Dashboards

A research quickly grows to thousands of patents and hundreds of products, and no single row tells you how a field looks as a whole. Is the technology heating up or cooling off? Who owns most of it? How much of a portfolio is still in force, and where does the infringement risk actually land? These are questions about distributions, trends and breakdowns, and a dashboard answers them with charts.

A dashboard visualizes the records of a research as a collection of charts. Where a table is for reading a dataset row by row, a dashboard is for seeing its shape. Each chart counts the patents or products in your research and draws them in a specific form, and the charts redraw themselves as records arrive and get mapped. Here is an example dashboard that answers four questions at once:

One look shows that:

• filings have risen steeply since 2018,
• most of the portfolio is still in force,
• one applicant leads the field, and
• a large block of recently filed applications is still working through examination.

Chart types#

We offer thirteen chart types, each suited to a different question. A single number answers "how many". Bar charts rank and break down. Line and area charts follow a count over time. Pie, donut and ring charts show how a whole splits into parts. A radar compares several categories at once. A funnel shows how a set narrows stage by stage. Sankey and sunburst charts trace how one dimension flows or nests into another. A graph draws a network of connections, e.g. inventors who work together or patents that cite each other. A map lays the count over world geography. Here is each one, drawn from the same research.

Single number#

A stat answers the simplest question of all: how many? It shows the headline count of whatever the source holds, without any grouping. Use it when one figure is the whole point.

Bar charts#

A bar chart lines categories up so you can compare them by length. It is the natural answer to "who files the most" on patents, or "which manufacturers turn up most often" on products, using any select or multi-select attribute the products carry. With long labels the bars turn horizontal, and a limit trims the list to the leaders.

Line charts#

A line traces a count across time, e.g. filings climbing year after year, so it shows the momentum of a field.

Area charts#

An area chart follows the same count over time but fills the space underneath. Split it by a second dimension and the bands stack up to the total. This one breaks each year down by status, so you read the growth and the mix at once.

Pie charts#

A pie shows composition, i.e. the share each slice takes of the total. Keep it to a handful of slices and it answers "how is this split" at a glance.

Donut charts#

A donut is a pie with a hollow center, a calmer take on the same composition.

Ring charts#

A ring plots the same composition as a set of concentric arcs, each filling the share it takes of the whole. It looks like a stack of gauges, so an exact proportion is easy to read off arc by arc.

Radar charts#

A radar spreads one dimension around a wheel of spokes and traces a shape across them. It answers "where is this portfolio strong" at a glance: the further the outline reaches on a spoke, the more the set concentrates there.

Funnel charts#

A funnel stacks stages largest first, so you see the drop-off at each step. Use it to follow a set as it narrows, e.g. every filing at the top and only the granted and in-force ones at the bottom.

Sankey charts#

A sankey traces how records flow from one dimension into another, with the width of each ribbon set by how many records follow that path. What makes it special is that each column can draw from its own source, so one chart can cross from patents to products. This one runs patents into the features their claims cover, and those features into the products that carry them, which shows you where the risk lands.

You can connect the two sides through any unscoped attribute. Because patents and products share one option pool, a patents column and a products column meet on the same nodes: create a multi-select called Features with no object set, and both sides draw on it. Pick each column's source in the axes, and use the record axis to draw one node per patent or product instead of a grouped dimension.

Sunburst charts#

A sunburst nests a hierarchy in rings: the center splits into broad groups, and each of those into finer ones as you move outward. It shows composition at every depth at once, e.g. a field and then the subclasses inside it.

Graphs#

A graph draws the connections in your research as a network of nodes and links. It answers questions about relationships, e.g. who works with whom, which technologies get filed together or which patents build on each other. You simply pick what connects:

• Co-occurrence links the values of one attribute that appear on the same record, e.g. inventors who file together or CPC subclasses that get assigned to the same patents. The more records two values share, the thicker their link.
• Two attributes links the values of one attribute to those of another, e.g. applicants to the CPC classes they file in.
• Citations links the patents in your research that cite each other, and each link points from the citing patent to the cited one.
• Mappings links patents to the products or reference patents they were charted against, and the higher the mapping score, the thicker the link.

