Quick Summary (TL;DR)
Start with Home Assistant’s built-in Energy dashboard and a small number of native dashboard cards. Move to Grafana only when a specific analysis needs data or visualisation features you cannot get economically from Home Assistant itself. The native Energy dashboard already presents electricity import, export, solar generation, batteries and device consumption. You can reuse Energy cards in custom views, and combine them with History Graph, Statistics Graph, Tile and Sensor cards. For most homes, this provides the best balance of useful information, low maintenance and family acceptance.
Grafana is a strong second tool for engineering-level questions: precisely when did a heat pump begin cycling, how did inverter output and grid voltage change together, or what happened across thousands of EV-charging sessions? Grafana does not automatically see Home Assistant entities and is not a time-series database. You need a compatible data source such as InfluxDB, PostgreSQL or another supported backend; its quality, retention and query configuration determine what Grafana can show. Adding Grafana does not repair inaccurate meters, missing readings or incorrectly modelled energy flows.
The buying decision should be made in this order: data quality → required questions → storage → dashboard software → display hardware. A wall tablet, mini PCAliExpress price or 24-inch monitor is a display choice, not a prerequisite for better energy monitoring. This guide compares dashboard features, data resolution, costs, operating risks and 24/7 screens. It also includes native dashboard card YAML, a Grafana deployment checklist, worked energy and cost examples, and an objective selection framework.
What are you actually trying to see?
Energy monitoring is often confused with visualisation. A household may already know its daily import in kWh but still be unable to explain a 2.5 kW evening spike. Another household has excellent second-by-second readings but no monthly totals that reconcile with its utility bill. Adding a sophisticated panel can make both situations look impressive without solving either. Begin with the decision you intend to take from the information: confirm the electricity bill, reduce unwanted standby loads, schedule an EV, prove a heat-pump adjustment worked, or investigate an intermittent inverter fault.
Home Assistant’s built-in Energy dashboard is principally for energy accounting and accessible daily feedback. Its state and history views answer “what is happening?” and “what happened recently?”. Grafana is strongest where multiple heterogeneous streams, precise time selection, transformations, historical comparison, and reusable analytical panels matter. None of these changes how a protection device, battery management system or EV charger implements its physical control and safety logic.
| Question | Minimum reliable input | Usually the simplest useful view |
|---|---|---|
| How many kWh did we import or export today? | Cumulative import and export energy counters (kWh) | Home Assistant Energy dashboard |
| What is pulling 3 kW right now? | Current power sensors (W/kW), with clear measurement boundaries | Native tile/sensor/history cards |
| How did each room use heat last winter? | Consistent temperature and heat demand history, suitable retention | Statistics Graph first; Grafana if time alignment is complex |
| Did an EV charging session overload a phase? | Phase current, charger current limit, timestamps and relevant electrical events | Grafana or retained high-resolution trace |
| Is the battery actually saving money? | Import/export and battery energy, tariffs, efficiencies, time boundaries | Energy dashboard plus separate validated cost model |
| What caused a brief inverter fault? | Device telemetry and events at sufficient sample resolution | High-resolution export + Grafana; not hourly statistics |
The three layers: energy accounting, time-series storage and visualisation
Separate what measures a quantity from where the measurement is stored and what ultimately plots it. Home Assistant integrates devices. Its Recorder stores detailed state changes and compiles statistics; its native dashboard and cards display those results. A separate InfluxDB installation can receive a chosen subset of entities from Home Assistant and keep a longer or finer-grained record. Grafana queries a configured data source and draws panels on top of it. Grafana can also connect to supported SQL data sources, but direct reads from Home Assistant’s internal Recorder schema should not be treated as a stable, officially supported integration contract.
The important Home Assistant distinction is between detailed history (retained by Recorder for a configurable period), five-minute short-term statistics (retained for the same relevant period), and hourly long-term statistics (not purged automatically for qualifying sensors). Hourly statistics are excellent for tracking changes over a year; they cannot recreate a 20-second event that was never retained at that resolution. Likewise, shortening a Grafana refresh timer from 30 seconds to two seconds does not make a sensor produce new data more frequently.
| Layer | Examples | Owns what? | Failure impact |
|---|---|---|---|
| Physical measurement | Grid meter, PV inverter, Shelly EM, heat meter, smart plugAliExpress price | Electrical or thermal readings; calibration and sensor limitations | Incorrect or missing source data |
| Home Assistant integration | Official and supported local integrations, MQTT, Modbus gateways | Entity state and metadata; availability and timestamps | Missing/stale entities and gaps |
| Primary history | Home Assistant Recorder, long-term statistics | State history, energy statistics and core HA history | Native dashboard history becomes incomplete |
| Optional analytical store | InfluxDB or compatible time-series/SQL store | Selected long-lived, fine-grained telemetry | Grafana panels may lose data; HA can still run |
| Presentation | HA Energy, Statistics Graph, Grafana | Layout, queries, user interaction, visual interpretation | Charts disappear but underlying sensors may continue |
| Display and access | Wall tablet, browser, mini PCAliExpress price, monitor | Physical viewing and secure authentication | Screen unavailable; monitoring/control should remain independent |
This architecture yields a useful rule: do not make a dashboard the single point of control. A tablet can be offline, a Grafana server can restart, and a Wi-Fi connection can disappear. The charge point, heating appliance and other protected hardware must remain safe without an attractive front-end.
