Learn

Concepts

What each number, chart, and object in the product actually means.

13 articles

• AUC, honestly AUC is a ranking statistic, not an accuracy score. What the number on your model card actually says, and what it does not.
• Feature analysis, decoded SHAP impact, splits, stability, direction, and the gap flag. How to read the feature table and which column to trust when two disagree.
• How your data is handled What leaves your CRM and what does not · what the PII badge means · how an email becomes a domain and a token · how people match to accounts without guessing.
• In-market timing Predicting which months an account is most likely to enter a buying cycle, how confidence tiers are assigned, and what the beta label on it means today.
• Precision at the top Of the top 20% of records by score, how many actually converted. The hit rate a rep feels, and why it always needs the baseline next to it.
• Reading the score distribution What the histogram, the bands, and the median tell you about a model. Healthy shapes, degenerate shapes, and why a small hot bucket is usually fine.
• Rules and the per-rule readout Rules route records to models inside one score, first match wins · and the per-rule readout shows where the current model ranks well and where it does not.
• Scores, models, and runs A score is the durable object that writes to your CRM field · models are the engine underneath it · runs are the dated executions that produced every number.
• The lift curve What each slice of your ranked list converts at, read against the rate across all records. The chart that shows where the good records run out.
• The score itself What the 0-100 number is, how it is derived from conversion likelihood, why the per-record factors do not add up to it, and how far to trust the level.
• Value grouping Collapsing spelling variants of the same CRM value into one group, per field, with the map versioned and pinned to each model that trained on it.
• What "records trained" means Find each of the three record counts where it actually renders, set the train / test split where it actually lives, and read the gaps between them correctly.
• Where key insights come from The plain-language findings under your KPI cards are observed conversion rates by field value, not model output. What that difference changes.

How to

The tasks you perform · getting data in, defining success, activating a score.

9 articles

• Activation and refresh What activating a score does, what a scheduled refresh recomputes, what happens when writeback is not configured, and what pausing keeps.
• CSV, CRM, and sample data The three ways data gets in · synthetic sample data, a CSV upload, or a connected CRM, and exactly what changes once the data is really yours.
• Defining success Set the success definition control by control · what defines success, the criteria, the success date field, and where non-conversions get cut off.
• Engagement features Set up the Engagement step control by control · sources, date field, dimensions, rate filters, windows, and why only pre-cutoff activity counts.
• From CRM to score The whole path in one pass · connect, define a win, check the data, train, read the results, deliver. Plus what the cutoff date does to activity features.
• Set up an at-risk model The At-risk customers starting point, control by control · the cohort that makes it a churn rate, the renewal check, and the at risk · watch · stable bands.
• Set up lead conversion scoring The Lead conversion starting point, control by control · who it trains on, what the preflight checks, and how to read the result for a top-of-pipeline goal.
• Set up renewal and expansion scoring The Renewal / expansion starting point, control by control · the cohort it shares with at-risk, the renewal-typing check, and when it cannot run at all.
• Set up won-business scoring The Won business starting point, control by control · why it trains on open deals only, what the preflight checks, and how to read win likelihood.

Integrations

Where scores land outside ax1om · CRM fields your reps read, and the live API.

2 articles

• Score factors in your CRM Per-record top reasons, how they are derived and rolled back to your own field names, and how to write them into a CRM field reps already read.
• The live scoring API Scoring records on demand · keys and test keys, the monthly meter and its grace band, and what an enriched flag of false honestly tells you.

Troubleshooting

The checks and flags that say something is off, and what to do about each one.

3 articles

• Dataset health checks Read the diagnostics wall block by block · the two counts, field signal, redundant fields, and the recommendations, with the action each one calls for.
• Redundant fields Read a Redundant fields card, decide which of the two to keep, remove the other, and re-run · plus when leaving the pair alone is the right call.
• Target leakage, and why we flag it What the Leakage risk badge and the orange banner are each telling you, why a flagged field may not be in your selection at all, and what to do in each case.