Digital Marketing for Startups: What to Do at Each Stage

Three stages, three questions. Spend follows the answers, not the other way round.

The most expensive marketing mistake we see founders make is buying ads before anyone has asked for the product.

Ads amplify whatever is already there. If people try the product and leave, a bigger budget just gets you more people who try it and leave, faster.

This guide covers what to do, and in what order, before product-market fit and after it. Stretching a small budget across channels in any business is covered in our guide to digital marketing on a small budget, and growth-hacking tactics have their own post; here we stay on sequence.

We will follow one hypothetical company throughout. Milaan Recon (made up for this guide) is a three-person, pre-seed B2B SaaS startup in Bengaluru. It sells a bank-reconciliation tool to chartered accountancy firms. It has a working product, 11 firms on free pilots and ₹40 lakh in the bank.

Why spending on ads before pull is the classic startup mistake

Paid ads tell you whether people click. A startup first needs to learn whether people stay, pay and tell a friend.

Zerodha is the loudest Indian counter-example to the "raise, then advertise" script. In a Business Today report of 20 January 2025, Nithin Kamath said the broker grew by word of mouth, with nearly 30% of customers coming through referrals, and put it bluntly: "Advertising is like cocaine. Once a business gets used to it, growth becomes dependent on it."

That does not mean every startup should never run an ad. Freshdesk used ads early, but as a test rather than a growth engine. Girish Mathrubootham's own account, published on the Freshworks blog on 18 March 2018, describes a beta sign-up form promoted on Google AdWords, Facebook and LinkedIn with $10, $25 and $50 budgets. Around $350 to $400 of spend brought more than 150 valid sign-ups.

That money bought an answer to "does anyone want this enough to sign up?" for the price of a nice dinner. It is the only kind of ad spend we would sign off on before product-market fit.

The stage map: one question, one channel, one metric

Each stage has one question to answer. Everything you spend should help answer it, and anything that does not can wait.

PMF = product-market fit. Treat the "do not spend on yet" column as the most useful part of this table: most wasted startup budgets end up in it.

Stage 1, before product-market fit: sell it yourself

Before fit, marketing is the founder talking to customers. Nobody else can do it for you yet, because nobody else knows which objections should change the product.

Aim for roughly 50 conversations with people who have the problem. Fifty is our working number: it is usually where the same complaints start repeating.

Meesho's founders did the unglamorous version of this. In a Meesho Tech interview published on 7 July 2020, Vidit Aatrey says the first model (Fashnear, a local-shops idea) was not working because buyers cared about deals more than about products being local. He also worked as the delivery person early on "because I wanted to hear my customer's feedback first hand." The pivot came from watching a shopkeeper sell through WhatsApp and Facebook groups, then noticing that women resellers used those groups differently.

For Milaan Recon, this means phoning all 11 pilot firms every week, watching someone run a reconciliation on screen share, and counting how many still use it in week eight.

To check for fit, run the Sean Ellis survey once you have enough active users to make the result mean something. It asks one question: how would you feel if you could no longer use this product? As Typeform's write-up of the test explains, Ellis found that companies where at least 40% of users answered "very disappointed" almost always had strong traction. With 11 pilots the percentage is noisy, so read the reasons behind the answers more closely than the score.

Pick one channel for this stage. For B2B in India, that is usually direct outreach to a list you build by hand; for a consumer app, one community where your first users already gather.

Stage 2, finding fit: one repeatable acquisition channel

You are looking for one channel that brings the right customer at a predictable cost, rather than five that each bring a few.

A channel counts as repeatable, in our view, when it gives you similar cost and similar conversion across three consecutive cycles (weeks for consumer, months for B2B) without the founder personally rescuing each deal. One good week is luck. Three is a pattern you can start to plan around.

Run two or three small tests, each with a budget ceiling, an end date and one success number written down before it starts. When a test fails, stop it, instead of letting it drift at ₹500 a day for four months.

Milaan Recon might test three things: Google search ads on high-intent terms CAs type (they will know these from the Stage 1 calls), a partnership with a regional CA association for a demo session, and a monthly LinkedIn post from the founder showing a real reconciliation before and after. Whether sales will run the motion, or the product will sell itself through a free tier, changes which test goes first. We cover that choice in product-led vs sales-led growth.

Cheap experiments like referral loops and onboarding tweaks are in our growth hacking guide. Use only the ones that feed the channel you are testing.

Stage 3, after fit: scale paid, start SEO, watch unit economics

Once one channel repeats and customers stay, spending more makes sense. The question changes from "does this work?" to "how much can we spend and still get our money back in time?"

