DK logodavorkarafiloski
AboutCase StudiesSequencerSEO ClockGlossary Schedule a call
Free tool

Build a 90-day SEO plan you can defend

Most plans fail because they contain everything. Tell me your site and your actual constraints, and this will give you a sequence for the next ninety days — plus the part almost nobody writes down: what you are deliberately not doing.

Nothing is sent anywhere and there is no email gate. It runs on the same timing model as the SEO Clock, so treat the estimates as a basic guide rather than a forecast.

1 · What the business is Pick the motion that pays the bills, not the one on the pitch deck. If you sell both self-serve and through a sales team, choose whichever closes the most revenue.
2 · Where you sell This is about where your customers are, not where your office is. A single shop serving one town is a different problem from a national brand with no premises.
3 · What the topic is Money, health or legal covers anything that could affect someone’s finances, wellbeing or rights — Google holds it to a higher bar. If you are unsure whether you qualify, you probably do.
4 · Your domain today Established means years of history and rankings for hundreds of terms. Growing means real traffic but still outgunned. Be honest: this one answer changes every estimate below.
5 · How big the site is Count indexable URLs, not products or posts. If a filter or facet generates pages Google can reach, those count too — the real number is usually far higher than people expect.
6 · Competition Search two or three of your most valuable terms and look at who owns page one. National brands, marketplaces and comparison sites mean brutal. Directories and small rivals mean low.
7 · Where you are Describe the last three to six months, not the last two weeks. A single bad month is noise; a sustained slide is a drop.
8 · Branded or non-branded Protect the brand SERP if people search your name and find something you dislike. Win non-branded if the problem is that nobody knows the name yet.
9 · Known problems · pick any Only tick what you have seen evidence for in Search Console or a crawl. Guessing sends the plan after the wrong problem — pick “I don’t know yet” instead and the first activity becomes finding out.
10 · Engineering access Be realistic about what you can get shipped, not what is theoretically possible. If tickets sit in a backlog for two quarters, that is closer to none.
11 · Content capacity Count people who can actually publish, not people who could in principle. A writer who also runs paid social is half a writer.
12 · What matters most One only. If everything matters, whoever is judging you will pick for you — and probably not the one you would have chosen.
13 · When you are being judged The date somebody decides whether this is working, not when the contract ends. Budget reviews, board meetings and probation periods all count.
14 · In your own words · optional If the options above do not describe the business, say so here. It travels with the plan when you copy, email or share it.
Before you use any of this

{{ verdictHead }}

{{ verdictBody }}
The link reopens this exact plan, the CSV drops straight into a tracker, and your answers are kept in this browser.
Picked up where you left off — these were your last answers in this browser.
What to tell your stakeholders

{{ promiseHead }}

{{ promiseBody }}

{{ countDo }}Activities in the plan {{ countDrop }}Deliberately dropped {{ firstSignal }}Earliest honest signal {{ loadSplit }}Jobs for you · engineering · writing
Read this first

Your answers contradict each other

Not a mistake on your part — these are the real tensions in your situation. Resolve them before the plan below means anything, because no sequence fixes a strategy that is pulling in two directions.

{{ c.title }} {{ c.body }}
The quarter as a schedule

Thirteen weeks, three lanes

Lanes run in parallel because different people do the work. Inside a lane the work runs in sequence, and a dependency pushes what follows it out rather than pretending two things can happen at once. If a lane looks empty, that resource is not the one holding you back.

Week 12345678910111213
{{ l.label }}
{{ i.name }}
Show Confidence {{ filterNote }}
{{ ph.when }}

{{ ph.label }}

{{ ph.blurb }}

{{ r.name }} {{ r.why }} ⇢{{ r.gatedBy }} If you stop:{{ r.decay }} {{ r.stance }} {{ r.confLabel }} {{ t.label }}
First signal {{ r.first }} {{ r.effort }} · stage {{ r.stage }} Scheduled {{ r.when }}
How to tell it is working

Checkpoints, and what it means if you miss them

A plan you cannot falsify is not a plan. Agree these before the work starts, so a bad quarter is caught in week six rather than argued about in month four.

