The First 90 Days Are the Most Honest Portfolio Piece

You are four tabs into an agency’s portfolio and every account looks accomplished, which is exactly the problem. None of it tells you how much the agency built and how much was already there on the day they got the login.

Why the flashiest month in a portfolio is the least diagnostic one

A mature account accumulates advantages that have nothing to do with whoever runs it this quarter. A follower base that already decided the brand is worth attention. A backlog of formats that proved themselves years ago, plus running bits the audience recognises on sight and rewards out of familiarity. A visual and verbal identity read at a glance, so a post is placed as this brand’s before anyone reads the caption.

None of that is attributable to current craft. A strong month here can be the payoff of years of prior work, some of it done by an entirely different team, not evidence of what this agency can do starting from nothing. It is a sharper version of what a portfolio leaves out, and a close cousin of how much of that work did the agency actually do. The work is right there. The credit is the ambiguous part.

What the first 90 days strips away

A launch is useful because none of that exists yet. On a cold start there is no audience to carry the first posts anywhere, no history of interaction to lean on, no visual or tonal memory in anyone’s head, and no library of formats that already earned a response. The premise is structural rather than technical: an account with a long posting and interaction history has accumulated audience signal that a two week old account has not.

That absence is what makes the window diagnostic. Whatever holds attention in those first three months came from decisions made inside it.

What to actually look for in the first quarter

• How early the system arrives. Does a recognisable visual or tonal grammar show up inside the first handful of posts, or does the account spend thirty posts trying on outfits before something sticks? Only one of those is a plan.
• Whether the captions sound written from post one. Read the first five captions and the fiftieth. If post one is generic and post fifty has an angle, the voice was found rather than designed.
• Whether there is a sequence. Look for a post that establishes a premise the account then riffs on. The failure pattern is every post reading as an independent attempt, unrelated to the one before it.
• Whether the account visibly adjusted. Around week four or six, does the rhythm change, does a format that did not land quietly disappear, does one that did get repeated with variation? Reacting to early evidence is craft.
• What the early comments say. Generic praise means little. People quoting a line, arguing with the premise, or asking about the thing the post was about means it gave them something to respond to.

The excuse that early work is unfair to judge

The objection is that a launch will always look rougher than a two year old account. Half of that is right. Production polish improves with budget and time, and early work will have thinner assets. Discount all of it.

The other half does not hold. A point of view is not a resource. Neither is a sequencing decision, a caption’s voice, or a visual system built from a font, a crop rule and two colours. Those are choices, and choices are as available on day one as on day five hundred. Discount the budget and the runway, judge the decisions.

A question that separates portfolios in a pitch

Ask for it in these words: show me the first 30, 60 and 90 days of a launch you ran, not the best month of an account you took over.

A strong answer is a team walking through decisions made before any performance signal existed. It might describe opening with posts that establish what the account is for before any product appears, holding a recurring format back until there were enough followers for it to land, dropping another once the comments came back polite and nothing more. The tell is that they can say why they made each call at the time.

A weak answer redirects. The team pivots to a mature account’s peak month, or describes the launch in language that could apply anywhere: they built the community organically, they let the content find its voice. Vagueness about a period you were responsible for is itself an answer.

Reading a launch you did not see live

You usually can. If the account is public and still active, scroll to the bottom. The earliest posts are there, in order, captions and comments intact. That is primary evidence rather than the agency’s retelling, in the same way that an agency’s own account is weak evidence of its work while a client account’s archive is not.

When the earliest posts are gone, the fallback is the conversation. Ask what went out first, what was held back, what got cut. A team that made those calls answers at that resolution without preparing, and a team that cannot goes smooth and unfalsifiable instead. Read across several launches where you can, since one excellent campaign tells you almost nothing.

Where this reading gets applied

This is the reading behind the list we keep here. See this year’s selections, or nominate an agency whose launch work deserves a closer look.

FAQ

What if an agency has no recent launch to show?

Common, and not disqualifying on its own. The fallback is to ask how they would approach a cold start: what they would establish in the first two weeks, what they would hold back until there is a following. A vague answer to that is itself informative.

Isn’t early work always rougher than mature work?

Yes, and you should expect it and discount it. The line to draw is between production polish, which improves with time and budget, and the presence of a point of view or a deliberate sequence, which costs nothing and is available on day one.

How is this different from judging one campaign in isolation?

A campaign is one moment. A launch’s first 90 days is already a running body of work across many posts. The criterion here is about which phase of an account’s life you read, not how many pieces of work you read.