Competitive analysis that survivescontact with Monday.

WHAT IS IN HERE

The route.

01

Why most competitive analysis quietly fails.

I have inherited a lot of competitor decks. They are almost always beautiful, almost always sixty slides, and almost always untouched since the day they were presented. The work was real. The output was inert. That gap is worth understanding before you fill in a single cell, because it is not a research problem — it is a design problem, and it shows up in three specific ways.

The first is the dossier failure. Someone collects everything knowable about five competitors and organises it by competitor. It reads like an encyclopaedia entry each. Nothing in it is comparable, because the same fact appears in a different place for each one, and no two people reading it draw the same conclusion. The fix is boring and total: organise by dimension, not by company. One tab per dimension, one row per competitor, and every cell in a column answers the same question. Comparison becomes possible only when the structure forces it.

The second is the undefined scale. Nearly every template on the internet — including the one that prompted me to build this — tells you to "score each area on the same scale" and then never says what the scale means. So your 4 and your colleague's 4 are different animals, the average of them is noise, and the ranking that comes out the other end is a ranking of who was in a generous mood. A scale that is not written down is not a scale. That is why this toolkit ships a rubric tab, and why the rubric is step two of the process rather than an appendix.

The third, and the one that actually kills projects, is the missing verb. Observation is not analysis. "Competitor A publishes three times a week" is a fact. "Competitor A publishes three times a week, owns the back-pain cluster outright, and we cannot beat them there this year — so we take Arabic and insurance, which they have left completely open" is analysis, because it ends in something you can do. Every scored tab in this file ends with a "gap for us" column for exactly this reason, and every finding worth anything is supposed to end up in the action plan tab with an owner and a date. If it does not reach that tab, it did not happen.

02

Decide who your competitor actually is.

This is the step people rush, and it poisons everything downstream. Analyse the wrong five companies with perfect rigour and you get a rigorous, confident, wrong answer. The instinct is to list the brands you find intimidating. That is a list of who you think about, not a list of who takes your money.

There are five kinds of competitor and they need different treatment:

Direct

Same offer, same buyer, same market. They appear opposite you in a shortlist. These get the full analysis.

Indirect

Different offer, same budget line and same problem. A physiotherapy clinic competes with a gym's rehab programme. They rarely show up in your keyword tool and frequently show up in your closed-lost notes.

Substitute

The thing your buyer does instead of buying anything. Doing nothing, doing it in-house, a spreadsheet, a friend who knows a guy. In most B2B categories this is the largest competitor by volume and it is never in the deck.

Aspirational

A market leader you are not competing with yet, tracked because their moves become the category's expectations in eighteen months. Watch, do not benchmark.

SERP-only

Publishers, directories and affiliates who outrank you but do not sell what you sell. They shape demand and they are frequently the page an AI assistant cites. They belong in the SEO and AI tabs and nowhere else.

The most reliable source for the direct list is not a tool. It is your own closed-lost reasons in the CRM, and the "who else are you looking at" question your sales team already asks. Second most reliable: the paid-search auction insights report, which names the accounts genuinely bidding against you rather than the ones a third-party estimator thinks are similar.

Then be ruthless about the count. Five competitors analysed properly beats twenty rows half-filled, and the difference is not effort, it is usefulness. The Competitor Map tab has a Track As column with three values — Core, Watchlist, Ignore. Use Ignore. It is the most valuable value in the file.

03

Define the scale before you score anything.

The rubric tab defines each score from 1 to 5 for all seven dimensions, in plain sentences, with the specific behaviour that earns it. Two people scoring the same competitor should land within one point of each other. If they do not, the scale is at fault, not the people — go and sharpen the wording before you continue.

A few rules that make scoring hold up:

Score the evidence, not the vibe

A 4 for content means "deep clusters, refreshed, named expert authors" — three checkable things. If you cannot point at the evidence, you cannot give the score.

Never score a dimension you did not research

Leave it blank. The weighted formula in the Scorecard ignores blanks and renormalises the weights, so a partial row still produces an honest number instead of a quietly deflated one.

Score against the category, not against yourself

A 5 means best in class for the market, not "better than us". Otherwise your own weaknesses inflate everybody.

Re-score, do not adjust

When you refresh next quarter, score fresh from the current evidence. Nudging last quarter's number is how a scorecard slowly becomes fiction.

04

Paid media — the fastest honest signal there is.

