Asset custody
Asset tagging and custody for fixed assets: who has each asset, since when, and under which term. Issuing it opens custody (check-out); returning it closes it (check-in). It doesn't change the asset's record — it only registers custody.
What you find on this screen
A portrait of the screen, read from the screen itself — it changes when the screen changes.
The list shows
SegmentUplift (p.p.)95% CInClass
Form fields
NomeComo o experimento é chamado; é por ele que a decisão é citada depois.
MétricaO desfecho medido pelo motor, como taxa de renovação ou conversão de proposta.
Variante AO braço de controle: o que já se fazia antes da mudança.
Variante BO braço de tratamento: a mudança cujo efeito se quer isolar.
Efeito mínimoDiferença mínima que vale detectar. É ela que dimensiona a amostra.
AlfaRisco aceito de enxergar efeito onde não há.
PoderChance de o motor detectar o efeito quando ele existe.
Amostra necessáriaAmostra total que o cálculo de poder pede para a leitura valer.
Amostra coletadaQuanto o motor já coletou. Enquanto estiver abaixo do necessário, a conclusão é provisória.
ResultadoA leitura descritiva que o motor escreveu, com intervalo de confiança e tamanho de amostra.
SituaçãoRascunho, em campo ou concluído.
DecisãoA decisão registrada ao fechar o experimento. Só é preenchida quando ele conclui.
Start here
O que esta tela responde
Whether the change you made CAUSED the result, or whether it would have happened anyway. It's the question that a before-and-after chart never answers on its own.
What this screen is
What would have happened IF. Tenka has a panel of companies with STAGGERED ADOPTION (each one changed price, cadence or team at a different moment) — a natural experiment. Difference-in-differences and synthetic control estimate the EFFECT of an intervention by comparing those who changed with those who have not yet changed, before and after; uplift shows in WHOM the action changes the outcome (persuadables) and in whom it does not (they would have stayed anyway). 🔴 Every output comes with an INTERVAL (95% CI), a stated n and the warning that it is an observational estimate — correlation is NOT an order: the engine measures, it does not tell you to raise prices. Visible, deterministic math. Read-only.
How to
Os três métodos, em uma frase cada
Difference-in-differences: compares the change among those who adopted with the change among those who did not. Synthetic control: builds an artificial "twin" of your company from those who did not adopt. Uplift by segment: shows in WHOM the effect was greatest — because an average of zero can hide one group that gained a lot and another that lost.
Persuadível, indiferente, contraproducente: as classes do uplift
Uplift compares the probability of the outcome between treated and untreated, segment by segment. Persuadable is the segment in which the action changes the outcome — that's where it pays off. Indifferent is whoever would have the same outcome with or without the action: spending there is waste. Counterproductive is the segment in which the action worsens the outcome — the case that most justifies looking at the breakdown before generalizing. Each row shows the uplift in percentage points, the 95% CI and the segment's n.
Rules
What this screen answers
If the whole market grew in the same month, your growth proves nothing about your change. The screen's three methods exist to build the missing comparison: what would have happened to those who did NOT change, in the same period.
Leia o intervalo antes do número do meio
Every result comes with a 95% CI and the stated n. An effect of "+12%" with an interval from −5% to +29% is not a 12% effect: it is information that you still cannot even assert the sign. When the sample is small, the screen shows the number anyway — with the caveat in plain sight, so you can disagree with it.
O motor MEDE; ele não manda
Nothing here changes price, campaign or process. The screen describes the measured effect; the commercial decision remains with the people accountable for it.
When the panel says "methodology example"
If the yellow banner "Methodology example — this is not your data" appears, the numbers shown are an illustrative sample to show how the engine reads an effect: the math (difference-in-differences, synthetic control, uplift) is real, but the reading does not describe your operation. In that state, use the screen to understand the method, not to decide. With enough data of your own the banner disappears and the panels start reflecting your history.
Frequently asked questions
Which question this screen answers
What the real effect of a change in price, cadence or team was — measured, with a stated confidence interval (95% CI) and n. The screen shows the same question through three lenses: difference-in-differences and synthetic control estimate the average effect of the intervention; uplift by segment shows in WHOM the action changes the outcome. Everything is an observational, read-only estimate: the engine measures, it does not command — correlation is not an order to raise prices.
Was this article helpful?
In the same module: The company over time5
Extension revenue shareThe apportionment of third-party extension revenue by PAID USAGE: for each extension (and its developer), per…Purchase ordersThe commitment made with the winning supplier: items, quantities, amounts, and lead time. The NF-e receipt ma…Stock movementsThe stock ledger: receipts, issues, adjustments, and transfers. A receipt recalculates the average cost; an i…Purchase quotesRequest pricing from N suppliers and compare them on a grid (rows = items, columns = suppliers, cell = price)…NF-e receivingImport the purchase invoice's XML: Tenka reads the items and amounts (a namespace-tolerant parser), you check…
Tenka Academy
Explore the AcademyStep-by-step learning paths for your industry, inside the platform.