: a channel has real switching costs or lock-in, so force is not freely reallocable in one quarter. • Counteracts: the reflex to be present everywhere. • May reinforce: over-concentration, if you dump every dollar into one owned channel before it can convert. Model 2: Mechanism design, on why the rules look like this Mechanism design runs the arrow backwards. Instead of asking what players will do given the rules, it asks how a designer sets the rules so that self-interested players produce the outcome the designer wants. The platform is the designer. Shared leads are not a bug or a capacity limitation. They are the mechanism. Selling one lead to five contractors multiplies revenue per lead by five while transferring all the conversion risk to you. The automated dispute systems, the credit expiries, the badge consolidations all point the same direction: maximize platform take, minimize platform obligation, keep the tenant unable to price the alternative. Once you see it as designed, you stop being angry about individual bad leads and start reading the incentive. The platform is optimizing its equilibrium. Your job is to design your own, where a customer who finds you once has a reason and a path to come straight back without the tollbooth. Mechanism design: the "why is it built this way" lens • Assumes: the platform is a rational designer optimizing its own take. • Fits because: the rules are engineered, not accidental, and they move one way over time. • Breaks when: a change is genuinely externally forced (regulation, a Google core update) rather than chosen, so reading intent misleads you. • Counteracts: treating each rule change as a random insult instead of a pattern. • May reinforce: conspiracy thinking, seeing deliberate design in what is sometimes just incompetence. Model 3: Network centrality, on who owns the demand node Model the market as nodes and edges. Customers are nodes. You are a node. The platform is a node. In the rented arrangement, almost every edge from a customer to you passes through the platform first. The platform is not the biggest node by revenue. It is the most central one, the intermediary that sits on the path between you and demand. Centrality, not size, is what makes it powerful. A node that controls the paths can tax every crossing, and can cut you off without losing the customer, because the customer's edge was always to the platform, never to you. The entire strategic goal reduces to one move in network terms: build direct edges from customers to your own node that do not route through the intermediary. A repeat customer who calls you directly is a direct edge. A referral is a customer minting a new direct edge on your behalf. Every direct edge lowers the platform's centrality over your book of business. Network centrality: the "who sits in the middle" lens • Assumes: power follows position on the demand path, not headcount or spend. • Fits because: the platform's leverage comes entirely from intermediating customer contact. • Breaks when: the platform is genuinely generating net-new demand you could never reach alone, so it is a source, not just a tollbooth. • Counteracts: the belief that spend equals control. • May reinforce: dismissing platforms that actually do expand your reach into new demand. The structure-break flag Here is the piece that makes the aggregator relationship worse than a normal supplier relationship. Every model above assumes a stable structure. Blotto assumes the fronts stay put. Mechanism design assumes the rules you are reading are the rules you will face. Centrality assumes the edges hold. The landlord can break all three overnight. Google removing manual disputes, killing geo credits, dropping the money-back guarantee. Bark expiring credits at 90 days and switching to a subscription. These are structure breaks, and they arrived as announcements, not negotiations. When you build a business on a channel whose rules can be rewritten without your consent, you are not exposed to market risk, which you can price. You are exposed to regime risk, which you cannot. Any plan that assumes today's aggregator terms will hold is already wrong. It just does not know when.
