100 companies, six sectors, 108,905 scored stories.
In all six sectors on mrsindex, the company that gets the most coverage is not the company whose coverage reads best. Not once.
Goldman Sachs is the most-covered bank on the index, on 2,748 stories in 90 days. On how that coverage reads (reputation impact) it is 18th of 20. Fidelity Investments is last of 20 on coverage and first on how it reads. Across the whole index, 63 of the 100 companies sit at least a quarter of their sector apart on the two.
If you lead communications at a large company, none of that surprises you. The industry retired advertising value equivalency in 2010, when the Barcelona Principles said out loud what everyone already knew: counting clips and pricing them like ads measures activity, not value.
Volume, reach and sentiment are the same kind of number. They are outputs, and the business is paying for outcomes. They tell you how much you were covered and roughly how it felt. They cannot tell you what the coverage was about, or whether that is what the business needed it to be about. If those three are still the headline of your reporting, you are where AVE users were in 2010: measuring activity and calling it value.
I have been making that argument for years. And the trouble with an argument is that it stays an argument: our field is full of well-reasoned positions on measurement, none of which you can check.
So I stopped arguing and built the instrument. One hundred companies, six sectors, every article scored on its own. A company's score is the average of its articles, so volume counts nowhere: a three-thousand-story company and a hundred-story company are judged the same way.
Not to find out whether the premise was true. To find out whether it would hold up in public, with every article published so anyone can take it apart.
It held. That is the table below, and it is the least interesting thing here, because what I did not expect is what came out of building it. A structural bias against retailers and carmakers that I had to find in my own scoring and remove, and an eighteen-point gap between sectors that no communications team can close.
I'm Evan Escobedo. I lead measuredI/O and wrote Reputation Intelligence, and I spent years building communications measurement inside large companies before that. Signals of Reputation is my monthly read on how the largest companies are actually covered, and on what that coverage says about them. Each edition takes one finding from the MRS® Index at mrsindex.com and answers three questions, because those are the only three a reputation report has to answer: what changed, why it matters, what to do.
It runs on GDELT, an open dataset anyone can query, and the method is published in full. This is the MRS® Public Signal, an open-data version of the model in the book; the full four-pillar MRS®, human-validated and weighted to a company's own priorities, runs on licensed coverage and is a different number.
A note on what this is not. Most corporate reputation research asks people what they think of a company and reports the average. That work is valuable and this does not replace it. It answers a different question. Opinion research tells you what people carry; coverage analysis tells you what they are being handed. The second is the one a communications team can act on this quarter, and it is the one I built, because it is the half that is measurable from the outside.
What you get every month: one finding from the coverage of 100 companies, the working behind it, and a board you can look your own sector up in. From next month, each edition also shows what moved and why. No gate, no sign-up wall, no pitch.
This is the first edition built on the index: the MRSIndex 100, six sectors, 108,905 scored stories from 496 outlets (156,378 placements, syndicated copies counted once) over the 90 days to 31 August. Here is the most-covered company in each sector next to the one whose coverage reads best. Every row opens to its articles at mrsindex.com.
| Sector | Most covered | Stories | Reads | Reads best | Tone | Coverage rank |
|---|---|---|---|---|---|---|
| Banking | Goldman Sachs | 2,748 | 18 of 20 | Fidelity Investments | 70.4 | 20 of 20 |
| Automotive | Toyota | 2,859 | 7 of 18 | Renault | 65.5 | 14 of 18 |
| Retail | Walmart | 2,214 | 8 of 15 | Best Buy | 70.7 | 8 of 15 |
| Tech | Alphabet | 25,574 | 15 of 20 | Salesforce | 68.6 | 20 of 20 |
| CPG | PepsiCo | 659 | 8 of 14 | Kraft Heinz | 75.1 | 8 of 14 |
| Pharma | Eli Lilly | 875 | 2 of 13 | Vertex | 56.4 | 11 of 13 |
Tone is how the coverage reads: every article scored on a 0 to 100 scale, averaged. Ranks are inside the sector. Alphabet is the most-covered company on the whole index, on 25,574 stories, and reads 15th of the 20 technology companies on it. Both of those are true at once, which is the whole point: you can open the row and read the coverage behind each number.