Nodes can be grouped by an attribute, e.g. each inventor by the assignee they file for most, or by communities that we detect from the network, i.e. sets of nodes that are linked more closely to each other than to the rest. The "Network" layout draws all nodes as one network and uses the groups only for color. The "Groups" layout draws each group as its own cluster and dashes the links that run between two groups. The nodes at either end of a dashed link get a ring, so you can spot the ones that bridge two groups. This one groups the inventors into three communities.

To pan, drag the chart, or drag with two fingers on a touch screen. To zoom, pinch or hold Ctrl or Cmd and scroll. Once you have clicked into the chart, you can also zoom by just scrolling, until your pointer leaves the chart again. To select a whole group, click its name in the legend, click an empty spot inside its cluster in the "Groups" layout or hold Shift and click one of its nodes.

To get started, simply pick one of the graph templates under "Add chart": "Inventor collaboration", "Citation network" or "Technology co-classification".

Maps#

A map shades each country by how many records land there, so geography does the grouping. On patents it answers where a portfolio is protected. On products it answers where they are sold, using a multi-select attribute whose options name countries, e.g. the Territories column the research agent keeps.

Patent maps read in two ways. Filings counts each publication once under the office that issued it, so a regional publication like an EP belongs to no country on the map and is listed beside it instead. Coverage asks the more useful question: it reads each patent's designated states and shades every country the patent actually reaches. One EP publication validated in eight countries shades eight. For coverage, you can pick which legal statuses count, with In-force on by default and Pending, Lapsed, Expired, Revoked, Rejected and Unknown available. In short, filings show where a company files and coverage shows where it can stop you.

Building a chart#

Every chart is assembled from the same few pieces: a source of records to count, optional filters that narrow what it counts, the axes that decide how it is drawn and, on some chart types, a limit that trims and orders the result. A chart created from a table column via "Visualize as chart" arrives with all of these already set.

Data source#

The source is the set of records a chart counts. Patent charts are drawn in aqua, and pointing the same chart at your products turns it orange.

• Patents: count and group the patents in your research.
• Products: chart your products instead, by any select or multi-select attribute they carry, e.g. category or manufacturer.

Every chart type works from either side, with a few exceptions. Line and area charts are patents only, since a product carries no date to run a timeline on. Citation graphs and graphs of mappings against reference patents are patents only too, since they connect nothing but patents. A graph of mappings against products draws both sides at once, so there the source just decides which side your filters apply to.

Filters#

Filters narrow which records a chart counts, separately from how it groups them. They are the same filters you use on a table or in search. While you can edit the dashboard, they show as chips above the chart so you can see what has been left out. A read-only viewer sees the same filtered chart, just without the chips. A chart created from a table's "Visualize as chart" carries that table's active filters with it.

Axes#

The axes decide how a chart is drawn. The x axis is the dimension you plot: an attribute (applicants or status on patents, any select or multi-select attribute on products, or one you added yourself) or a date. For a date, a second control rolls the timeline up, by year, by quarter or by month.

The y axis is always a count of records, so every point on the x axis shows how many. To compare two dimensions at once, replace "No breakdown" with a second attribute and the bars or bands split into stacked or side by side series. On a radar the same control reads "No overlay", since it lays one outline over another rather than stacking.

Pies, donuts and rings trade the x axis for slices. The relational types chain dimensions together: a sankey into columns, a sunburst into rings and a graph into a network of nodes and links. The underlying choice is always the same: pick a dimension and count the records.

Beyond attributes and dates, two more axes are worth knowing by name. The mapping score axis reaches across the two sides of your research: plot patents or products on it and each record lands in a band by its best match against the other side, so you can chart how exposed a set is without leaving the source you are on.

The record axis, offered as "Each patent" or "Each product", draws one node per record instead of grouping them. It only makes sense in two places. On a sankey, it lets a flow start at individual patents and end at individual products, and in a graph of two attributes, it links each patent or product to the values it carries.

