Tomasz Gęstwicki

Causal / Lucanet

Causal is a tool for financial modeling and simulation. It gives simple access to Monte Carlo simulation but breaks the conventions of the familiar spreadsheet, so despite a clean interface and many tutorials, people used to spreadsheets could feel lost at first. The main challenge was to understand their mental models across many use cases and to answer them with new interface solutions.

period
2021–2023
through
flow/control
with
Szymon Kaliski and the Causal team
my part
Interaction flows, prior art, interface explorations, final UI

The representation problem

Variables used to be large rounded cards, each with its formula in a popover. For people used to spreadsheets, this made a model harder to read: neighboring values were out of view, references to earlier periods felt indirect, and many operations were hard to find.

The earlier Causal interface: variables as large cards, with a formula editor open in a popover and charts beside it.
Fig. 1 -- Earlier model: variables as cards, formula in a popover.

How we explored it

In the first workshops I learned how Causal runs its simulations, went through the more complex use cases and read user feedback on time-series modeling. We rebuilt the full user flow, element by element, and interviewed a small group of users about how they think of time in financial planning. Then came prior art, from research papers to other software, and many quick experiments in different directions — code sketches, wireframes, user flows — to test ideas and ask better questions.

A map of interface explorations linked by arrows: nested lists, a grid, a row view, time-based views and autocomplete variants.
Fig. 2 -- Solution map: the directions we explored.
Close-up of the map: the trello-like, grid, row view, separate input cells and formula bar variants.
Fig. 3 -- Detail of the solution map.

The revised interaction model

Variables now sit in rows, with time across the columns. You edit a formula where it is, and references to other variables appear inside it as small tokens. Editing and presenting are separate views of the same model. Neighboring values and earlier periods stay in view, the way people used to spreadsheets expect.

The revised Causal interface: variables in rows, months in columns, formulas with green references to other variables.
Fig. 4 -- Revised model: rows, time across columns, formulas edited in place.

Three decisions

The expression editor with a menu of operations: variables, formulas, recipes and data sources, with growth templates below.

Formula assistance

Users could not tell which functions existed without opening the help, which knows nothing about what they are doing. Autocomplete and autocorrection now help with formula semantics, and a small menu at hand offers ordered, labeled templates.

A formula with green variable tokens, each with a time selector such as “current” or “t - 3”, and autocomplete suggestions below.

Time and previous values

In interviews, “prev” felt confusing, especially to people coming from Excel: it is indirect and hides the value it stands for. Time references became explicit — relative (previous, t-12), fixed (final) and ranges (t-6:t-1) — so a formula says which period it reads.

A chart wireframe with an editable title, axes, granularity settings and a legend of variables and scenarios.

Charts

Any variable can be charted from its cell, or through a wizard with instant previews. On the chart, range selectors let you pick and compare periods directly.