{{ c.label }} {{ c.emptyNote }}
{{ i.name }} ✓{{ i.see }} ✕{{ i.miss }}
Not this quarter

What you are dropping, and why

This is the half of the plan nobody writes down. Take it to the meeting — a list of decisions is far easier to defend than a list of tasks.

{{ d.name }} {{ d.reason }}
Not doing, on purpose

Things this plan refuses

Popular in your situation, and not worth your quarter. If an agency proposes one of these, ask what it displaces.

{{ n.name }} {{ n.reason }}
Where this plan is fragile
{{ k.title }} {{ k.body }}
How to read this

What the labels mean, and where this model is weak

Every activity carries two labels: what it is for, and how confident I am that the mechanism works. A tool that asserts everything with equal certainty is not one you should trust.

MeasureProduces no traffic on its own. It tells you whether everything else is working, and stops you fixing the wrong thing.
ProtectDefends traffic or revenue you already have. Capped upside, uncapped downside if you skip it, and the first thing an inexperienced plan leaves out.
CaptureGoes after traffic you do not have. Higher ceiling, longer runway, and the first casualty when capacity runs out.
DocumentedGoogle documents the mechanism, or the effect is mechanical: remove the block and the page can rank. Plan around these with confidence.
ConsensusNo official confirmation, but consistent enough across sites and practitioners that I would plan around it. Most of SEO lives here.
Judgement callA bet. Plausible mechanism, I have seen it work, and outcomes vary enough that promising one would be dishonest. Fund these as experiments, never as commitments.
Where this model is weak
The priority order is my judgement, written down as numbers. Every activity carries a base score and a set of adjustments for your model, scale, competition and market, and I set each of those figures by hand from client work — not from a dataset, and not from anything I can show you a regression for. They are calibrated opinions whose job is to produce a defensible order, not a measurement. Two items a few points apart are effectively tied, which is why this tool never shows you a score: a two-digit number would imply a precision that is not there. Argue with the sequence, which is the real claim. Capacity is modelled crudely, on purpose. Three separate budgets — your time, engineering, writing — with a heavy activity costing roughly five times a light one, divided across the three phases. It is a sanity check against a plan that assumes more hours than exist, not a resourcing model. It is also why the summary counts jobs per person rather than quoting a percentage of your capacity: a percentage of an invented budget is precisely the number that ends up in a board deck and cannot survive the first question about it. It has never seen your site. Fourteen answers is not an audit. Real plans change once somebody opens Search Console and finds the thing nobody mentioned, and that happens more often than not. It deliberately does not forecast traffic. A traffic forecast requires assuming a target position, a click-through rate and stable search volume — three numbers nobody can know in advance. Multiply three guesses and you get a precise-looking figure carrying no information. Sequencing and timing are the parts that are actually knowable, so those are what you get here. It does not model who owns page one. A result dominated by marketplaces, an AI Overview absorbing the answer, or a page mostly of ads changes what is winnable more than several of the inputs above. Check the SERPs for your ten most valuable terms before committing to any of this. Every timing assumes the work is done well. A refresh that adds words rather than value produces nothing on any timeline, and the model cannot tell the difference between the two. A core update inside your window overrides all of it. Rankings move hardest during rollouts, and banked work often lands in bulk at exactly that moment. Annotate the dates, and never judge a quarter that contains one.
The resourcing argument

What one more resource actually buys you

Same engine, one input changed. Take the number, not the opinion — this is the difference between asking for headcount and proving you need it.

Start the conversation here {{ bindHead }} {{ bindBody }}
{{ w.label }}{{ w.delta }} {{ w.noneNote }} Unlocks {{ w.names }}.

A tool can sequence. It cannot look at your site.

This plan is built from thirteen answers. A real one comes from your Search Console, your competitors and your codebase — and usually changes two or three of the choices above. Bring me both and I will tell you where this got it wrong.

Copied
© 2026 Davor Karafiloski — SEO & AI search consultant AboutCase StudiesGlossarySEO ClockSchedule a call