Best way to start an advertising agency compared, and the media money problem that makes it a different business from a marketing agency

Start an advertising agency on a fee, not on a percentage of media, and do not touch client media money through your own bank account until you have a reason to. That is the short answer, and the reason is cash rather than ethics. An advertising agency is distinguished from a marketing agency by one thing: it places paid media on behalf of clients. The moment you do that, you are standing between a client who pays on terms and a platform that charges immediately, and more new agencies die of that gap than of bad creative.

If what you are building is a services shop that sells strategy, content and search, the general version of this decision is a different page: the four models and the first service worth selling covers how to shape a marketing agency and how to price it. This page is about the part that is specific to advertising, which is the money and the credentials.

The four ways an advertising agency gets paid, compared on cash

One, a flat management fee. You charge a monthly fee for running the account, the client pays the platforms directly with their own card or billing account. Your invoice is a service invoice like any other. Cash positive, no exposure, no float, and it is the only model we would start on. The ceiling is that your revenue does not grow when the client's spend does, which founders mind more than they should at the beginning.

Two, a percentage of media spend. The traditional model. Your fee scales with the budget, which sounds excellent and introduces an incentive the client can see, so expect it to be negotiated and expect to justify it. Cash is fine as long as the media is still billed to the client, and dangerous the moment it is not.

Three, you front the media. You hold the billing relationship with the platforms, you pay them, and you invoice the client afterwards. This is where the business becomes a lending operation with an advertising department attached. A client on thirty day terms with a spend of fifty thousand dollars a month means you are carrying fifty thousand dollars of somebody else's cost, every month, permanently, and a single late payer can end the company while every account is performing well. If you do this at all, do it after the business is established, on written terms, with a credit limit per client that you set and enforce.

Four, production and creative on a project fee. Not media at all. Deliverable-based, checkable, paid on milestones. The safest revenue in this list and the easiest to start selling tomorrow.

The pattern across all four: the model is chosen by who carries the float, and a new agency should not be the answer.

The criterion for your own first offer, and it is the same one your clients will use

One test settles what to sell first: can the client verify, on their own, that you did the thing. A campaign that ran, with a report pulled from the client's own platform account, is checkable. A landing page that did not exist last month is checkable. A brand strategy is not checkable. An impression count on a slide you made is not checkable either, because you made the slide.

This is also why point one above matters more than it looks: when the media is billed to the client, the client's own platform billing is the independent record of what you spent. You have handed them verification for free, and verification is what a first client has no other way to get.

The credential nobody budgets for, with the published numbers

Platform partner status is the thing new agency founders assume is a formality and it is not. Google publishes its requirements and we read them on support.google.com on 21 October 2026. To be a Google Partner, a manager account "needs to have a minimum optimization score of 70%", it "needs to maintain a 90-day ad spend of $10,000 USD across managed accounts", and you need "a minimum of 50% of your account strategists certified in Google Ads (capped at 100 users)", "with at least one certification in each product area with campaign spend of $500 USD or more in 90 days". Premier Partner status additionally requires being "in the top 3% of participating companies within a given country (determined annually)". And "The Google Partner requirements are checked on a daily basis."

Read what that means for a founder. The $10,000 of 90-day managed spend is a floor you cannot reach without clients, so the badge is not available to you on day one and should not be in your pitch. The daily check means it is not a certificate you earn once; lose a client and you can lose the status. The 50% certification rule is a staffing constraint that arrives the moment you hire your second strategist. None of this is a reason not to start. It is a reason not to promise it in month one.

We also tried to read Meta's partner directory the same day and got a login wall, which served us a notice in Gujarati because our requests leave from an Indian IP address. So we have no Meta requirement to quote, and rather than repeat a figure from somewhere else we are telling you we could not read it. Check it yourself from your own market before you plan around it.

The first six months, in the order that keeps you solvent

• Decide the billing model and write it into the contract. Who pays the platform, on what day, and what happens if the client is late. This clause is the business.
• Sell a project before you sell a retainer. A paid project is a reference, a cash injection and a test of whether you like the client, and it costs them less to say yes to.
• Put every account in the client's name. Ad accounts, analytics, domain, billing profiles. Created by them, access granted to you. Agencies that do the opposite think they are retaining clients and are actually creating the dispute that ends the relationship.
• Set a credit limit before you ever front media. A number, per client, in writing, that you will not exceed. Decide it while you are calm.
• Price your delivery cost honestly. Your own tool and labour cost per account, calculated before you quote. Underpricing delivery closes more new agencies than losing pitches does.

