Scoring a marketing agency at ninety days in India, and the four things it is fair to judge that early

Ninety days is long enough to judge four things about a marketing agency and not long enough to judge the fifth. You can fairly judge whether the measurement was put right, whether the work in the scope actually shipped, whether the agency ran and reported real tests, and whether it communicates when something goes wrong. You cannot fairly judge compounding results, because search and content work does not finish arriving in a quarter and a paid account rebuilt in month one has had six weeks to run. Most ninety day reviews go badly because the buyer scores the fifth thing and the agency defends the first four, and both sides leave the meeting annoyed.

This page is the buyer's scorecard. Twelve items, each one pass, partial or fail, with the Indian measurement gaps that make the first four items harder here than the published guides assume.

Agree these four numbers before day one, or the review has no baseline

• The business objective in one sentence, with the number attached. More qualified enquiries from Maharashtra, from 40 a month to 70, is an objective. Grow the brand is not.
• The baseline, measured before anything changes. Last three months of enquiries, spend and cost per enquiry, written down and signed off by both sides.
• What is in scope, as deliverables with volumes. Six landing pages, two campaigns, twelve posts, one tracking rebuild.
• What you owe them, and by when. Logins, product information, approvals, a named person who answers in a day. More ninety day reviews fail on this line than on the agency's work.

The twelve item scorecard

Items 1 to 4 are the India specific half, and they are where the quarter is won

Three measurement gaps are routine here and each one quietly invalidates every result figure above it.

• Account ownership. An ad account opened in the agency's name means your campaign history, your conversion data and your audience lists leave when they do, and the ninety day numbers are unverifiable. Our page on judging a Google Ads agency in India on account ownership and billing quotes what Google itself publishes about the payments profile and the billing address, which is the detail that decides this on day one rather than at day ninety.
• Phone and WhatsApp enquiries. In most Indian service categories these are a large share of real enquiries and no advertising platform counts them. An agency that reports form fills only is reporting the smaller half, and the cost per enquiry it prints is too high rather than too low.
• Regional language campaigns. If you ran in English and one Indian language, there are two cost per enquiry numbers. One blended average hides which half worked.

Four things not to score at ninety days

• Organic search results. A content and technical programme started in month one is still arriving. Score whether the pages shipped and whether they are indexed, not what they rank for.
• Brand effect. Real and not measurable in a quarter without a study nobody at this budget is running.
• Revenue, where the sales cycle is longer than the quarter. Score qualified pipeline against the written definition instead.
• A single month's cost per enquiry. Indian demand moves with festivals, exam seasons and the monsoon, and a three month average read against the same months last year is the honest comparison.

How to read the score

• Nine or more passes. Extend, and spend the meeting on the next quarter's three priorities.
• Five to eight, with items 1 to 4 passing. Extend one quarter with the three weakest items written into the next scope. The measurement is sound, so the next ninety days will actually prove something.
• Five to eight, with any of items 1 to 4 failing. Fix the measurement before extending, and say so plainly. A result you cannot verify is not a result either of you can use.
• Fewer than five. Give thirty days' notice as the contract allows, ask for the account handover list in writing, and keep the logins. The nine trades that sell under the phrase marketing agency in India is worth re reading before you appoint the next one, because a surprising share of failed engagements were the wrong trade rather than the wrong firm.

Questions about judging an Indian agency at ninety days

Is ninety days the right first commitment

It is the common one here and it works when the scope names deliverables and the agreement names a review date. Shorter than that and the agency spends the whole engagement on access and setup. Longer than that and a poor fit costs a year.

What if the agency spent the first month on tracking and nothing was launched

That is usually a pass on items 1 to 4 and a partial on item 5, and it is the right order of work. An account launched on broken conversion tracking produces three months of numbers nobody can use. Ask for the tracking work to be evidenced, then judge it.

Should the ninety day review be in the contract

Yes, with the date, who attends and what will be presented. A review nobody scheduled becomes a conversation about whether to pay the next invoice. For the full list of what the agreement should carry, our page on what a marketing agency contract has to contain covers the clauses two assistants agreed on, and the ownership and exit clauses read the same way wherever you are.

The agency wants six months before being judged. Is that reasonable

It is reasonable for the results and not for the first four items, which are all deliverable inside the first month. Agree a ninety day review on measurement and delivery and a six month review on outcomes, and both of you get a fair hearing.

Who should attend the review

Whoever signs the invoice and whoever receives the enquiries. The second person is usually missing and is the one who knows whether the leads were any good.

What should I have in writing before the review

The baseline, the scope with volumes, the conversion definitions and the qualified lead definition. All four fit on one page and all four are the things that get argued about at day ninety.

The item to add to the scorecard next quarter

There is a thirteenth item worth putting on the next version of this list, and almost no Indian agency reports it yet. When a buyer describes your requirement to ChatGPT or Gemini and asks who does this work, the answer names some firms and not others. On 22 October 2026 we put 198 India buying questions to both engines, 50 of the readings blind with the country written into each question, and a business was named in the answer text in 19 of them while we were named in none. Asking your agency what that answer says for your own category, and keeping the result beside the twelve items above, turns an invisible channel into a line somebody is accountable for. That reading is what AI Knows Us does, and the 597 reading benchmark across three markets publishes the method in full.