How the Agency List Score is measured
ALS measures how well an agency demonstrates its capability publicly; it does not measure client outcomes.
Everything an assessment reads is published by the agency, on the agency’s own website. We do not survey clients, we do not audit results, and we do not see a single engagement from the inside. What follows is exactly what we do, what the thresholds are, and what the score cannot tell you.
What the score is
The Agency List Score is a number from 0 to 100 built from the evidence an agency has published about its own work. An assessment fetches the agency’s homepage, then follows the links that homepage offers to its case studies, portfolio, client list and testimonials, and reads those pages too. A score landing on one of the bottom rungs gets a second, deeper read before it is stored, because a bottom score is the least trustworthy thing the process produces and a site that merely looks empty often is not.
A separate confidence figure records how much of the site we managed to read. An agency below 50% confidence still gets a score and is not ranked against anyone, because a site we could barely read is a fact about our reading and not about the agency.
How a score is built
Three steps, in this order, and the order is the whole design.
1. The kind of evidence sets the band
The strongest single piece of evidence on the site decides which range of the scale the agency can occupy. The ladder is below.
2. The best three pieces fill the band
The three strongest pieces are combined, weighted 50 / 30 / 20%, so the strongest dominates and reaching the top of a band takes three genuinely strong pieces rather than one good one repeated.
3. Volume adds a little, inside the band only
How much an agency has published contributes at most 12% of its own band — about 2.3 points at the widest one. Publishing more can order an agency above a peer with identical evidence. It can never carry it past an agency a rung above.
That last rule holds for everything in the score: a secondary signal moves an agency inside its band and never across one. It is stated here because it is the property that makes the number mean something — an agency cannot out-publish a better-evidenced rival.
The evidence ladder
Each rung requires something harder to fake than the rung below it. Naming a client is a commitment the client can publicly object to; a number attached to a named client is the hardest claim to make falsely. The bands are the ranges of the 0–100 scale each rung maps to.
| What is published | Why it ranks there | Band |
|---|---|---|
| A named client, with a result carrying a number, and the page says the agency's work produced it. | The strongest thing an agency can publish: a number, attached to a company that can publicly object to it. | 80 to just under 99 |
| A result carrying a number and attributed to the agency's work, where the client is described but not named. | The number is checkable in shape but not in source, because no one is standing behind it by name. | 76 to just under 80 |
| A named client with the work written up, and no confirmed number. | The engagement is described and the client is named; nothing quantifies what changed. | 68 to just under 76 |
| A testimonial credited to a named person at a named company. | Someone identifiable is standing behind the quote, but nothing about the work is measured. | 64 to just under 68 |
| Client names or logos with no write-up of the work. | A logo grid asserts a relationship and describes no work at all. | 50 to just under 56 |
| No published client evidence found. | A statement about what we could read on the site, not a judgement about the agency. | no score published |
A zero is an absence, not a low score
An agency whose site we read and on which we found no published client evidence sits at the bottom rung. That is a statement about what we could read, not a verdict on the agency, so we never publish it as a score. Those profiles show no score at all.
An agency we have not assessed also shows no score. Not a zero, not a placeholder, and not a number from the retired first version of this score — nothing.
Cohorts, standings and awards
Agencies are compared inside a cohort: one category or service, in one country, region or city. A cohort is drawn from the agencies we have assessed, not from every listing in the directory. Whenever a claim is made from a cohort, the size of that cohort is published with it. “Top 10%” is not a credential; “top 10% of 340 assessed” is one a reader can weigh.
When a position is published it is a band with its denominator, never a rank. The scores at the top sit very close together, and ordering them is defensible in a way that separating them is not.
The cohort size a band requires
| Band | Minimum assessed agencies | Ranked positions inside the band |
|---|---|---|
| Top 25% | 8 | 2 |
| Top 10% | 20 | 2 |
| Top 5% | 40 | 2 |
| Top 1% | 300 | 3 |
A cohort must be large enough to express the band being awarded: at least 8 assessed agencies for a top 25% award, 20 for top 10%, 40 for top 5%, and 300 for top 1%.