How clients will shortlist you, which is not how you expect

New agency founders prepare for a credentials conversation and get a proximity one. Buyers shortlist on who they have heard of, who somebody recommended, and who is nearby, in roughly that order, and the strongest of those three is the weakest predictor of good work. Our page on why distance is the weakest criterion on the shortlist is worth reading from the other side of the table, because the arguments a buyer should use against proximity are the arguments you will need to make. We measured our own internal linking on 21 October 2026 and 1,015 of our 2,598 live pages carry no durable inbound link; that page is one of them.

The practical version: get one checkable result, write down how it was measured and over what dates, and lead with the method rather than the number. A founder who can say where a figure came from beats a deck of percentages, especially in front of a buyer who has been burned once.

The service nobody in your market is selling yet

Here is a genuine opening for a new agency, and it is the reason this page names us at all.

Every incumbent agency reports on search positions, impressions, clicks and conversions. None of them reports on what an assistant says. When a buyer asks ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews who to use in a category, the answer is composed for that person on the spot and names a handful of companies, and there is no position to report, no impression to count and nothing in anybody's media report about it. Clients are starting to ask the question and their current agency has no answer.

AI Knows Us sells that one measurement and nothing else. We are not an advertising agency, we do not place media, and we are not competing with the business you are about to start; we are the instrument you could put in a pitch that nobody else in the room has. We ask your client's buyer question in all five apps on a schedule and record which companies were named. From our pricing page read 21 October 2026, one credit is one buyer question on one app once, and "A Claude check uses 3, because Claude costs us more to run." The cheapest plan that reads all five surfaces is Growth at $299 a month with 100 buyer questions watched and 1,200 scheduled answers a month. One credit is one buyer question, asked on one AI app, once, and a Claude check uses 3; the pricing page sets Growth at 1,200 scheduled answers a month. The $149 tier is SEO Start and carries no AI answer checks at all, so do not plan a client offer around that price.

Two things to know before you build an agency offer on us, because they are against our interest. First, our own pricing page contradicts itself on exactly the thing an agency needs: it renders "More than one business" as No for all eight plans, while a separate row on the same page lists 10 businesses covered for Agency at $1,299. Both lines render. If you intend to run several client brands, settle that with us in writing before you buy, and do not take either line as the answer. Second, where a rival beats us: Profound, read at tryprofound.com/pricing on 21 October 2026, publishes "Up to 9 Answer Engines tracked" on Enterprise against the five surfaces we read, "SSO/SAML + SOC 2 compliance", and "Both plans include unlimited seats", and it publishes an "Agency Growth plan" with "400 credits per month per client workspace", though no price for it. Searching our own pricing page's HTML on word boundaries the same day returned no SSO, no SAML, no SOC 2, no ISO, no white label and no DPA. An agency reselling a measurement to its own clients should weigh unlimited seats and per-client workspaces seriously, and today Profound publishes both where we do not.

Questions about starting an advertising agency

How much money do I need to start

On a fee model where clients pay their own platforms, very little beyond your own living costs and tooling. On a model where you front media, you need working capital equal to your largest client's monthly spend plus their payment terms, held in reserve, permanently. The gap between those two answers is the whole decision.

Do I need to be a Google Partner to pitch

No, and you cannot be at the start. The published requirement includes a 90-day managed spend of $10,000 USD, so the status follows clients rather than preceding them. Pitch your method and your certifications instead, and add the badge when the spend arrives.

Should I specialise or stay general

Specialise, on either one channel or one industry, because a new agency's only real advantage is knowing something the generalist does not. Pick the specialism you can prove with one checkable result, not the one with the biggest market.

What goes wrong most often in year one

Fronting media for a client who pays late, and underpricing delivery. Both are decisions made in week one that only become visible in month six, which is why the contract clause and the delivery cost calculation are the first two tasks on this page rather than the last.

Is an advertising agency still worth starting

Yes, and the reason is that the measurement layer is changing. The clients most willing to move are the ones whose current agency cannot answer a new question, and there is a new question on the table that almost nobody is answering yet. A new agency has no legacy reporting to defend, which for once is the advantage.