How the standing score works

Your standing score is the number a lender will look at when they decide whether to extend you credit. The tier is the band it falls into. Everything else on the platform — matching, alerts, advance offers — uses the same signal in the background. This page is the model from a user’s perspective. Architectural detail lives at ML layer → Standing model.

​

The four tiers

The thresholds aren’t strict — the model is a classifier, not a rule engine. But the rough boundaries above are the practical guide for what gets you to the next tier.

​

Building (0–299)

You just joined. The model doesn’t have enough signal to commit. This is the “show us what you can do” tier. Most Job Seekers spend two to four weeks here. You can browse and apply to jobs from this tier. You can withdraw earnings to your bank (once you’ve completed Tier 2 KYC). You can use the wallet normally. You just can’t yet take credit through the partner lenders.

​

Emerging (300–549)

You’ve shown up. You’ve completed jobs. Reviews are decent. This is where most users are after their first month or two. Lenders rarely underwrite at this tier. Employers see you as a serious candidate.

​

Trusted (550–749)

You’ve earned trust through consistent work. Your reviews are solid (4.0+). Your on-time rate is high. Employers prioritise you in matching. This is where credit becomes accessible. Partner lenders extend tool-equipment loans for Job Seekers and inventory advances for traders.

​

Established (750–1000)

The top of the curve. Long tenure, very high completion rate, excellent reviews. The matching model surfaces you to employers first. Partner lenders extend their best terms. A small gold ring appears around your standing badge in the dashboard. (Yes, that’s a deliberate visual cue.)

​

The signals

The model reads the ledger and weighs a small number of signals. For traders, the trader-specific signals are: The model’s calibration is regularly updated as more real-world data flows. The signals listed are the headline drivers; the model uses subtle interactions among them that don’t lend themselves to a flat table.

​

How the score updates

The model recomputes lazily — on the next dashboard read after a material event, or on the nightly batch retrain pass. Material events that invalidate your tier cache:

• A new job is marked completed.
• A new review is left for you.
• A loan repayment lands or fails.
• A direct debit (savings or repayment) fails.
• Significant time passes without activity (a quiet stretch dampens recent-activity signal).

You see the tier in real time on the dashboard. The chart shows the trajectory over the last 90 days.

​

When the model is wrong

The model can be wrong. Three common cases:

1. A genuinely good Job Seeker has a tough month. Bad weather closes work sites; family emergency pulls them off. The score drops. The next month of consistent work recovers it.
2. A new Job Seeker who’s genuinely excellent. The model is conservative on thin history. The fastest path up is completing a number of jobs in quick succession, all with good reviews.
3. An employer leaves a bad-faith review. Rare but happens. You can flag a review for moderation; if it’s removed (e.g., the employer is found to have been retaliating against a legitimate dispute), it stops affecting your score.

The model doesn’t claim to be infallible. It claims to be calibrated to Nigerian informal-economy patterns and to improve with more data.

​

How a lender sees the score

When a lender pulls your signed snapshot (after you grant them consent), they see:

• Your tier (Building, Emerging, Trusted, Established).
• Your raw score (0-1000).
• The signals behind it (counts, averages, ratios).
• An aggregated view of your audit ledger window (totals, counts).

They make their own underwriting decision. Some lenders weight the tier heavily; some weight specific signals (repayment streak especially, for repeat borrowers). The snapshot gives them the signal; their model decides. You don’t see the raw score on your dashboard — only the tier and the underlying signals. This is intentional: chasing a single number invites optimisation behaviour that doesn’t match real-world performance. Focusing on the signals (complete the work, show up on time, get good reviews, save consistently) is the right behaviour.

​

What’s not in the score

A few intentional exclusions:

• Credit-bureau lookup. We don’t pull your CRC or FirstCentral report. The score is built entirely from your Guild activity.
• Demographic data. Age, gender, ethnicity, religion — none of these affect the score. The model doesn’t even see them as features.
• Friend-of-friend or social graph signals. Your standing isn’t affected by who your friends or family are.

The score reads work; that’s the design constraint.

​

What to read next