Two numbers decide that. CAC payback is how many months of gross margin from a customer it takes to earn back what you spent to acquire them. LTV:CAC compares the total margin a customer brings over their life with that acquisition cost.

A worked example, using Milaan Recon's hypothetical figures: if a firm pays ₹8,000 a month at an 80% gross margin, each firm brings in ₹6,400 of margin a month. If acquiring one firm costs ₹40,000 across ads, demos and sales time, payback is about six and a quarter months. Work out your own with our CAC calculator, and for the other side of the ratio, our guide to customer lifetime value shows how to estimate LTV when you only have a few months of churn data.

Calculate both per channel. A blended CAC hides the channel that is losing money behind the one carrying it.

This is also when SEO starts to pay. Organic search takes roughly three to six months to show meaningful results and compounds after that, so it matures around the time paid costs begin to rise. Start with pages answering the questions your Stage 1 customers asked on calls.

Paid scale is where outside help earns its fee: bidding, creative testing and clean conversion tracking. If you are here, our performance marketing team can run the proven channel while you keep the product moving. As a directional India range, performance marketing management runs about ₹25,000 to ₹1,00,000 a month, and ad spend is always separate from that fee.

Before you scale, write down what each quarter is for (pipeline, payback, a new city). Our guide to setting marketing objectives shows how to phrase them so they can be checked.

India-specific realities founders underestimate

Language comes first. If your buyer in Surat or Coimbatore thinks about the problem in Gujarati or Tamil, your Stage 1 calls should happen in that language, and your first landing page may need to as well. Test before translating the whole site: some B2B buyers want English in the product and their own language on the phone.

WhatsApp is a channel, with rules. The WhatsApp Business Messaging Policy says you may contact people only if they gave you their number and opted in to receive messages from you. On the paid API, Meta has charged per message since 1 July 2025 for marketing, utility and authentication templates, while replies inside the 24-hour customer service window are free. Before fit, use WhatsApp for conversations customers start, not broadcasts.

UPI changes how trials convert. Under the RBI's 16 June 2022 e-mandate notification, recurring debits up to ₹15,000 per transaction can run without an extra authentication step once the mandate is set up. That makes a "₹1 trial, then monthly" offer easy to build. It also makes it easy to annoy people, so state the renewal amount and date on the offer itself. A surprise debit comes back as churn and angry reviews.

Consent is now a legal question. The government notified the Digital Personal Data Protection Rules on 14 November 2025, with an 18-month phased timeline and a requirement for clear, standalone consent notices, according to the PIB release. Build consent into your sign-up and lead forms now, while your list is small. Our explainer on the DPDP Act for marketers covers what that means for forms, lists and retargeting.

The first 90 days: a worked plan

Here is how we would spend the first 90 days of marketing at Milaan Recon, our hypothetical pre-seed B2B SaaS startup in Bengaluru.

1. Days 1 to 7: write down the one question. "Will CA firms with 5 to 30 staff use this every month-end without being chased?" Every activity for 90 days has to help answer it.
2. Days 1 to 30: run 50 customer conversations. Weekly calls with all 11 pilots, plus about 40 new firms found by hand through LinkedIn and alumni networks. Record the exact words buyers use for the problem.
3. Days 1 to 30: set up measurement before any spend. A simple sheet or CRM with one row per firm: first use, last use, month-ends completed, and a consent flag for every contact. Retention at week 4 and week 8 is the headline number.
4. Day 30: decide the product changes. Fix the two complaints you heard most. Do not start acquisition while a known blocker is still in the product.
5. Days 31 to 60: run the Sean Ellis survey and read every reason. Then pick two channel tests, each with a written budget cap (say ₹25,000), an end date and one success number, such as five qualified demos.
6. Days 31 to 75: run the tests in parallel. For example, search ads on phrases heard in calls, pointed at a demo-booking page, plus one CA association demo session. Log cost per qualified demo weekly.
7. Days 61 to 90: write the answers-first pages. Three or four pages answering the questions CAs asked most on calls. This is the seed of SEO later, and a sales asset now.
8. Day 90: keep one, kill one, calculate payback. Keep the channel that repeated. Stop the other. Run the surviving channel's numbers through a CAC calculator to see the payback period before raising its budget.

Notice what is missing: no brand film, no agency retainer, no Instagram calendar. At day 90 they would only add noise to the signal you are trying to read.

What to do next

Work out which stage you are in by checking retention. If users drift away after a month, stay in Stage 1 and keep talking to them. If one channel keeps bringing the right buyer, measure its payback and give it more money.