Each floor is the smallest cohort in which the band holds enough ranked positions to be robust: 2 positions for the wider bands, so one agency moving past another does not falsify the claim. The top 1% band requires 3, and so a cohort of 300, because scores at the very top sit close together and a band holding one agency would be drawing a line the evidence cannot support.
A standing and an award are different objects
A standing is where an agency sits in a cohort right now. It moves when a competitor is assessed or when the cohort grows, so it is only ever read live and shown on pages we control. It is never frozen into anything that leaves this site.
An award is a standing frozen with the year it was earned, as in “Top 5% · Digital Marketing · Canada · 2026”. The year is what makes it safe to put on someone else’s website: it stays permanently accurate even after a competitor overtakes them, where an undated crest would quietly start asserting something no longer true. One award per cohort per calendar year, issued when it is earned rather than on a calendar date, and awards do not expire.
An award is revocable. If the evidence one rests on is later found to be wrong — an outcome we read that the page does not support, prose captured from an error page — the award is withdrawn rather than left standing.
What is published today, and what is not
No Agency List award has been issued to any agency. The rule above is settled, and nothing anywhere on this site should be read as an award until one exists.
No per-agency standing is published either. Category and location pages state how many agencies in that cohort have been assessed and the strongest band a cohort of that size could support; they do not yet say where any individual agency sits in it. Individual placements are held back until each cohort’s leading rows have been read page by page against the evidence they cite, because a placement is a published claim about a real company and an automated check is not enough to make one.
Ranking position, cohort membership and the score itself are not for sale, and paying us changes none of them. Paid placement exists, is labelled as paid wherever it appears, and sits outside the ranked list.
Every assessment carries a date
A score without a date is a claim nobody can check. Every profile carrying a score names the month it was assessed — and, where we hold the record, how many pages that assessment read. The score reflects what the agency had published by that date. Anything published since has not been read.
This matters more than it looks on a page listing several agencies. The directory is assessed in waves, so two agencies sitting next to each other can rest on reads weeks apart. The per-agency date is what stops that being misleading, and it is why we do not present the ranking as a snapshot taken on a single day.
The assessments currently published were made between August 2026 and September 2026.
Cohort standings are recomputed when any member of the cohort changes, and a listing page carrying a cohort claim names the month that cohort was last computed.
What the score does not measure
ALS measures how well an agency demonstrates its capability publicly; it does not measure client outcomes.
It does not measure client satisfaction, price, or availability. It does not measure whether the work was any good. It cannot: we read what an agency chose to publish, and an agency that does excellent work and writes none of it up will score below one that does ordinary work and documents it carefully. That is a real limitation of the method and not a subtlety of it.
A few more things worth knowing before you rely on a number here.
The top of the scale orders agencies; it does not separate them
Scores at the top sit within a point or so of each other. Right now the twenty highest published scores span 1.13 points in total. Treat the difference between two high scores as noise, and treat the band with its denominator as the claim. It is why a published position reads “top 10% of 340” and never “#3 in Toronto”.
One reading of a site is a sample, not a measurement
Bot protection is intermittent and servers fail. A site that refuses us today may answer tomorrow. When we cannot read a site we record that we could not read it; we do not record it as an agency with nothing to show.
Measured 20 August 2026
Re-reading the same sites twice with the same code moved about 2.2% of agencies a full rung. The sites had not changed; the reading of them had. Anything we publish as a movement or a trend would need agreement across passes, so we do not publish one.
Measured 20 August 2026
Re-fetching the same 100 sites a second time changed the outcome for 3 of them, and hand-checking found only one was our crawl. The other two were the agency's own server returning an error at the moment we looked.
Measured 19 August 2026
In a random sample of 100 listings that had signed up, 9 were not agencies by our own criterion — they sell a product or a service to end customers rather than agency services to businesses. So a cohort is a set of listings, and some of them do not belong in it.
How much of the directory is assessed
2,203 agencies carry a published score. 811 were assessed and no published client evidence was found, so no score is shown for them, and 336 could not be read at all. The directory lists 60,464 agencies, so most of it is not yet assessed and shows no score. A cohort’s denominator is the assessed population, never the directory.
If we have this wrong
An assessment rests on a machine reading a website, and machines reading websites get things wrong. If a score, a standing or the evidence behind one is wrong about your agency, tell us and we will look at the page again.