Built-in Home Assistant Energy dashboard: what you get without Grafana
The Energy dashboard is included in Home Assistant and is configured by selecting suitable grid, solar, battery, gas, water and individual-device entities. You can start with a grid import counter and expand later. In the current Home Assistant documentation, Energy cards such as the energy usage graph, solar production graph, distribution and device consumption views can also be placed on a normal custom dashboard. It is no longer accurate to describe all Energy components as permanently locked inside the dedicated Energy page.
The Energy dashboard works with historical statistics rather than blindly graphing raw sensor values. Home Assistant checks the state class and units of eligible entities. A total-increasing energy counter in kWh is not interchangeable with an instantaneous W sensor; the latter must be integrated correctly if you need energy, and real-world reconciliation still matters. See our existing guide to correct grid, solar and battery entities for the sensor model itself. This article is about choosing the display, not reimplementing the Energy setup.
| Native feature | What it does well | Limit you should account for |
|---|---|---|
| Energy usage and distribution | Daily/period energy flows, solar, grid and battery accounting | Depends on correct energy sources and long-term statistics |
| Energy date selection | Select and compare supported reporting periods | Card collections and date selection must align |
| Statistics Graph | Hours, days or months using aggregated sensor statistics | Aggregates cannot show every transient |
| History Graph | Recent exact sensor history; long-term statistics for older spans | Raw historical detail depends on Recorder retention |
| Tile and Sensor cards | Fast everyday status and interaction | Not designed for forensic cross-device time analysis |
| Sections views | Responsive multi-section family dashboards | Many dense charts can still overload small screens |
| Energy power badges/cards | Current consumption from configured compatible sources | Derived totals must be checked against measurement boundaries |
A clear family energy screen can have four panels: current power and battery state at the top, today’s import/solar/export in the centre, flexible-load readiness such as EV departure or hot-water status, and one actionable exception (“grid import is unusually high”). A wall dashboard is most useful when a user knows what action each number supports. Displaying every inverter and sensor register merely because they exist is counterproductive.
Grafana: why people install it, and what it does not solve
Grafana can combine time-series visualisations, dashboard variables, repeated panels and data links. It lets you move through time ranges quickly and compare multiple series at chosen temporal resolution. That is valuable for homeowners who also maintain server infrastructure, for installers investigating intermittent faults, and for technically curious households building an experimental optimisation project. These advantages come with work: you must select a data source, manage credentials and retention, learn its query language, and check that every plotted series means what its legend claims.
For example, a panel combining grid power, EV charging current and heat-pump compressor operation must consider whether measurements represent different electrical phases, whether positive power means import or export, and whether timestamps represent sensor observation or database arrival. Plotting data on a shared graph is easy; making it comparable is an engineering task. When different units are shown together, give them clearly labelled axes or separate panels rather than producing a colourful but misleading chart.
Do not assume that every Grafana feature is available on every managed Grafana service or every self-hosted version. Query editor interfaces, alerting support and data source plugins change. Confirm the deployed Grafana edition, source database version and access permissions before copying an older community JSON dashboard. A Grafana panel is a report, not evidence that a meter was calibrated.
| Capability | Home Assistant native | Grafana with suitable data source |
|---|---|---|
| Quick daily electricity totals | Excellent | Possible, but requires additional queries/model |
| Simple family controls | Excellent, integrated actions | Not a substitute for HA controls |
| High-resolution multi-series traces | Limited by retained history and chart type | Strong where source retains the required granularity |
| Time-range and variable-driven comparisons | Useful built-in choices | Strong, reusable custom variables |
| Long-term energy statistics | Built in for qualifying sensors | Can visualise stored series; retention configured elsewhere |
| Reusable engineering dashboards | Reasonable, manageable templates | Strong with provisioning and variables |
| Installation effort | Low once sensors are correct | Medium to high if database and auth also needed |
| Wall-tablet ease | Strong mobile-friendly interface | Depends heavily on dashboard design and kiosk/browser management |
| Offline operation | Yes with local HA and sensors | Yes when self-hosted and data sources stay local |
| Additional infrastructure | None for existing HA | Usually at least Grafana and a compatible data-source path |
Which charts should you use for which measurements?
