Startup marketing glossary

The terms that keep coming up in the essays and guides on this site, each defined the way the articles use it, with the number behind it where there is one and a link to the piece that explains it.

Budget and metrics

CAC (customer acquisition cost)

What it costs to win one new customer. The ads-only CAC that Facebook and Google report counts ad spend and stops. The fully loaded CAC adds the salaries of whoever runs the campaigns, the software they use, a share of the founder's own hours and, where you can price it, the opportunity cost. Most decks show the first and should show the second.

Formula Fully loaded CAC = (ad spend + campaign payroll + software + a share of founder hours) ÷ new customers

Incremental CAC

What the next customer costs you, as opposed to the average of the ones you already have. Facebook and Google report one blended cost that leans on your cheapest early buyers, so it cannot be measured cleanly. Raise spend on one platform, hold the rest flat and watch the blended cost: if it climbs, the low figure was demand you already had.

LTV (lifetime value)

What a customer is worth over their whole stay with you. One blended LTV hides customers who cost the same to win and behave nothing alike: at PadSplit, a renter who books a room and never moves in, and one who stays for years. Split the customers who stayed from the ones who left before you set LTV against CAC.

LTV to CAC ratio (LTV:CAC)

Lifetime value divided by the cost of winning the customer. The popular 3:1 target usually rests on a CAC counted short. Atticus LeBlanc of PadSplit puts a working business nearer 1:1 or 1.5:1 once labor, software and the founder's hours are in. A 1.5:1 you can defend means the business works on a thin margin; it is a harder number to raise on and a safer one to run on.

Formula LTV ÷ fully loaded CAC

Payback

How long a dollar of marketing spend takes to come back, and whether it comes back at all. At Artlist most revenue came from annual subscriptions, so a dollar spent came back several times over within the same month, and Ira Belsky kept raising the budget while that held. The moment to stop is when the next dollar stops coming back, not a date on the calendar.

ROAS (return on ad spend)

The revenue that comes back for every dollar spent on ads, often reported as marketing ROI. A very high number is a starting line, not a finish line. An early Artlist report showed 500 percent, five dollars back on every one, and Ira Belsky reads a return like that as proof of underspending: demand nobody has paid to reach yet. Keep raising spend until the return drops.

Formula ROAS = revenue from ads ÷ ad spend

MER (marketing efficiency ratio)

Total revenue divided by total ad spend across every platform together. Its partner, POAS, is profit divided by spend. Neither leans on any single platform's credit for a sale, which makes the two the most reliable answer to whether your media is profitable overall. Google Analytics alone cannot give a true cross-platform return, because it only has the spend from linked Google Ads.

Formula MER = total revenue ÷ total ad spend · POAS = profit ÷ total ad spend

Net dollar retention (NDR)

The revenue a customer base keeps and grows year over year. Didi Azaria, CEO of Workiz, benchmarks small-business SaaS at 110 to 130 percent and enterprise well above 150. Measure it per segment, not as one company-wide figure: invest in a segment above 110 percent and cut one below 100. At $100 million in revenue, a 10 percent loss is $10 million to earn back.

Churn

The rate at which customers leave. One blended churn rate is an average, and the cause hides inside it. Split it by how each customer was won: chasing many kinds of customers at once brings in some who were never going to stay, so the experimenting that drives growth can drive the churn too. Serhat Pala of Cross Ocean Ventures found exactly that at a growing company.

Incrementality

The extra sales an ad truly caused, the ones that would not have happened without it. Platform analytics cannot show it, because the platform grading the ads is the one selling them. Ron Oren of Imagen measures it with a blackout: turn every paid campaign off for a set window and count the growth that keeps coming. For brand spend, run small geographic tests.

Attribution

Tracing which marketing touch produced a given sale. An early startup rarely has it clean: Maya Szutan Azoulay of lool ventures calls it a very hard nut to crack, and the ad platforms grade their own work, reporting one blended cost that mixes buyers who were coming anyway with the ones the ads won. So tie marketing's variable pay to a revenue number it shares with sales.