How to read this. Coverage is how many articles were written about a company in openly published news over 90 days, from a vetted list of credible outlets. The wires and paywalled papers are thin in open data, so this is a directional read, not a census. Every limit is spelled out at mrsindex.com, and in the fine print below.
The table is the part you already knew. The next three findings are the part most reporting still misses.
Volume is not neutral, and it is not neutral in a direction. The more a company is covered, the more of the world's bad news its name is standing next to. That is a familiar observation. What is less familiar is testing your own index for it. I did, and it was there.
Articles where the company was the location of someone else's story were being scored as the company's reputation. A crime in a Dollar General parking lot. A stolen Mercedes. A man who crashed a Tesla into a shop. Violent vocabulary, so machine-derived tone reads them as catastrophic, and the company's name is in the headline either way.
Index-wide it was 3.2% of scored articles. That sounds survivable. It is not, because it does not land evenly:
| share of its scored coverage | how those stories read | how the rest reads | |
|---|---|---|---|
| Dollar General | 16.3% | −6.3 | −1.2 |
| JPMorgan Chase | 13.1% | −5.1 | −1.1 |
| Dollar Tree | 12.9% | −6.5 | −0.6 |
| Nissan | 9.4% | −6.0 | −0.9 |
| Honda | 8.2% | −5.6 | −0.5 |
| Mercedes-Benz | 7.7% | −6.0 | −0.6 |
Companies with stores were being charged for crimes committed on their premises. Companies that make cars were being charged for how people drive them. Sentiment is 35% of the score, so this was a structural penalty against two sectors, and it was invisible in any aggregate view.
This is the trap in letting a model build your index unsupervised. The machine tagged every one of those stories correctly: the company is named, the tone is hostile, the theme is trust. It has no way to know the company was the car park rather than the subject. No amount of model quality fixes that, because it is not a language problem, it is a judgement about what counts as the company's coverage. Somebody has to make that call and write it down.
Removing it moved 77 of the 100 companies, and every company that rose is a retailer or a carmaker.
That is the finding, and it is not really about me. An index that never tests this measures traffic accidents and calls the result reputation against a brand. If you run a score, this is the question to take to whoever built it: when my company is the setting for someone else's bad news, does your number go down?
Your baseline is your sector, not the market. Coverage of a consumer brand reads warmer than coverage of a bank or a drugmaker. The median CPG company on this index reads 66.6 out of 100. The median pharma company reads 48.4. That is an 18-point gap that no pharma communications team can close, because it is the beat, not the brand. It is also why the overall column on the index ranks every company against its own sector's norm rather than against the market. If your reporting compares your sentiment line to a cross-industry benchmark, you are grading yourself on somebody else's reality.
One number hides the driver that decides your bad day. Every article is also tagged to the reputation driver it is about: trust, innovation, performance, responsibility, leadership.
Costco is the third most-covered retailer on the index, on 1,012 stories in 90 days. A dashboard built on volume, reach and sentiment would have told its team they had an excellent quarter, and on reach they had. Its overall position is 51st of 100. On trust it is 83rd, its own weakest driver, on 474 stories, while leadership sits at 21st and responsibility at 27th.
Sixty-two places between its best driver and its worst. Same company, same window, same articles. Both numbers are true, and only one of them survives if you report a single figure. Costco is not unusual in this: 33 of the 100 companies move 10 or more places when trust is scored on its own, each measured against its own sector.
Those three findings have something in common. No single metric could have produced any of them. Not one of volume, reach or sentiment, reported on its own, would have shown you the retailer being charged for its car park, the drugmaker that cannot out-run its beat, or the company whose trust number and overall number disagree by sixty-two places.
That is the case for a composite. It is the argument Reputation Intelligence is built around, and it is worth stating plainly here, because I keep being asked what one actually is.