Start here, because paid media is the one area where competitors are legally obliged to show you their homework. Meta's Ad Library lists every active ad on any page, with the date it started running. Google's Ads Transparency Center does the same for Search, Display and YouTube. TikTok has a Commercial Content Library. LinkedIn shows the last six months of a company's ads on its own page, under Posts → Ads. All free, all public, none of them requiring a tool subscription.

The single most valuable field on that tab is Longest-Running Ad, and almost nobody fills it in. Ad start dates are visible, and an ad that has been running continuously for six months is an ad that is paying for itself. Nobody funds a loser for six months. That one date tells you more about what works on your audience than any amount of guessing about their strategy — and it hands you a hypothesis you can test with your own budget next week.

Read the set, not the individual ad. Thirty ads with one message is a company that has found something and is scaling it. Thirty ads with thirty messages is a company still searching. Two ads running for a year is a company with a small, disciplined budget and a proven angle. Each of those calls for a different response from you.

Then click through. The landing page is where most competitors are weakest, because the media team owns the ad and nobody owns the page. Score it on the same rubric you use for your own, note the offer, note the form length, and record the CTA verbatim — you want their exact words, not your summary of them, because the wording is the thing you will later test against.

05

SEO — read the shape, not the number.

Every third-party organic traffic estimate is a model, not a measurement. It is routinely off by a factor of two or three in either direction, and it is worse in non-English markets and in small niches. This does not make it useless — it makes it useful for exactly one thing: relative comparison. The ratio between two competitors, measured by the same tool on the same day, is informative. The absolute number is not, and it should never appear in a board deck as a fact. Write "est." next to it or do not write it at all.

What actually matters on this tab is shape. Look at three things:

Branded versus non-branded split

If 60% of their organic clicks are people typing their name, they do not have an SEO advantage — they have a brand, and their SEO is a fulfilment channel. That is a completely different competitor to one earning non-branded demand, and you attack them differently.

Concentration

Pull their top pages. If one cluster produces most of the traffic, they are strong and fragile at once. A single algorithm shift on that topic takes most of it. Note it in Threats — and do not build your own plan on the same single point of failure.

Referring domains, not raw links

Link counts are inflated by site-wide footers and vary wildly between tools. Referring domains is the stabler figure. Whatever you choose, use one tool for the whole column. Mixing two tools in one column produces a comparison that means nothing, and it is the single most common data error I see.

The output of this tab is not a score. It is the content gaps column: the queries with commercial intent that nobody in your set covers properly. In practice these cluster in four places — pricing and cost content, language editions nobody has built, "X versus Y" comparisons, and the unglamorous operational questions buyers actually ask before committing. Those four are open in more markets than you would believe, because they are boring to write and nobody wants to publish a price.

06

AI search — the surface almost nobody is measuring.

A meaningful share of category research now happens inside an assistant. Someone asks "best physiotherapy clinic in Dubai" or "which CRM handles Arabic properly" and gets a short answer naming two or three companies. If you are not one of them, you never entered the consideration set and no analytics tool will tell you that it happened.

You can measure this by hand, and the method matters more than any tool:

What earns the citation is fairly consistent: a direct answer in the first sixty words, a real structure the model can lift (a table, a numbered list, a definition), a named author with verifiable credentials, and corroboration on third-party pages the model already trusts. Schema markup helps parsing; it does not manufacture authority. And this is genuinely uncontested in most markets — I have audited categories where not one competitor had ever run the prompts. That is a rare thing in 2026: a channel where the work is cheap and the field is empty.

07

Social — throw the follower count away.

Follower count is exact, easy to find, and nearly meaningless. It is a historical number that says what a brand did once, not what it does now. Three fields on the social tab carry the actual signal.

Engagement rate, calculated by you. Take their last twelve posts, average the likes and comments, divide by followers. Do it yourself rather than trusting a platform figure, because vendors define it differently and you need one consistent definition across the whole column. The absolute number varies wildly by category — 1.8% is strong in healthcare and weak in a hobby niche — so what you are reading is the spread within your own set.

The format that repeatedly outperforms. Not the single best post — the format. If their top five posts of the quarter are all the same shape, that shape is their strategy, and you have just found it without a single strategy document. Write it down in "Why It Worked" as a mechanism, not a compliment: "a clinician answering a real patient question to camera, no music, subtitled" is usable. "Great video content" is not.