KPI (key performance indicator)

The number a team is judged on. At an early startup, track one, the number that tells the story of the business, and read a longer list as a sign you have not found it yet. Shaul Olmert ran PlayBuzz on how often people shared its content, and runs KwaKwa on whether creators come back to build a second course.

OKR (objectives and key results)

A goal and the measurable results that show it was reached. For marketing, set them from the one number the company is judged on that quarter, usually revenue or growth: how many deals it needs, how many prospects the team must reach to close them, and what each person does this week. Anything that does not trace back to that number stays out.

Vanity metric

A number that looks good and does not trace back to the company's current top-level goal. Maya Ber Lerner of Chiefy borrows impact mapping from product work: every initiative names the goal it serves and the action it should cause. A visitor count that leads to no meeting fails that test and stops being reported.

The 70/70 rule

Udi Ledergor's test, taken from sales teams: a target is set right when about 70 percent of the team hits about 70 percent of it. If everyone clears it easily, the bar was too soft and you pay accelerators you did not need. If nobody comes close, the number was fiction and people quit. It works on the number you hand a first marketing hire, too.

Positioning and go-to-market

ICP (ideal customer profile)

The customer a product is built for. When your last few deals closed with completely different kinds of buyer, a large enterprise, a small startup and a government agency, the ICP is not defined yet, and a new marketer has no motion to scale. Lior Handelsman's rule: change the ICP to someone the product fits, or change the product to fit the ICP.

Positioning

What the company is and who it is for. Technical founders fund lead generation and postpone positioning until the product sells at scale; Barrel Kfir of Dell Technologies Capital calls ignoring it one of the biggest mistakes technical teams make. It is cheap to get right early, often with a proven branding agency rather than a full-time hire.

Value proposition

The one sentence that says what the customer gains: the money you save them or the profit you add, before any feature. Orlie Gruper of Mobilitech Capital tests it against a slide that shows the company in three icons. If the sentence needs a paragraph to explain itself, the offer is still too broad.

GTM (go-to-market)

The work of reaching and selling into a market: who to sell to first, what to say, and which channels in what order. Where a regulator or a multi-party purchase stands between you and the buyer, being there matters; Talor Sax of eHealth Ventures says meeting the US regulator in person saves two years over doing it by email.

Fractional CMO

A part-time senior marketing hire. It makes sense once a company knows which numbers to move and can fund someone to own them. Before that, while the founder is still working out what the churn and the growth are made of, a consultant fits the stage better. At seed, Talor Sax sees most companies with no marketing team at all, maybe one person or someone fractional.

AI and search

GEO (generative engine optimization)

The work of getting ChatGPT, Gemini, Claude, Perplexity and Google's AI Overviews to cite you, put you in the answer and recommend you. It is a layer on top of SEO, not a replacement. The engine cites a short list of brands, so it is closer to pass or fail than to a ranking, and citations usually start moving between week 6 and week 12.

AEO (answer engine optimization)

The older name for getting your content picked as the direct answer, starting with Google's featured snippet above the results. In practice people use AEO and GEO interchangeably, and the GEO guide on this site covers both. Brand strength moves either more than technical tricks: in a study of 75,000 brands, Ahrefs found web mentions tracked AI visibility far more closely than backlinks.

AI Overviews

The AI summaries Google shows at the top of ordinary search results, and where the most people meet an AI answer. BrightEdge saw them on about 48 percent of its tracked queries in early 2026; Semrush puts them on about 16 percent of US searches. When one appears, the click-through rate on the first result falls about 58 percent below expected, per Ahrefs.

llms.txt

A proposed standard: a plain-text file at a site's root that gives AI engines a clean map of its content. As of September 2026 no major engine confirms it reads the file, and Ahrefs found 97 percent of the files it checked got no requests at all. It is cheap to add and harmless, and this site has one; server-side rendering, indexing and answer-first pages come first.

MCP (Model Context Protocol)

An open standard that connects an AI agent to an outside system, like the one power socket everyone finally agreed on. Google Ads, Google Analytics, Meta and TikTok each have an official MCP connection, so Claude can read an ad account in plain language. The guide's rule is read-only first: Meta and TikTok can also change budgets and campaigns, and a person reviews every output.