When an invoice becomes overdue and nobody follows up
You know the situation: an invoice has been overdue for days. Reminders are sent too late or not at all, replies get lost in an overflowing inbox, and accounting spends hours manually matching payments. This is precisely where a structured open‑items workflow with powerful fuzzy matching comes in. In practice, it reduces routine work, speeds up clarifications, and ensures that payment reminders are sent precisely and in a controlled manner.
What fuzzy matching should achieve – concise and practical
Fuzzy matching correctly identifies payments even when details differ: name variants ("Müller GmbH" vs. "Mueller"), typos in references, partial payments, or missing invoice numbers. It is important not to decide everything automatically: automation should relieve routine cases, and when uncertainty exists a release workflow must apply. Debentra offers configurable rules and thresholds that let you adjust the degree of automation to your risk and verification model.
Concrete rule sets and recommended thresholds
Below you will find practice‑oriented baseline settings that have proven effective in many B2B scenarios. Start conservatively and adjust the values based on operational feedback.
1) Name discrepancies
- Normalization: convert strings to lowercase, unify umlauts, remove punctuation, remove common legal‑form tokens (GmbH, UG, AG, Ltd., etc.).
- Fuzzy similarity: use a string‑similarity score. Recommendation: ≥85% → auto‑match; 70–84% → verification queue; <70% → no match.
- Whitelist/blacklist: add frequent partners with known name variants to a whitelist to reduce false‑negative matches.
2) Amount tolerances
- Exact amount match: if the payment amount and invoice amount are ≈ 100% identical (tolerance ±0.5% or ±1.00 EUR), → auto‑match.
- Partial payment detection: if the amount matches a defined installment plan amount (tolerance ±2%), → auto‑match as partial payment and assign to the installment plan.
- Deviations up to 5% or larger rounding differences → verification queue with a note about possible cash‑discount/fee deductions.
3) Reference heuristics
- Invoice number in the remittance text: present invoice numbers are weighted highest – immediate auto‑match.
- Partial matches: if the last 6–8 digits of the reference match the invoice number, increase the score.
- IBAN/bank references: partial IBAN matching (e.g., last 4–6 digits) can add points, especially when name score is low.
- Keyword detection: order numbers, customer numbers, or fixed keywords ("Rechnung", "Invoice") in the remittance text → increase the score.
4) Confidence scoring and thresholds
Develop a points system where each rule contributes to the total score. Example scheme (recommendation):
- Name exact match: 50 points
- Reference/Invoice number: 30 points
- Amount match: 40 points
- IBAN partial match: 10 points
Overall thresholds: ≥90 points → Auto‑match; 70–89 points → approval workflow (review); <70 points → manual allocation required.
Approval workflows for uncertain allocations
Automation must not replace control. Define clear procedures for all cases in which the system does not find an unambiguous match.
Workflow building blocks
- Review queue: All cases with scores in the intermediate range are placed in a central queue, visible to the receivables and accounting teams.
- Assignment & roles: A ticket has an assignee; escalation if no decision is made within, e.g., 48 hours.
- Approval: Options "Auto‑assign", "Hold", "Do not allocate"; every action is logged (audit trail).
- Bulk approvals: For cases with similar patterns (e.g., recurring partial payments) group approvals can be allowed to increase efficiency.
- Notifications: Email/Slack notification for new review cases or missing decisions.
Rules for reminders and communication
Important: Payment reminders should only be sent automatically for clearly allocated or consciously approved cases. For review cases:
- Disable automatic sending while a match is pending.
- Employee approval required when score <90 and reminder thresholds are reached.
- Visible preview and edit option for reminder texts before sending – so control always remains with the user.
Practical implementation tips
- Start conservatively: A higher auto‑match threshold at the beginning reduces misallocations; gradually relax as experience grows.
- Monitoring: Define KPIs (e.g., share of automatic allocations, error rate in the review queue) and review them weekly.
- Create a whitelist: Frequently misidentified counterparties with correctly preconfigured aliases save a lot of time.
- Maintain installment plans: If you expect recurring partial payments, store installment plans in Debentra so partial payments are recognized automatically.
- Audit & documentation: Every allocation and approval is logged – important for traceability and later adjustments of the rules.
How Debentra supports this
Debentra offers configurable fuzzy matching, a point‑based confidence system and flexible approval workflows that adapt to your internal roles and authorization processes. In the PRO edition many manual allocations are eliminated by advanced matching rules; in PREMIUM, installment plans and integrated accounting interfaces can further simplify tracking – while the final decision on reminders and dunning always remains with the user.
Quick check: How to implement the first 7 days
- Day 1–2: Import bank files and post a few test payments; enable basic normalization.
- Day 3: Activate recommended thresholds (Auto ≥90, Review 70–89); create a whitelist for your top‑10 customers.
- Day 4–5: Monitor the review queue, adjust rules (name filters, amount tolerances).
- Day 6–7: Test the approval workflow, configure notifications, define bulk approval processes.
Conclusion
With a clear rule set for name matching, amount tolerances and reference heuristics, manual allocations can be drastically reduced – provided you define sensible confidence thresholds and a practical approval workflow. Debentra supplies the tools: configurable matching rules, review queues and full audit logs. This lets you automate routine cases, retain control over critical decisions and speed up your receivables management without blind trust in technology.
Act now: Choose the right edition for your team and reduce manual allocations. Choose an edition
Next step
Try Debentra yourself
Central overview and process control for open items and payments with full data sovereignty; reduces manual allocation work (PRO) and enables integrated accounting (PREMIUM); selectable installation and licensing models.