The choice between a line chart, a step chart and a bar chart is not cosmetic. Power in watts is a rate, energy in kWh is an accumulated quantity, and a thermostat demand entity is a state. Combining them indiscriminately produces invalid conclusions. A cumulative energy counter that increases every hour will look like a steadily rising line even if the underlying load changes drastically. To compare consumption across days, calculate daily increments or use energy statistics; do not compare the raw cumulative counter’s height.
| Signal | Unit or type | Good visualisation | Misleading alternative |
|---|---|---|---|
| Grid import/export instantaneous power | W or kW, signed or separate | Line or area over time, clearly signed | Daily kWh bars without converting units |
| Daily grid energy | kWh over defined reporting day | Daily bar chart | Plotting lifetime meter total as daily energy |
| Heat-pump flow temperature | °C | Line with return/outdoor temperature separately labelled | Stacked bar with energy in kWh |
| Boiler burner state | Binary on/off | State timeline; separate runtime summary | Treating demand state as proof of flame |
| Battery state of charge | % | Line bounded 0–100% | Assuming SOC itself represents delivered kWh |
| Dynamic electricity price | Currency/kWh and interval | Step or bar by valid tariff interval | Connecting hourly prices as a continuous ramp |
| EV session energy | kWh per session | Bar by session or day | Mixing charger kWh with utility meter import kWh |
| Meter availability/data gaps | Boolean, age or quality flag | State timeline and gap highlights | Interpolating missing values as if measured |
Sampling and aggregation need explicit disclosure. If the power meterAliExpress price updates every ten seconds, a five-minute average compresses thirty nominal points. That may be suitable for a monthly power profile but hides fast cycling and some current-limit responses. A thirty-second dashboard refresh cannot recover the discarded details. Conversely, trying to plot millions of points on a kitchen tablet seldom helps a household decision. Configure downsampling for the question being answered, not to maximise points on the screen.
Recommended architecture for four real household types
Scenario A: a normal home with only an electricity meter
With a smart-meter integration or suitable local CT meter, start with the native Energy dashboard, a current-power Tile or Sensor card, and a weekly Statistics Graph. You can answer whether nighttime loads are excessive, when consumption changes and whether a tariff switch reduced peak-hour energy. The host can remain a Home Assistant Green or an existing supported Raspberry Pi setup if performance is otherwise adequate. A new NAS and a separate dashboard server are not justified by the existence of graphs alone.
Scenario B: solar PV, a home battery and an EV charger
Use the Energy dashboard for accounting and a purpose-built dashboard with live grid power, battery charge/discharge, PV power, current tariff and charger session status. Validate that each source refers to the same physical boundary: many inverters present battery DC power while household meters read grid-side AC power, and their numbers are not directly additive without accounting for conversion losses. Start with Home Assistant cards. Add Grafana when comparing full charging sessions, changing loads by phase or conducting monthly tariff experiments becomes genuinely difficult in the native UI.
Scenario C: a heat pump and multi-zone heating system
Native History Graph and Statistics Graph cards can show flow/return temperatures, compressor demand, room temperatures and daily electricity consumption. That is enough for most tuning. A specialist investigating defrost sequences, repeated startup or short cycles may need accurately timestamped, frequently sampled events retained beyond Recorder’s detailed-history window. Store the necessary subset in an analytical database before installing Grafana. Check flow, meter resolution and sampling intervals; a chart cannot turn a thermostat relay into a real compressor-runtime sensor.
Scenario D: large, technically managed smart home
A house with hundreds of sensors, three-phase PV, an energy store, a workshop and multiple EV chargers can justify a compact mini PCAliExpress price or NAS providing the analytical database and Grafana. Home Assistant remains the automation controller and primary native dashboard. Give each person an intentionally simplified household view, and reserve diagnostics and raw device traces for an engineering dashboard. Document backups, encryption, restart dependencies and data flows. A homelab integration can be sophisticated without forcing the household to rely on the homelab for basic comfort.
| Home profile | Recommended first platform | When to add Grafana | Suggested display |
|---|---|---|---|
| Apartment or small house | Energy dashboard + a few standard cards | Usually unnecessary | Existing phone or tablet |
| Solar / battery / EV household | Energy + native live status views | Longer session analysis and cross-series diagnostics | 8–11 inch tablet or family PC |
| Heat-pump tuning and fault finding | History and Statistics Graph | Retained fine-grained thermal/electrical traces | Desktop browser; wall display optional |
| Large local homelab | Home Assistant for operation, Grafana for diagnostics | At the outset only if clear data requirements exist | Mini PC server + chosen client screen |
| Professional multi-site monitoring | Site-local control plus central reporting | Strong case with access control and separated tenants | Admin workstation; authenticated reports |
Native dashboard example: energy summary without extra software
The following examples are dashboard card snippets, not a new Home Assistant configuration.yaml. Add them through the dashboard card editor or insert them under an appropriate cards: array in a custom view. Names such as sensor.grid_power are placeholders; replace them with entities on your actual system. In particular, a power sensor for W and an energy counter for kWh cannot be swapped without changing what is being shown.