Section 3
The solution: GEER, then RADAR, then CHAIN
Three moves. Rank the levers, schedule the actions against uncertainty, then check the plan against history so you do not talk yourself into a fantasy. GEER: rank the levers Net exposure here is high and negative. The dominant channels of harm are price (rent rises), competition (shared leads), and rule risk (structure breaks). Map those to levers and pull the cheap, reversible ones first. 1. Capture contact on every rented job. The cheapest, highest-return lever. When a platform lead becomes a customer, get them onto a channel you own before the job ends: your own booking link, your phone, an email list, a service reminder. You already paid the rent. Extract the deed. This converts a one-time rental into a direct edge at near-zero marginal cost. 2. Stand up one owned channel you control end to end. A Google Business Profile you own, a fast site with your own booking flow, a review asset that lives on your domain. This is the front Blotto says to concentrate on, because spend here compounds instead of evaporating per lead. 3. Build the referral loop deliberately. Referred customers convert at higher rates, cost less to acquire, and carry 16 to 25 percent higher lifetime value than customers from other channels (Wharton, Schmitt/Skiera/Van den Bulte). A referral is a customer building your direct edge for free. Ninety-two percent of people trust a peer recommendation; 33 percent trust a banner ad. You are competing on the trusted channel. 4. Reprice the rented channels as pure overflow. Do not fire the aggregators on day one. Demote them. They become the marginal buyer of your idle capacity, used only when owned demand does not fill the calendar, and evaluated on true cost per booked job, not cost per lead. 5. Track owned share of bookings as your north-star. The single number that says whether you are becoming an owner or staying a tenant. Financing flag: levers 1, 3 and 5 cost time, not cash, and pull first. Lever 2 needs modest investment. None of this requires a war chest, which is the point. RADAR: schedule it against uncertainty You do not know when the next rule change lands, so build a dated portfolio instead of a forecast. • DO NOW (zero regret, reversible): Turn on contact capture for every job this week. Claim and complete your Google Business Profile. Ask your last twenty happy customers for a referral or a review on your own asset. These pay off under every scenario, including the one where the platforms behave. • HEDGE (cheap tail insurance): Export whatever customer data the platforms let you export, now, before a terms change locks it. Keep a running copy of your reviews off-platform. Diversify across at least two rented channels so no single landlord can end your month with one email. • DEFER + TRIGGER (irreversible, wait for a signal): Do not cut a profitable aggregator while it still books jobs. Pre-commit the trigger instead. If owned bookings cross 40 percent of volume, or if a platform raises effective cost per booked job past your defined ceiling, or if it changes dispute or pricing terms against you again, that is the signal to cut spend to that channel by a fixed step. Decide the number now, in calm weather, so you are not renegotiating with yourself under pressure. CHAIN: check it against history Match this to the right reference class by structure, not surface. The structure is: independent operators dependent on an intermediary that owns the demand and can reprice access. The base rate from that class is not kind to tenants. OpenTable and restaurants, Amazon and third-party sellers, DoorDash and kitchens, franchisors and franchisees, App Store developers and Apple. The recurring pattern is that the intermediary raises take over time, and the operators who survive with margin intact are the ones who used the platform to build a direct relationship and then shifted volume off it. The ones who stayed pure tenants got their rent raised until the business was working for the landlord. Present-state modifier: local service demand is geographically sticky and reputation-driven, which tilts the odds toward the operator more than in pure e-commerce. Your customers live near you and talk to their neighbors. That is a real edge the App Store developer never had. Use it. Matrix-break flag: the base rate assumes the platform keeps behaving like a rent-seeking intermediary. If a platform genuinely reinvents itself as a demand generator that reaches customers you never could, the reference class shifts and pure exit would be a mistake. Watch for that. It is rare, but it is the honest exception.
Section 4
What this framework cannot see
Three blind spots, stated plainly, because a framework that hides its limits is just a sales pitch. First, it cannot tell you the true incremental value of a platform. Some aggregator demand is net-new, customers you would never have reached, and some is customers who would have found you anyway. The models treat the platform as a tollbooth, but for a brand-new operator with no reputation and no list, it is also a starting gun. If you have zero owned demand, rented leads are not rent, they are seed capital, and cutting them early is malpractice. The framework is built for operators with something to own, not for week one. Second, it cannot price your own execution. Owned channels only compound if you actually run the follow-up, answer the phone, earn the review, deserve the referral. The deed is worthless if the house is a mess. Plenty of operators escape one landlord and simply fail to build anything, then blame the market. Third, it says nothing about demand collapse. If the underlying local market shrinks, owning your channel just means you own a smaller pie. This is a distribution-ownership framework, not a demand-creation one.
Section 5
The fitness test
Run one number this quarter. Of every job you booked, what share came from a customer you can reach again for free, without paying anyone for permission? Owned bookings over total bookings. If that number is flat or falling, you are still a tenant, and every dollar you call a marketing budget is rent. If it is climbing month over month, you are buying back the deed one job at a time, and the day a platform posts its next rule change on your door, it will be an annoyance instead of an emergency. Measure it. Then go take a channel back. --- Sources: LeadTruffle, "Complete Guide to Angi Leads" and "Angi Leads Cost 2026" (leadtruffle.co, 2026); Searchlight Digital, "Google LSA Cost Per Lead by Trade 2026" and BlueGrid Media LSA statistics 2026; BGCollective, "LSA Lead Credits: How Google's New System Works" (bgcollective.com, 2026); Bark.com Help Center, "What is a credit"; sidehustles.com Bark review 2026; Schmitt, Skiera, Van den Bulte, "Referral Programs and Customer Value" (Wharton, 2011). Conversion, credit and cost figures are platform- and trade-specific and move over time; treat them as order-of-magnitude, not fixed.