greek-electricity-bill-analysis

name	greek-electricity-bill-analysis
description	Use when comparing Greek electricity providers/tariffs, parsing ΔΕΗ/ΗΡΩΝ/Protergia/Elpedison/etc bills (λογαριασμός ρεύματος), deciding whether to switch supplier, or backtesting what a bill would have cost on another offer. Covers προμήθεια vs ρυθμιζόμενες χρεώσεις, the Διακύμανση/ρήτρα adjustment clause, ΕΦΚ, ΦΠΑ 6%, ΥΚΩ/ΕΤΜΕΑΡ, Γ1Ν night tariff, έναντι/εκκαθαριστικός bills, and RAAEY Μπλε/Πράσινο/Κίτρινο colors.
license	MIT

Greek Electricity Bill Analysis

Overview

A Greek electricity bill splits into one provider-controlled block and three pass-through blocks. Comparing suppliers correctly means isolating the provider block, computing the effective energy rate (not the headline), and holding everything else fixed.

The #1 mistake (verified): trusting the headline energy rate. Floating products (ΗΡΩΝ PROTECT, most "Κίτρινο/Yellow" tariffs) print a low base rate like 0,0825 €/kWh and add a separate "Διακύμανση Κόστους Αγοράς" / ρήτρα αναπροσαρμογής line that often doubles it. A bill that looks like 0,0825 is really ~0,157 effective. Ignore the clause and your switch recommendation flips to the wrong answer.

The four blocks (only block A changes when you switch)

Block	Lines	Provider-dependent?	In ΦΠΑ base?
A. Προμήθεια (competitive supply)	πάγιο + energy €/kWh + Διακύμανση/ρήτρα + discounts	YES — the whole comparison	yes
B. Ρυθμιζόμενες (regulated)	ΑΔΜΗΕ (Σύστημα Μεταφοράς/transmission), ΔΕΔΔΗΕ (Δίκτυο Διανομής/distribution €/kVA+€/kWh), ΕΤΜΕΑΡ, ΥΚΩ	NO — identical for all suppliers	yes
C. Taxes	ΕΦΚ (excise), Ειδικό Τέλος 5‰	NO	ΕΦΚ yes; 5‰ no
D. Pass-through	Δημοτικά Τέλη/Φόρος (ΔΤ/ΔΦ), ΤΑΠ, ΕΡΤ, late interest, roundings	NO — property/municipality based	no

See greek-tariff-reference.md for current unit rates, ΕΦΚ values, RAAEY colors, night-tariff hours, and provider consolidation (WATT+VOLT→Protergia, Elpedison→Enerwave).

Procedure

1. Extract with pdftotext -layout bill.pdf — ΔΕΗ and ΗΡΩΝ bills are text PDFs, no OCR needed. Pull: provider, παροχή (supply no.), product, period+days, kWh (day/night split if dual), each block's lines, ΦΠΑ base, totals, and Προηγούμενο Ανεξόφλητο (carried debt).
2. Reconcile to the cent: A + B + Γ(supplementary) + E(municipal) + ΣΤ(ERT) + ΦΠΑ = current-period total. If it doesn't balance, you missed a line (often a discount like Έκπτωση ΕΝ.Α or a credit).
3. Compute the EFFECTIVE energy rate = base €/kWh + Διακύμανση/ρήτρα €/kWh − discounts. Never compare on the headline base alone.
4. Verify ΦΠΑ: it is 6% (reduced) of (A + B + ΕΦΚ). Municipal, ΕΡΤ, 5‰, and interest are OUTSIDE the base. ΕΦΚ ≈ 0,0022 €/kWh residential, 0,005 €/kWh business.
5. Backtest an offer — replace only block A, recompute ΦΠΑ, keep B/C/D as billed:

   fixed_part = actual_total − actual_A − actual_VAT          # B + C + D, unchanged
   new_A      = offer_paygio × days/30 + energy_at_offer_rate # day/night aware
   new_VAT    = 0.06 × (new_A + (actual_VAT_base − actual_A))  # B+ΕΦΚ stay
   new_total  = fixed_part + new_A + new_VAT
   

   The reusable engine is backtest_engine.py.
6. Decide STRATEGY before price — then compare within the tier (not across it). Floating, 1-year fixed and 2-year fixed are different products on a risk axis, not interchangeable price points; ranking them in one flat list is apples-to-oranges. First ask: do you want to commit to a price (fixed/Μπλε) or follow the market (floating/Κίτρινο-Πράσινο)? Then rank within each commitment tier and present the cheapest of each with its term, so the choice is risk-stance first, price second. Treat contract duration/lock-in as a first-class factor: a multi-year fixed rate is only "cheaper" if you'd actually hold it, and a long lock in a volatile market (fuel/geopolitical shocks) carries opportunity risk — not free certainty.
7. Compare across the WHOLE market — don't name a winner from memory. Pull the current active-supplier list from the official ΡΑΑΕΥ tool energycost.gr (the field consolidates — see reference); the cheapest option is frequently a non-incumbent or a recent entrant, not ΔΕΗ/ΗΡΩΝ. For each candidate: (a) filter by segment (residential vs business/κοινόχρηστα) and meter type (single vs dual-zone) — an offer your meter can't take is not a candidate; (b) compute its effective rate; (c) group by commitment tier and rank_within_tiers(...) (see engine) — not one flat rank_offers across tiers; (d) tag each with term length, an eligibility flag (accepts your κοινόχρηστα/kVA/dual meter?) and a data-confidence flag (live energycost.gr value vs aggregator/stale vs formula-indexed). Never present a rate you cannot source as if it were available.