Example 1: current demand and recent history
type: vertical-stack
cards:
- type: sensor
entity: sensor.grid_power
name: Grid power now
graph: line
hours_to_show: 12
detail: 2
- type: history-graph
title: Grid and solar power (last 12 hours)
hours_to_show: 12
entities:
- sensor.grid_power
- sensor.solar_ac_powerThe Sensor card gives an immediate reading and a short trend; History Graph compares the grid and solar values with timestamps. Give the chart a sufficiently short time span to stay inside detailed Recorder retention. Inspect each entity’s unit and sign convention first. If solar output is positive while export is negative, label the plot accordingly. A supplier bill cannot be validated by adding raw W samples without proper integration.
Example 2: daily energy and solar cards
type: vertical-stack
cards:
- type: energy-date-selection
- type: energy-usage-graph
- type: energy-solar-graph
- type: energy-distribution
link_dashboard: trueThese cards use the energy sources defined in Home Assistant’s Energy configuration. They inherit a shared date selection within the same dashboard collection; if multiple independent selections are desired, use deliberately matching collection_key values. The solar graph will be meaningful only if a compatible solar energy source is actually configured. Energy cards provide a good household overview without Grafana or a separate InfluxDB server.
Example 3: long-term grid demand profile
type: statistics-graph
title: Grid demand - hourly profile
entities:
- sensor.grid_power
chart_type: line
stat_types:
- mean
period: hour
days_to_show: 14This chart uses the statistics available for a qualifying power sensor, not every original state change. If it says the statistics are unavailable, check sensor metadata and the Home Assistant Statistics repair tools. A sensor supplying instantaneous power should normally have compatible measurement metadata; an arbitrary text entity cannot simply be selected and expected to gain historical statistics. The mean profile is useful for typical consumption patterns, not short inrush events.
Example 4: energy consumption by day
type: statistics-graph
title: Grid import by day
entities:
- sensor.grid_import_energy
chart_type: bar
stat_types:
- change
period: day
days_to_show: 14For an energy sensor with appropriate cumulative statistics, change expresses the increment over each interval. If the underlying meter resets, loses its state class or provides duplicate increments, correct the statistics before trusting the bars. This chart’s output is a visual check, not a guaranteed substitute for reconciling utility import from a properly identified meter boundary. More details are in our energy-data reconciliation guide.
Grafana example: choosing a sane data path
The most understandable starting architecture is meter or inverter → Home Assistant integration → selected InfluxDB export → Grafana. That keeps Home Assistant running normally. Its built-in Recorder and Energy dashboard continue to function even if the analytical stack is temporarily stopped. Do not install a second database merely because Grafana offers an InfluxDB data-source dropdown: use the one you actually maintain, and restrict exported data to the entities needed for a specific purpose.
The Home Assistant InfluxDB integration is separate from Recorder, as the official InfluxDB integration documentation explains. InfluxDB version and setup matter: InfluxDB 1.x, 2.x and 3.x do not have identical authentication, query modes or storage rules. Grafana requires a datasource configuration matching the deployed database and permissions. A random Grafana dashboard downloaded from a forum may use a different field schema, tag name or query language. Validate those assumptions before importing panels or assuming the query is broken.
- Select six to twelve starting metrics: grid import/export power, PV AC output, battery power and SOC, EV charging power, heat-pump input and a few availability/event signals. Expand only as questions arise.
- Use stable, distinct measurement names and tags: home, device, phase and source. Resist dumping hundreds of volatile attributes into tags; that can increase time-series cardinality.
- Time synchronisation: use reliable system clock synchronisation on the HA host and analytical database; document whether data timestamps are in UTC or local time.
- Export selectively: exclude unnecessary battery levels, clock sensors, chatty diagnostic entities and personal information from a long-retention archive.
- Keep source ownership clear: an inverter can report one value directly while the external meter reports another. Avoid calling both “total house power” without stating the measurement boundary.
- Reconcile: compare summed energy against a trustworthy counter before trusting Grafana calculations or inferring efficiency.
Grafana dashboard design that genuinely helps
A useful engineering Grafana screen has one contextual header, a timeline and a clearly defined exception panel. The top row can show the period, energy and tariff context; the middle can show PV output, grid import/export, battery flow and EV charge power using comparable signs and units; the bottom can show alarms, connectivity and data completeness. Avoid automatically refreshing annual charts every five seconds. For long horizons, appropriate aggregation reduces query cost and makes charts more legible.
