Coding rules are the highest-leverage configuration decision in ReconLink. A well-built rule library means the majority of a client's transactions are classified the instant they land — no human touch required. For a busy practice managing 15, 20, or 30 clients, that compounds into hours recovered every week.
This guide walks through exactly how to set up, test, and grow a rule library in ReconLink — from first login to a mature automation rate of 70–80%.
How coding rules work in ReconLink
A coding rule is a pattern-match instruction. When a new transaction arrives, ReconLink checks the description against every rule in the library. If a match is found, the rule assigns the account code and GST code automatically — and the transaction is classified without going to any other layer of the engine.
Rules run before the machine-learning model and before the LLM fallback. This makes them:
- Fast — pattern matching is near-instant
- Deterministic — the same description always produces the same result
- Auditable — every automated classification links to the rule that produced it
Rules in ReconLink are practice-wide. A rule you build for one client applies to all clients whose transactions match the same condition. This is what distinguishes ReconLink from per-organisation tools — you invest once in building the rule, and every client in the practice benefits.
Understanding vendor key recognition
ReconLink normalises every transaction description through a vendor key algorithm before rules are evaluated. The normaliser strips bank-channel prefixes — EFTPOS, OSKO, BPAY, INTERNET, PAYPAL AUSTRALIA, and similar — and removes trailing reference IDs, invoice numbers, and date stamps.
The result is that transactions like:
BPAY ORIGIN ENERGY REF 847392018INTERNET ORIGIN ENERGY 10NOV26EFTPOS ORIGIN ENERGY MELBOURNE
all resolve to the vendor key ORIGIN ENERGY and match a single rule.
This matters when building rules: write your pattern against the core vendor name, not the full raw description. Patterns written against raw descriptions are brittle — they break when the bank changes a prefix or the reference number format changes. Patterns written against the normalised vendor key survive those changes.
Step 1: Build your starter rule library
Open the Rules Library
Navigate to Settings → Coding Rules in the ReconLink sidebar. Click New Rule to begin.
Choose the right match type
ReconLink offers three condition types:
| Match type | Use when |
|---|---|
| Contains | The vendor name appears somewhere in the description, but the full format varies |
| Starts with | The description always begins with the vendor name |
| Equals | The description is always identical (rare in practice) |
For most rules, contains is the right choice. It is tolerant of prefix and suffix variation without being dangerously broad. Reserve "equals" for highly specific, low-volume patterns where you need precision.
Assign the account code and GST code
Every rule requires:
- Account code — select from your chart of accounts (e.g., 6-1050 Electricity, 6-1200 Telephone, 4-1100 Trading Income)
- GST code — one of GST, FRE, INP, N-T, or CAP
Common GST coding decisions for Australian rules:
- Bank fees and interest charges → INP (input-taxed, no ITC available)
- ATO payments (PAYG, FBT, PAYG instalments) → N-T
- Superannuation payments → N-T
- Most utility and telco bills → GST (confirm supplier is registered)
- Fuel → GST (note: fuel tax credit rules apply separately)
- Insurance → GST (most general insurers are GST-registered)
Add a rule note explaining the GST decision. This note appears in the audit trail and is invaluable when someone reviews the rule in 18 months and cannot remember why a particular code was chosen.
Use the live preview before saving
The rule builder shows which recent transactions would have matched your pattern. Review this list before saving. If the match set includes transactions it should not, tighten the condition. Saving a rule that is too broad will miscategorise legitimate transactions silently.
Target vendors for your starter library
A first session of 60–90 minutes should produce 25–35 rules covering:
- Utilities: AGL, Origin Energy, Synergy, AusGrid, Jemena, SA Power Networks, Sydney Water, Yarra Valley Water
- Telecommunications: Telstra, Optus, TPG, Vodafone, Aussie Broadband
- ATO payments: ATO PAYG withholding, ATO PAYG instalment, ATO GST, ATO FBT
- Superannuation: ATO SUPER (SuperStream), individual super fund clearing houses
- Software subscriptions: Xero, MYOB, Microsoft 365, Adobe, Dropbox, Slack — all GST
- Banking: Monthly account fees (INP), international transaction fees (INP), bank charges (INP)
- Insurance: CGU, QBE, Suncorp, IAG, Allianz — confirm GST registration
These 25–35 rules typically handle 50–60% of transaction volume across most client types from day one.
Step 2: Extend the library during client onboarding
When you add a new client, review their first 30 days of transactions. Identify any high-frequency vendors that are specific to that client's industry and add rules for them.
A hospitality client will have regular payments to food and beverage suppliers. A construction client will have plant hire, materials, and subcontractor payments. A professional services firm will have industry association fees and CPD costs.
Most practices add 5–10 client-specific rules per new client. These rules help the entire practice if any future client in the same industry has the same supplier.
Step 3: Promote AI codings to rules
ReconLink's AI coding layer runs when no rule matches. When you accept an AI-suggested coding in the review queue, ReconLink can automatically create a rule keyed on the normalised vendor key for that transaction.
This is controlled by the Rule Promotion setting in practice preferences. When enabled, accepted AI codings above the confidence threshold (default 0.70) become deterministic rules — meaning the next transaction from that vendor codes instantly via Layer 1, without going through the AI at all.
The practical effect: you do not need to build rules manually for every vendor. Code new vendors through the review queue, accept the AI suggestions when they are correct, and the rule library grows automatically. Within three to six months, most practices find their rule library has doubled in size through promotion alone, and their automation rate climbs from 60% toward 80%+.
One exception to be aware of: if you edit the AI's suggestion during review (e.g., the AI suggested account code A but you changed it to B), promotion is bypassed for that coding. ReconLink infers that the AI was not accurate enough on that vendor to warrant a deterministic rule.
Step 4: Audit and maintain rules
Monthly rule-health check (10 minutes)
Filter the Rules Library by "Last matched: more than 30 days ago". These rules have not fired recently — either the vendor stopped transacting, or the description format changed and the rule is no longer matching.
For each zero-match rule, check recent transactions for that vendor to see if the description format changed. Update the pattern if needed.
Quarterly GST review
Any rule coded as N-T or FRE deserves a quarterly sanity check, particularly for smaller suppliers. A supplier that was not GST-registered when the rule was created may have crossed the $75,000 registration threshold. If your rule still codes them as N-T while they are now charging GST, you are missing input tax credits on every transaction.
Client-level overrides
Where a vendor has a different GST treatment for one specific client, use a client-level override rather than changing the practice-wide rule. Overrides take precedence over the shared library for that client only.
Putting it together
Coding rules are the foundation of a fast, low-effort reconciliation practice. A starter library takes one session to build. AI promotion grows it automatically over time. Monthly maintenance keeps it accurate. The compound return is significant: every rule eliminates a manual decision that would otherwise be repeated indefinitely across every matching transaction, for every client, for the life of the practice.
View the full ReconLink feature set or book a demo to walk through rule setup with your own client data.