How this variant's sales have moved when its price moved

What past price changes did to this variant's sales velocity, as a tier with the evidence behind it.

tier is HIGH, MODERATE, LOW, INCONSISTENT, STILL_LEARNING or null. confidence, n_events (the qualifying price changes counted), sign_agreement, shrunk_effect_up and shrunk_effect_down (the effect of a rise and of a cut, as signed fractions) and data_through (the last event analysed) are the figures under it. worth_considering and summary are those figures written as sentences.

A variant with no classification yet answers every field null with n_events: 0. That is the honest answer, not an error, and not a tier of zero. An id no variant has answers the same shape rather than a refusal, so this read cannot tell you an id is wrong. Check it against a variant read.

basis and the fields beside it say why a tier is what it is: periods_observed, price_variation against required_price_variation, expected_error, and a reference_elasticity with its reference_basis and reference_source where the variant's own history is too thin. A tier of STILL_LEARNING with a price_variation well under required_price_variation means the price has not moved enough to measure, not that the reading is pending.

engine names the run behind the tier: state is not_run, current, borrowed or insufficient, the figure's state and never a run's, with computed_at (the run's, never your clock), inputs_through, stale_after, basis and run_id.

Scope: products:read.