Grafana supports dashboard variables and time-range controls; these are particularly helpful when comparing different EV charge sessions, circuits or devices. You can configure variables for equipment ID, phase or tariff period rather than duplicating one dashboard for each component. The Grafana variables documentation and dashboard usage documentation explain those capabilities. Save and export the dashboard definition and document the source data it requires.
| Design rule | Correct implementation | Common mistake |
|---|---|---|
| Time axis | Display explicit timezone and accurate interval boundaries | Assuming every day contains exactly 24 hourly intervals |
| Power sign | Describe import/export and battery charge/discharge polarity | Mixing source-specific positive/negative conventions |
| Units | Clearly label W, kW, kWh, £/kWh and % | Plotting kWh and W as if identical |
| Missing data | Show gaps; keep availability series | Connecting outages with misleading interpolation |
| Refresh interval | Match source sample interval and use case | Refreshing rapidly for hourly long-term statistics |
| Long horizon | Use suitable aggregation and series limits | Fetching a year of ten-second raw samples for a small screen |
| Financial panels | Document tariff model, fixed charges and export credit | Presenting imported market prices as final household bill |
| Access | Viewer role and authenticated screens | Public anonymous admin panels for convenience |
How much resolution does an energy dashboard need?
For daily import kWh, one hourly aggregate can be sufficient to show broad time-of-use behaviour. For a heat-pump cycling investigation, a 10–30 second sample may be more revealing. For a brief protective current response, even that may be too slow and you should inspect the device’s event log or approved test equipment. If a monitor updates every minute, no dashboard can recover the intra-minute waveform. Choose the meter and its supported update rate before choosing a plot refresh value.
Consider three practical retention tiers. Operational measurements shown at home can use current state and recent Recorder history. Planning measurements such as long-term electricity, temperature trends and bills can use hourly statistics. Diagnostics may require an appropriately sampled time-series archive or a dedicated device log. Set a retention policy that matches the value of the data and the actual size of the server’s storage. Longer history is not free: you must back it up, protect it and test restores.
| Question | Illustrative sample period | Storage horizon | Practical approach |
|---|---|---|---|
| What is our grid import today? | Meter interval as supplied; hourly aggregation useful | Years of energy statistics | Energy dashboard |
| Is an EV session respecting tariff windows? | 30–60 seconds where device supports it | Weeks or months | HA live view then Grafana if multi-session analysis |
| Does the heat pump cycle excessively? | 10–30 seconds if supported by telemetry | Weeks to diagnose | History + selected InfluxDB export |
| Did a breaker trip after a short surge? | Device event or approved high-frequency instrument | As needed | Device event log; normal dashboards may miss event |
| How did winter heating vary by weather? | Hourly means and totals | Multiple heating seasons | Statistics Graph or Grafana grouped query |
| Have solar savings improved year on year? | Daily/monthly validated energy counters | Multiple years | Energy dashboard and explicit cost model |
Sizing and cost: the screen and server are separate purchases
A cheap wall tablet can make an already good Home Assistant dashboard immediately useful. A mini PC can provide a reliable Home Assistant host or analytical server. A NAS can preserve telemetry, but it is typically more expensive and complicated than necessary if you only require a few charts. The right purchase depends on where the software runs. You do not need a powerful computer behind each wall-mounted screen; a browser is only a client and can connect over the local network to the actual server.
A dedicated display that operates every hour contributes to the electricity bill. For example, a 7 W average tablet left on continuously consumes 0.007 kW × 24 h × 365 = 61.3 kWh/year. At an illustrative £0.25/kWh this is about £15.33/year. A 30 W monitor used 24/7 consumes 262.8 kWh/year, about £65.70/year at the same illustrative rate. Real screens vary with brightness, standby, supply and device model; measure their actual draw if ownership cost matters.
| Display option | Main advantage | Principal drawback | Buying check |
|---|---|---|---|
| Existing phone / laptop | No new hardware cost | Not always visible | App and browser access |
| 8–11 inch wall tablet | Compact, familiar touch controls | Battery ageing, charging strategy, screen burn-in risk | Browser support, mounting, power, kiosk mode |
| 13–16 inch portable display | More room for plots, easy moving | Often needs external host/power | Video input, supported resolution and supply |
| 24-inch monitor + mini PC | Readable multi-chart diagnostic display | Higher power, cables and physical footprint | Host CPU/RAM, storage, mount, remote input |
| Touchscreen wall panel | Strong in dedicated installations | Installation effort and vendor lock-in | Mount depth, power, HA integration, local use |
| Existing smart TV / old display | Low initial cost if owned | May be slow, unsupported or unreliable for kiosk display | Browser updates, sleep/restart behaviour |
When you choose a tablet for fixed installation, confirm the manufacturer’s charging guidance. Permanent connection, high ambient temperatures and tight wall enclosures can accelerate battery deterioration. Use accessible, certified low-voltage power supplies; do not hide an improvised mains power extension behind a combustible panel. Commercial tablets or purpose-built panels may offer power-management features, but such features and pricing differ by product and model. Buy for supported availability and mounting, not for a pretty marketing photograph.