A composite is one score, built from several named dimensions of the same coverage, scored article by article, and weighted to what the business cares about. The weighting is not a refinement you add later. It is part of the definition. A composite without it is an industry average; a composite with it is a measure of your company against your own priorities.
Four things follow from that.
The dimensions are named, so the number comes apart. What the coverage says about you, how credible the outlet is, how prominent you are in the story, how far it travelled. One figure at the top, and every layer underneath it visible. Three metrics reported side by side do the opposite: they let everyone in the room choose the one that flatters the quarter.
The narratives are scored separately before they are combined. Trust, innovation, performance, responsibility, leadership, or whichever set your business actually runs on. Costco's overall number is carried by leadership and responsibility while trust runs the other way underneath it. Both are true. Only the split tells you which conversation to go and have. That is the difference between a score and a diagnosis.
The weights are agreed before the quarter, not after. Weights chosen once the numbers land will always flatter them. Weights chosen first will tell you the truth about them, and they are the step where measurement stops being somebody else's standard and becomes an account of what your business said mattered.
An index is that method run on a fixed population you did not choose. The hundred were named before the scores existed, so nobody picked the list to suit the answer, and it puts you against the companies you will be compared to anyway.
Which is why mrsindex.com is public. A composite you can only see on your own data is a claim. A composite running on a hundred companies, with every article downloadable, is something you can verify.
Demote volume, reach and sentiment to inputs. Keep them. They feed the score and they have their uses. But they stop being the headline. Then have the weighting conversation with the business before the quarter starts, because that conversation is the measurement: it is where you find out whether your team and your CEO agree on what this year is supposed to be about.
Neither of those needs a new platform. Everything after it is a question for whoever builds your reporting, which is the next section.
Every company on the index, with every article and all five drivers, is at mrsindex.com. If yours is not on it, look up your sector and your closest competitors, and start with the Reputation Gap panel: it shows how far coverage and regard sit apart for every company in the sector.
And again: this runs on public GDELT data. Run the same method on your own licensed coverage and the picture is fuller, because the wires and the paywalled press are properly in it.
If your platform, your agency or your team is still reporting volume, reach and sentiment every month, you are being handed an answer to the first question and nothing on the other two. Ask for the rest.
Ask for a composite, weighted to your priorities. Not three metrics on three slides. MRS® is one published model; any model with named dimensions, weights agreed in advance and a floor under every number will do the job.
Ask for it broken out by narrative. Trust scored separately from performance, from innovation, from responsibility, from leadership. If the overall number and the trust number would tell your CEO two different stories, that difference is the reporting.
Then ask the two questions that separate a measurement system from a reporting tool. Does a negative story ever make this number go up? And when my company is the setting for someone else's bad news, does it go down? If nobody can answer either one, you have a report.
I hold mrsindex to the same standard, which is why the method, the audit log and every one of the 131,375 rows behind those stories are published rather than described.
Check the work.
Signals of Reputation is published by measuredI/O. The index, the method, the audit log and every scored article are at mrsindex.com. No gate, no sign-up, no pitch. Written by Evan Escobedo, author of Reputation Intelligence.
mrsindex.com: the fine print. Every figure is an MRS® Public Signal, computed from GDELT through BigQuery on earned coverage only, from an allowlist of outlets. Each article is scored on the same four pillars a client project uses, with the tags derived from open data rather than assigned by a human: sentiment from article tone, prominence from where the company is named, salience from how many tracked companies the story names, reach from the outlet. A company's pillars are the average of its articles, so volume raises nothing. A company needs 100 scored articles and 4 sources about it to appear; a sector needs 8 companies; a driver needs 12 articles. Each sector is rescored 200 times on resamples of its own coverage, and a sector lead is only called clear when it holds in at least 80% of them. The priority weights are illustrative and published. The full four-pillar MRS®, human-validated, key-message-aligned and weighted to a company's own goals, runs on licensed data and is a different number.