Reply behaviour. Sample their comments and their DMs and record how long a reply actually takes. This is the most consistently neglected competitor weakness in existence, and it is the one you can beat this week without a budget or a headcount request. If their DMs go unanswered for nine hours and yours are answered in twelve minutes, that is not a nice internal fact — it is a marketing claim you can prove.

08

The content engine test.

The difference between a blog and an engine is whether anything compounds. Three things are hard to copy, and they are the three the tab is built around.

Topic ownership. Does one competitor own a cluster so thoroughly that the whole category defers to them on it? Ownership looks like twenty to sixty interlinked pages, updated, internally consistent, with the hub outranking everyone on the head term. Where you find it, the honest strategic conclusion is usually "we cannot take this cluster this year". That is not defeat — it is what frees you to go and take an open one.

Original data. One annual survey, benchmark or index earns more links, more citations and more AI mentions than a hundred derivative posts, and it cannot be copied because it is theirs. If a competitor has one, note it and note when they publish it. If nobody in your set has one, that is the highest-leverage row in your whole action plan.

Refresh discipline. Check the visible update dates. A site where a third of posts show a current-year revision is running a maintained asset. A site where everything is dated at publication is running an archive. The second one decays on a schedule, and you can wait it out.

09

Website and CRO — score the journey, on a phone.

Do this properly: a real phone, mobile data, logged out, in a private window, as somebody who has never heard of the company. Not a desktop browser at 1440px with their cookie banner already dismissed. The conditions are the finding — most site audits are wrong because they were done in comfortable conditions no real visitor experiences.

Count the steps between arriving and being able to buy or book. Then find the step where a normal person would stop. There is almost always exactly one, and it is almost always a form field asking for something the visitor is not ready to give: a national ID number before showing availability, a phone number before showing a price, a company size before letting anyone read the demo page. Write that step into Conversion Friction in plain language. It is the most valuable single cell on the tab, because if they fix it their numbers move a lot, and until they do, it is yours to exploit.

Record the value proposition and the primary CTA verbatim. Their exact wording is a research artefact — it tells you what they believe their buyer responds to, it is the thing you will test your own wording against, and a paraphrase destroys all of that.

Use field data for speed, not lab scores. PageSpeed Insights shows both; the field section is real users from the Chrome UX Report and the lab section is a simulation on a modelled device. Reporting a lab score as user experience is a small lie that gets repeated a lot.

10

Pricing and offers — the offer is the strategy.

Record the numbers, but do not stop there. The interesting columns are the ones that explain how they avoid competing on price at all: what they bundle, what risk they absorb, what they guarantee, what they lock in, and what they charge extra for. A competitor with a real guarantee and a two-week onboarding commitment is not selling the same thing as one with the same price and neither.

Two techniques earn their place here. First, if they do not publish prices, that is itself a finding — write down what they say instead ("contact us", "packages from", "insurance accepted") and, if you can, why. Hidden pricing usually protects a channel margin, a negotiated rate, or an inconsistent one. It also creates a specific, exploitable gap: in almost every category where nobody publishes prices, buyers complain about exactly that, and the first brand to publish takes a disproportionate share of the high-intent traffic.

Second, run their pricing page through the Wayback Machine. Twenty-four months of price history tells you about their margin pressure, their positioning drift and their confidence in a way no current snapshot can. A brand that has quietly added a cheaper tier twice is a brand under pressure, whatever their marketing says.

11

Share of voice — the only tab with a time axis.

Everything else in this file is a snapshot. This one has a date column and one row per competitor per month, and it is the tab that eventually justifies your budget, because it is the only one that can show movement.

The four component percentages — organic, paid, social, AI citation — do not need to be precise. They need to be consistently computed. Pick a fixed keyword set, a fixed prompt set and a fixed platform list, compute each competitor's share within your own tracked universe, and never change the method. A share of voice number that is 15% wrong but computed identically every month is enormously more useful than a perfect number computed a new way each quarter.

Use Google Trends for branded demand rather than a keyword volume tool. Trends is relative and normalised by design, which is exactly what share of voice needs, and it lets you plot brand names directly against each other. Set your first month to an index of 100 and track relative change from there.

The last column, What Moved, is what makes this tab survive into a board meeting. "Social SoV +9 points" is a number. "Social SoV +9 points — they launched a Reels series with four creators on 1 August" is a story, and it is the version that gets a decision made.

12

A SWOT that ends in a verb.

A four-box SWOT is a workshop artefact. Useful for forty-five minutes, dead by Friday. This version has five extra columns and they are the entire point: Implication For Us, Our Counter-Move, effort, owner, due date.