Worked five-year ownership comparison
The following is a planning illustration, not market pricing or a quotation. It assumes Home Assistant already operates on a reliable host, energy sensors are installed, labour is valued separately and electricity costs £0.25/kWh throughout. Hardware purchase prices are example assumptions and should be replaced with current local offers, particularly where taxes, shipping and labour differ.
| Five-year example | Native on existing device | Tablet wall view | Separate Grafana mini PC + monitor |
|---|---|---|---|
| Additional hardware purchase | £0 | £170 tablet and mount | £300 mini PC + £125 monitor |
| Extra setup / mounting allowance | £0 DIY | £70 | £90 |
| Estimated average extra electrical load | 0 W incremental | 7 W | 40 W combined |
| Energy over five years | 0 kWh | 306.6 kWh | 1,752 kWh |
| Five-year incremental electricity @ £0.25/kWh | £0 | £76.65 | £438.00 |
| Five-year illustrative total | £0 | £316.65 | £953.00 |
The 40 W example comprises an illustrative 10 W mini PC and 30 W active monitor. If the monitor sleeps for most of the day, its energy cost falls sharply; if the mini PC replaces hardware you were already using, the incremental total may also be substantially lower. Conversely, a database appliance, UPS and backups add costs that this simple example omits. The comparison is meant to stop people buying a full workstation to answer a question the existing Energy dashboard already handles.
A break-even calculation must identify a real mechanism of savings. Suppose better visibility helps avoid an otherwise wasted 50 kWh/month of electricity over ten months a year. That is 500 kWh/year; at £0.25/kWh it is £125/year. A £317 five-year wall-display investment would recover that amount in approximately 2.5 years if the avoided load is genuine, repeatable and attributable to the display. But the same benefit might be achieved by an automated alert on an existing phone, at little incremental cost. A dashboard does not produce savings merely by being installed.
Wall tablet reliability and security: the part most setups neglect
An energy screen in a kitchen, hall or home office is often used for several years. Choose a manageable browser, local access, sensible screen-sleep policy and restricted user permissions. Avoid exposing Grafana or Home Assistant administrative accounts on an unattended tablet. Home Assistant’s view visibility choices control presentation and are not a substitute for real authorisation boundaries. Use a dedicated low-privilege user where the platform allows and separate household controls from sensitive infrastructure dashboards.
For a local-only installation, do not assume that Wi-Fi coverage is adequate because the phone works nearby. Test the mounted position, reboot the wireless access point, check that the screen reconnects, and verify that the dashboard does not show stale “live” numbers as current. If remote access is needed, use supported HTTPS and authentication rather than publishing a Grafana service directly onto the internet. Review browser frame/security restrictions before promising an embedded Grafana panel inside Home Assistant; the Home Assistant Webpage card documentation warns that not every site can be embedded, and HTTPS pages cannot embed insecure HTTP content.
| Failure | What the screen may show | Safe design response |
|---|---|---|
| Internet unavailable | Cloud-fed price or weather tiles stale; local data may still work | Show feed age and retain local functionality |
| Home Assistant restarts | Blank or disconnected wall view | Independent charger/heating safety, reconnect logic |
| InfluxDB offline | Grafana panels missing history | Native Energy remains on Recorder |
| Grafana service offline | Diagnostic screen unreachable | Home Assistant still controls devices |
| Tablet asleep or battery failure | No dashboard at wall | No critical action depends on wall screen |
| Wi-Fi reconnect fails | Stale last rendered measurements | Display last-update time, log events and recover automatically |
| Clock mismatch | Spurious gaps and shifted tariffs | Synchronise clocks; use source and local display time carefully |
| Exposed view credentials | Private energy/activity data leakage | Restricted viewers; authentication and secure remote access |
How to audit accuracy before trusting either dashboard
Start from the physical grid meter, not from the chart. Confirm whether the meter exposes instantaneous power, separate imported/exported energy, net flow or per-phase values. Check which entities count battery charging from the grid and which count PV used directly. Read the device documentation for polarity and conversion. Record a controlled before/after measurement, such as a 1–2 kW appliance operating for a known interval, and compare the observed change. Do not perform experiments involving exposed live mains conductors or unsafe isolation of heating equipment.
Next, choose a period with a reliable utility bill or externally read meter and compare imported energy. Investigate any large differences in source period, time zone, missed intervals, duplicates, sensor resets and unit conversion. A display being numerically precise to four decimals is not evidence of physical accuracy. A ±3% error on a 20 kWh day is 0.6 kWh, potentially worth more than all the additional chart styling effort. If the question is whether export payments are correct, make sure the comparison uses the tariff’s contractual terms, not just the intraday energy graph.
| Validation step | Evidence | Accept / investigate |
|---|---|---|
| Confirm energy meter boundary | Import and export counter definitions | No duplicate sources or swapped signs |
| Confirm power units | W, kW, kWh metadata | Power and energy not conflated |
| Confirm updates and gap handling | Timestamp, last seen, unavailable history | Chart is explicit about missing data |
| Confirm time zone and DST | Local and UTC alignment | No duplicate or omitted tariff buckets |
| Confirm battery boundary | AC versus DC and inverter location | Conversion loss not counted twice |
| Confirm chart aggregation | Raw, 5-minute, hourly, daily | Resolution suitable for the question |
| Reconcile day/week totals | Trusted source and dashboard totals | Discrepancies explainable and within justified tolerance |
| Inspect retention and backups | Actual restore or export test | Critical periods recoverable |
Common dashboard mistakes that waste money and time
- Installing Grafana before you have reliable sensors. A perfect panel cannot fix a missing CT or an incorrectly configured import/export entity.
