Analytics

Analytics shows you how your Mediastilo pipeline is performing, how much material it's finding, how on-brand that material is, and how it's turning into drafts over time. Everything is scoped to the brand you're currently in.

Analytics is a plan-gated feature. If you don't see it, check your plan in Credits & Billing.

Summary

The summary gives you the headline numbers at a glance:

How many signals Mediastilo has scored, and their average alignment score, a quick read on how well your sources match your Brand.
How many drafts you have, broken down across the draft lifecycle (suggested, in progress, ready, published).

If your average alignment score is low, that's a signal to sharpen your Brand or review your Sources.

Source breakdown

The per-source breakdown shows, for each of your sources, how many items it contributed and how many were included versus excluded in Signals. Use it to spot:

High-signal sources: lots of included items. Keep and lean on these.
Low-signal sources: mostly excluded items. Consider replacing them or tightening what they feed in.

How the coverage reads

This panel splits the stories your sources published into positive, negative, and neutral, and shows the mix as one bar.

Sentiment is about the story, not about what it means for you. A factory closing is negative for everybody; whether that story is worth writing about is what the alignment score already answers. Reading the two together is the useful move: a run of negative coverage that scores high on alignment is a story you probably want a position on.

Two things worth knowing:

Most municipal and administrative coverage is neutral, and that is not a problem with the reading. A council agenda has no sentiment.
The denominator is what we scored, not what we ingested. Sentiment is worked out during the scoring pass, so items that fell outside your daily scoring capacity have no reading at all. The note under the bar says how many stories the split covers.

On the Signals feed, positive and negative items carry a small chip. Neutral ones don't — a chip on almost every card is a chip nobody sees.

Timeline

The timeline charts your signals and draft activity over the last 30 days, bucketed by day or week. It's the quickest way to see whether your pipeline is producing steadily, and to correlate spikes with changes you made to sources or Brand.

What's being covered

This panel ranks the topics your sources keep returning to over the last 7 or 30 days, and shows how each one moved against the period immediately before it. An arrow up means a topic is getting more attention than it was; an arrow down means it's fading. Hover a topic to see its count in the earlier period.

A few things worth knowing about how it's counted:

It reflects the items Mediastilo analysed for you, not everything your sources published. Items filtered out before scoring aren't counted here, so treat this as "what came through your pipeline", not a census of the outlet.
Topic labels are shown in your interface language, but a topic your sources name in two different languages can still appear as two rows (for example sport and sports). We're tightening that.
Very broad section labels are left out on purpose. A tag that sits on a quarter or more of everything your sources published — sport, politics, lifestyle — names the desk the item came from, not what the story is about, and it would otherwise fill the panel every single day.

If your plan covers a shorter window than the one you picked, the panel says so and shows the period your plan covers.

Use it to spot a topic your sources are crowding into before you commit a draft to it, or one that's quietly cooling off.

Share of voice

Track the names you care about — your own brand, your competitors, a topic — and this panel shows how many of the stories your sources published in the window mentioned each one.

The share is each name's slice of the names you track, not of all coverage. A share of everything published would move whenever an unrelated source had a busy week, and would tell you nothing about the question you actually asked.

Add a name with the input at the bottom of the panel, and mark it as us or a competitor. Remove one at any time; its counts go with it.
A new name is counted back over the last 30 days by a background job, so it can read zero for a minute after you add it. The panel says so while that's happening.
Counting happens once a day, over the same items the rest of Analytics covers — what your sources published and Mediastilo read, not the whole web. It is not a press-clipping service.
Matching is on the name as written, anywhere in the headline, summary, or body, and on whole words — tracking HRT finds the broadcaster, not the German word Fahrt. A name that is also an ordinary word will still over-count, so a distinctive name is a better name to track.
Croatian declines nouns, so a name matches in the form you enter it: Večernji finds Večernji list but not Večernjeg lista. Enter the form that actually appears in headlines. Suffixed acronyms like HRT-a do match, because the hyphen ends the word.
A source never counts toward a name that belongs to it. A paper mentions itself in its own bylines and footers constantly, and counting that would describe its page templates rather than its coverage. A source is treated as its own if the name appears in its web address (24sata on 24sata.hr) or if one of the name's words is that address (Večernji list on vecernji.hr). Whole words only, so Sportske novosti is not silenced by sport.hr. A masthead sharing nothing with its own domain — Kronen Zeitung against krone.at — is still counted.

When a name gets unusually loud

Tracking a name isn't only a number to read later. If one of your tracked names turns up in far more headlines than it normally does, Mediastilo tells you the same day, in Waiting on you on the dashboard.

The comparison is against that name's own normal, not against a fixed number: it is its share of what your sources published that day, measured against the middle of the last fourteen days it was in. A busy news day lifts everything, so a raw count would cry wolf every time collection had a good morning.

Two things follow from that. A name that is almost never mentioned can't spike — there is no normal to be unusual against — and neither can one that is loud every single day. What sets it off is a name that is usually quiet and suddenly isn't.

Noticing costs nothing, and nothing is generated unless you ask. Following the notification takes you to Ideas, where you can turn the coverage into angles worth publishing.

Using analytics to tune your pipeline

Analytics is most useful as a feedback loop:

Low average alignment → refine Brand.
A source with mostly excluded items → fix or swap it in Sources.
Few drafts reaching Ready → revisit your alignment/auto-generate thresholds in Account & Settings.