The discipline is that a weakness you cannot exploit is not a weakness — it is trivia. If a competitor has no Arabic content and you also have no route to producing Arabic content, that is not an opportunity, it is a fact about the world. Either it converts into a counter-move with a name and a date, or it comes out of the file. Being strict here is what keeps the analysis honest, and it is why so many competitor decks are long: nobody applied this filter.

One more thing worth writing down: their strengths are usually built on something you cannot replicate — brand equity, a cost base, a decade of accumulated links, a founder who is genuinely famous in the niche. Copying the leader's tactics without the leader's underlying assets is how a marketing budget disappears. The correct response to an unassailable strength is rarely to attack it. It is to find the adjacent thing they have left open and take that instead.

13

ICE, and the first ninety days.

Everything worth doing lands in the Action Plan tab and gets three scores from 1 to 10: Impact, Confidence and Ease. The file computes the geometric mean, which is deliberate — the arithmetic mean lets one high score carry a plan that is quietly impossible, while the geometric mean punishes any single low input. An action scoring 10 on impact and 2 on ease should not outrank a solid 7-7-7, and with a geometric mean it does not.

Confidence is the honest column, and it is the one people fudge. It is your confidence in the impact estimate, not your enthusiasm for the idea. If you are guessing, say 4. A plan full of 9s in the Confidence column is a plan nobody stress-tested.

Then split the ranked list by horizon:

Fill in Expected Outcome before you start and Actual Outcome after. Doing this consistently for two quarters is what turns a competitive analysis from an opinion into a calibrated instrument — you find out which of your predictions were right, and the next round of estimates gets better. Almost nobody does this, which is why almost nobody's ICE scores improve.

14

Setting the weights for your business model.

The Scorecard produces one weighted number per competitor, and the weights live in row 6 where you can change them. They do not need to add up to 100 — the formula normalises them, and a weight of 0 removes a dimension entirely. This matters more than it sounds: an equal-weighted scorecard implicitly claims that social and pricing matter identically to your business, which is true for almost nobody.

Four starting points I would defend:

Set the weights before you score. Setting them after you see the scores is how you accidentally rig the answer towards whatever you already wanted to do, and you will not notice yourself doing it.

15

Cadence — monthly, quarterly, never again.

A competitive analysis with no refresh cadence is a photograph of a market that has already moved. But refreshing everything every month is how the whole practice gets abandoned by March. Split it:

16

Where the lines are.

Everything in this toolkit is filled from public sources, and that is not a limitation — it is most of what there is. Ad libraries, live pages, sitemaps, review platforms, public registries, job listings, the Wayback Machine, and your own analytics will answer nearly every field.

Three things I will not do and would advise against. Do not pose as a customer under a false identity to extract a quote or a demo — beyond the ethics, it is the fastest way to make your own brand the story. Do not ask a new hire for numbers from their previous employer; it puts a person at real legal risk to satisfy your curiosity, and any figure obtained that way is unusable anyway because you can never cite it. And do not use scraped data obtained in breach of a platform's terms, particularly personal data — in the UAE, the EU and increasingly everywhere else, that is a regulatory problem attached to your company name.

The practical point is that the ceiling on public data is much higher than most teams assume. In fourteen years I have never needed anything else, and the fields people imagine require inside information — spend, conversion rates, margin — are the ones you should be inferring from behaviour anyway. What a company keeps paying to say for six months tells you more about what works than the number attached to it.

17

Eight mistakes worth naming.

Ranking competitors by how much they intimidate you

Rank them by who appears in your closed-lost reasons. That list is usually different, and it is the true one.

Scoring on an undefined scale

Two people, two meanings, one average, zero information. Write the rubric first.

Reporting an estimate as a fact

Third-party traffic figures are directional. Use them to find gaps, never to report performance.

Mixing two tools in one column

Different indexes, different definitions. The comparison silently becomes meaningless while still looking rigorous.

Copying the leader

Their tactics sit on brand equity and a cost base you do not have. The tactic transfers; the conditions that made it work do not.

Filling in twenty competitors

Five, properly, beats twenty half-filled. Use the Ignore value — it exists for this.

Stopping at the SWOT

No owner, no date, no counter-move, no analysis. Just a wall of adjectives.

Doing it once

The value is in the second and third pass, when you can see direction. The first pass is only a baseline.

18

How to use it in your first week.

FAQ

Questions people ask.

KEEP READING

Related field guides.