- Treating “near real time” as a fixed update promise. Sample period, integration polling, database write interval and browser refresh all contribute to latency.
- Displaying a cumulative total as a daily load. Distinguish kWh counters from daily interval increments.
- Publishing raw home data on an open web server. Energy patterns can reveal occupancy and routines. Protect login credentials and endpoints.
- Assuming a touchscreen is a saving device. Its cost should be evaluated against improvements actually achieved and alternative alerts or automations.
- Combining incompatible power boundaries. Grid AC power, PV DC power and battery terminal power can be valid individually yet impossible to sum without conversion.
- Copying a public Grafana JSON dashboard unchanged. Entity naming, field keys, measurement schema and query syntax may all differ.
- Making device control depend on a browser. Automation and safety logic belong in supported controllers and Home Assistant logic, not an unattended kiosk.
- Ignoring long-term maintenance. Tablet batteries, database disk space, backups, users and certificate renewal matter over five years.
- Overloading the screen. Four reliable actionable measures often outperform forty tiny panels no one reads.
Seven-day deployment and commissioning plan
| Day | Task | Pass criterion |
|---|---|---|
| 1 | Verify the installed meter entities, units, provenance and energy-source boundaries | Daily kWh broadly reconciles and no obvious duplicates |
| 2 | Create a native dashboard using Energy cards plus current power | A household member can read the key values in seconds |
| 3 | Check Recorder state class, retention and long-term statistics | Statistics Graph works over intended periods |
| 4 | Record one question native cards cannot answer; design selective export only if justified | Concrete decision/use case for analytical stack |
| 5 | If using Grafana, test a sample datasource and compare same values/time stamps in HA | No unexplained sign or timing mismatch |
| 6 | Mount a display only after network, sleep and secure-access tests | Reboots and reconnections restore a correct view |
| 7 | Repeat bill/energy comparison, document backups and train users | Known limitations and rollback recorded |
Keep the result maintainable. Save dashboard configuration exports where available, note the firmware and integrations of source meters, and record whether displayed statistics are raw, hourly or calculated. When you replace a meter, revisit source sign conventions and the statistical history. Use version-controlled copies for Grafana dashboards if the setup is substantial. Most importantly, make the dashboard readable and useful without requiring a technical explanation each time someone walks past the screen.
Buying recommendations by budget and use case
Best value for most households: existing Home Assistant host and built-in Energy cards, viewed from a phone, tablet or PC already owned. Spend additional money on an accurate energy meter or well-supported local integration before spending it on a dedicated dashboard screen.
Best dedicated family display: a reliable 8–11 inch tablet, mounted with accessible power and sensible power management, showing a carefully designed native Sections view. Choose a model with supported software updates and enough screen brightness for its location, rather than chasing maximum RAM or benchmark speed.
Best engineering workstation: a computer you already use, running a browser to a self-hosted Grafana installation fed by selected, well-timestamped measurements. This is particularly practical for fault-finding and comparing many hours or months of history. No second wall screen is necessary.
Best larger data deployment: a modest mini PC or existing NAS that already serves the household’s approved self-hosted applications, with dedicated storage, backups and access controls. Avoid choosing a NAS solely to host a dashboard if Home Assistant’s own storage is meeting the need. Read our Recorder vs InfluxDB vs Grafana storage guide (article #072, available after publication) before buying a new server; check the publication status of linked guides if importing out of sequence.
A practical workflow for selecting dashboard cards
When creating a new Home Assistant view, begin with a list of questions rather than a list of installed devices. Group those questions by audience: someone managing the family budget, someone adjusting the heating, and someone troubleshooting networked meters. They should not share one overcrowded screen. For example, a family energy view can contain a current-import Tile, a solar and battery Energy card, today’s energy graph, and an EV charge status. An engineering view can include heat-pump flow/return history, per-phase current, connection health and a link to the diagnostic Grafana service when available.
Use the built-in visual editor for the first version; move to card YAML only where it demonstrably reduces repetition or documents a precise configuration. Home Assistant Sections view lets you arrange cards in groups and choose a layout that adapts to phone and tablet widths. Test the dashboard in portrait and landscape modes. Resize cards deliberately, avoid long unbroken tables on small displays, and pay attention to units: the person glancing at a kitchen tablet should not need to decode eight registers and an inverter status bitfield.