Chart options#

A few controls belong to particular chart types. Subdivided bar and area charts choose between "Stacked" and "Side by side". Line and area charts can show a "Running total" or the "Value per period". A bar chart can switch between "Vertical bars" and "Horizontal bars", and long category names fit better on horizontal ones. A sankey takes up to six columns via "Add stage" and a sunburst up to four rings via "Add ring". A graph lets you pick what connects and how its nodes group ("No groups", "Communities" or an attribute), and once they are grouped, it can switch between the "Network" and "Groups" layouts. It leaves out nodes without a single link unless you turn on "Show unlinked nodes".

Limits#

Most chart types hold more values than fit on screen, so they take a limit that keeps the leaders and folds the rest away: "Top 5", "Top 10", "Top 20" or "Show all". A sort then orders what remains: "Highest first", "Lowest first" or "A → Z".

Not every chart has a limit. A stat is a single figure, and line, area and map charts have none either. Neither does a bar chart plotted on a date, since a timeline already runs its whole range. The sort has exceptions too: a radar keeps its limit but always draws its spokes in a fixed order, a graph has a limit but no sort, and a mapping score axis is always ordered by band, from critical down to low.

A graph counts its limit in nodes, from "Top 50" up to "Top 10,000" or "Show all". A graph of two attributes or of mappings applies it to each kind of node, so it reads e.g. "Top 250 per kind". Where a link counts shared records, you can also set a minimum link weight to hide the weak ones: the control reads "All links" until you pick e.g. "2+ shared", which keeps only the links with at least two records behind them. A mapping graph weighs its links by score instead, so there the control reads "Any score". Pick "Scored only" to hide the mappings without a score, or "61% or more" to keep only the medium, high and critical matches.

Titles#

A chart names itself. Patalyze reads the configuration and writes a matching title, so an applicants axis becomes "Top applicants" and a subdivided chart becomes "X by y". You can rename a chart from its menu, but be aware that the next change to its axes writes a fresh title over yours, so set a custom name once the chart has settled.

Arranging a dashboard#

"Add chart" at the top of a dashboard offers a set of templates such as filings over time, filings by status, top applicants or a classification breakdown, and you can configure any chart from scratch. You can drag charts into a new order. The handle in a chart's corner resizes it on a grid of twelve columns and up to six rows, so a headline number can sit small beside a wide timeline. A chart's menu offers "Rename" and "Delete".

Reading a dashboard#

A new dashboard is seeded from what the research already holds. An empty research starts blank. A research with only patents gets six charts: the total number of patents, filings by status, top applicants, the status distribution, a classification breakdown and a bar chart of countries. A research with only products gets a total plus one breakdown per select or multi-select attribute its products carry, up to four, e.g. by category and manufacturer. A research with plenty of both gets a mix. If you keep a shared Features attribute, the patents-to-features-to-products sankey is added on top. Everything after that is yours to change.

Most dashboards then lean on the same handful of questions. Once you can read these four, you can read almost any dashboard:

• Filings over time. A line or area chart of patents by date shows whether a field is heating up or cooling off. A steep, late climb means most of the art is recent, and a long block of pending filings is still working through examination.
• Top applicants. A bar chart grouped by applicant names the companies that own the space and shows how far ahead the leader sits.
• Portfolio breakdowns. A pie or donut split by status, territory or classification shows the make-up of the set: how much is in force, where it is filed and which technologies it concentrates in.
• Risk distribution. Chart patents or products on the mapping score axis to see how their best matches stack up. A single critical match carries far more weight than a long tail of low ones.

A chart is also a way into the data. Click a bar, a slice, a country, a funnel stage, a ring, a sunburst segment, a sankey node or flow, or a graph node, link or group, and Patalyze opens a table panel with exactly the records behind it, already filtered. You can read them there, download them, or keep the selection as a new table with "Save as table". So when you see a spike in one year, you can click it and read the filings that caused it.