Gotchas (each one flips a number)

• Έναντι vs εκκαθαριστικός: a clearing bill (εκκαθαριστικός) nets out a prior on-account charge via a negative Αξία Ρεύματος Έναντι line. The printed payable is then LESS than the period's real cost — reconstruct the gross (add the έναντι back, plus its VAT) for a true comparison.
• Carried debt: Προηγούμενο Ανεξόφλητο Ποσό and final-payable include old arrears. Compare on the current-period charge, not final payable.
• Night/dual tariff (Γ1Ν, διζωνικό): has separate day (κανονική) and night (μειωμένη) registers. Don't apply a single flat offer rate to its total kWh — price each register, or note the ~25–30% night share. Few suppliers offer dual-zone (ΔΕΗ EnterTwo fixed, Zenith Go Electric Plus floating); ΗΡΩΝ does not.
• Consumption level sets the lever: at low kWh/month the πάγιο (standing charge) dominates → a zero-πάγιο plan usually wins even at a higher rate. At high kWh the energy rate dominates. Compute, don't assume.
• Business/κοινόχρηστα supplies (common areas, often ≥25 kVA) use business tariffs and higher ΕΦΚ (0,005) — residential offers don't apply. Confirm a candidate explicitly accepts a shared κοινόχρηστα meter before recommending it; the cheapest business product may not, and that caveat can decide the answer.
• Don't compare floating vs fixed on price alone — they're different risk products. A floating tariff can be cheaper today than a fixed one, but they answer different questions (track the market vs lock a price). Decide the strategy first (see step 6), rank within the tier, and weigh the fixed term's length. "Cheapest right now" across tiers is a misleading headline.
• A short sample misses the seasons. Greek consumption swings hard (summer A/C, winter heating). 6–8 bills can't reveal that, and naively annualising a short sample (×365/days) over- or under-weights whatever season you happened to capture. Use ≥12 consecutive months where possible, or weight by season. For a forward "what will a year cost" figure, prefer the seasonally-weighted comparator annual_cost(offer, profile) (a reference profile of 4-month bands × day/night + kVA, all-in incl. regulated + VAT + municipal) over scaling a short sample.
• Tiered & formula-indexed tariffs aren't a single rate. Many products are tiered (e.g. 0–700 / >700 kWh, day >200 kWh) or indexed (a×DAM+b, hourly MCP). Reduce them to an effective €/kWh at the supply's actual monthly kWh and the current wholesale index before feeding the engine — see tiered_effective_rate in backtest_engine.py.
• Discounts are conditional (έκπτωση συνέπειας/on-time, πάγια εντολή, e-bill). Headline rates assume you keep them; a missed payment reverts to a much higher nominal rate.
• Sign-up gifts (δώρα) are real but volatile and conditional. Vouchers / bill credits (e.g. a €50 Πλαίσιο voucher, monthly referral credits) materially cut the first-year cost — count them in the first-year total only, never as a recurring rate, and record the condition (e-contract, "not claimed in the last 6 months"). Track per offer as gift_eur (+ gift_conditions); exclude from steady-state.
• How a floating price is set varies (pricing_basis). Same color can differ: green-monthly (announced on the 1st, known for that month) vs index-prev-month (ΤΕΑ of the previous month) vs dynamic-hourly. The more it's announced ahead, the more predictable — note it; don't treat all "floating" as equally knowable.
• Per-offer behaviour caveats matter. Some suppliers claw back the consistency discount on the final-bill consumption, change discounts improperly, or attach an exit fee per remaining month. Capture a caveat and exit_fee_per_month rather than a vague "terms apply".
• Floating rates reset monthly. Verify any Κίτρινο/Πράσινο rate on the official ΡΑΑΕΥ tool energycost.gr before relying on it. Fixed (Μπλε) rates lock for the term.

Deliverables that work

For a multi-supply comparison build an Excel workbook (openpyxl) with live formulas so rates can be tweaked: Bills_Raw, Supplies, Bill_Line_Items, Fee_Model, Market_Offers, Scenario_Model, Backtest, Tier_Ranking, Recommendation. Tier_Ranking groups offers by commitment tier (floating / fixed-by-term) and ranks within each — never one flat list across tiers. Validate the embedded formulas by a headless soffice --headless --convert-to xlsx recalc and reading values back with data_only=True.

Red flags — stop and recompute

• Comparing on a headline 0,08xx €/kWh without finding the Διακύμανση/ρήτρα line
• Naming a "cheapest" provider from memory of the big names instead of ranking the live energycost.gr list (the winner is often a non-incumbent)
• Ranking floating vs 1-year-fixed vs 2-year-fixed in one flat list (apples-to-oranges) — decide commit-vs-market first, then rank within tier
• Ignoring contract length / lock-in when recommending a fixed product (a 2-year lock in a volatile market is a cost, not free certainty)
• Annualising a 6–8-bill sample by scaling as if it were seasonally representative
• Recommending a product without confirming the meter/segment fits (κοινόχρηστα, kVA, dual-zone)
• Treating a tiered or DAM/MCP-indexed tariff as one flat rate
• Applying VAT to municipal fees or ΕΡΤ (it doesn't apply there)
• Using final-payable (includes arrears) instead of current-period cost
• Recommending a low-rate/high-πάγιο plan for a near-dormant supply
• Quoting a floating rate as if it were stable, or not checking energycost.gr