Apply a strict naming convention to presented measurements. A label such as “Solar power” is ambiguous if one entity means total DC panel output and another means inverter AC export. Prefer labels such as “PV inverter AC output”, “Grid import now”, “Battery charging power” and “EV session energy”. This matters if you later copy the series into Grafana, where data labels are often built from database tags and the same ambiguity can be multiplied across several installations.
| Display question | Best initial element | When to improve it |
|---|---|---|
| Is the home importing electricity right now? | Tile or Sensor card in W or kW | Add history when unexpected spikes recur |
| How much electricity came from the grid this week? | Energy usage graph | Add tariff model for financial analysis |
| What is the solar contribution at present? | Current PV power and Energy distribution | Add separate battery power to explain imports |
| What did the heat pump do overnight? | History Graph of temperatures and demand | High-resolution diagnostic archive for cycling investigation |
| Which EV charge sessions were most expensive? | Native session totals and tariff sensors | Grafana/queries for many historical sessions |
| Is a meter still reporting? | Last-update or availability indicator | Alert/availability history if gaps recur |
Accessibility also matters. Use readable text, sufficient contrast, understandable names and clear missing-data labels. Colour should reinforce an already explicit meaning rather than be the only indication of import versus export. Test the display while the room is bright and when the tablet dims at night. Persistent animated gauges and flashing graphs can make a control screen harder to use and draw attention away from genuine exceptions.
Worked example: time-series sampling and storage before buying hardware
Suppose you monitor 20 electrical or thermal metrics at a ten-second sampling interval, and each source actually provides one update per interval. Thirty days contain 30 × 24 × 60 × 60 ÷ 10 = 259,200 intervals per series. Across 20 metrics, that is 5,184,000 samples. One year at that rate is about 63.1 million samples. These counts do not translate directly to a fixed disk size because database storage, compression, metadata, indexes and source cardinality vary. They do show why a decision to retain everything indefinitely should be deliberate.
A five-minute historical summary, by contrast, has 30 × 24 × 12 = 8,640 periods per series over thirty days, or 172,800 periods for the same 20 metrics. That is 30 times fewer time buckets than the ten-second example. It is a useful trade-off for energy budgeting, but may obscure a heat pump that cycles every four minutes. A sensible policy might preserve fine-grained data for the troubleshooting window and store longer-term aggregates after that. Such a policy is configured in the database or export process, not by setting the Grafana panel refresh rate.
| Sampling illustration | One metric, 30 days | 20 metrics, 30 days | Suitable question |
|---|---|---|---|
| 10 seconds | 259,200 points | 5,184,000 points | Detailed equipment cycles when sensor supports it |
| 30 seconds | 86,400 points | 1,728,000 points | EV and heat-pump session shapes |
| 60 seconds | 43,200 points | 864,000 points | Short-term household demand trends |
| 5 minutes | 8,640 points | 172,800 points | General demand profile and planning |
| 1 hour | 720 points | 14,400 points | Monthly and yearly comparison |
Another important distinction is missing samples. A sensor that reports only when its value changes does not generate one sample in every nominal ten-second bucket. Filling missing buckets with zero implies the appliance was off; forward-filling implies its previous value persisted. Either choice can be wrong, depending on device semantics. Inspect the original timestamped data and the integration’s behaviour before calculating energy from power or drawing conclusions from an apparently flat line.
A dashboard query should return a manageable number of points for the displayed interval. A 1,000-pixel chart cannot visibly distinguish one hundred thousand plotted samples in any straightforward way. For a one-month view, aggregate to hourly or daily values; for a one-hour diagnostic, retain sub-minute values if they exist. Grafana time-series panels and query macros can make adaptive grouping convenient, but verify the aggregation function: averages suit power and temperature while energy often needs interval differences or correctly calculated totals.
Troubleshooting checklist: when the graphs disagree
There are three common causes of apparently contradictory graphs: the measurement boundary is different, the stored record is different, or the aggregation is different. Investigate them in that order. If the grid meter is on the AC incoming supply while an inverter reports DC PV production, they can never be expected to balance perfectly. If Home Assistant shows 10 days of detailed history while Grafana has 90 days of high-resolution exports, their historical appearance will differ. And if one panel plots mean power while another plots a sum of raw power readings, the latter is not a valid kWh total.
- Confirm the source IDs: copy the entity ID and the source device into a validation sheet. Similar friendly names are not sufficient.
- Compare the same UTC interval: daylight-saving changes make local midnights and price intervals particularly easy to misalign.
- Inspect raw and aggregated traces: a query using a five-minute mean is not equivalent to a ten-second instantaneous trace.
- Look for restarts, resets and gaps: Recorder purging or exporter downtime can cause a divergence even when each currently displayed state agrees.
- Check conversions: 1 kW maintained for one hour equals 1 kWh; summing kW readings without multiplying by elapsed time is incorrect.
- Validate display freshness: add a recent-update or availability value so a stale browser panel cannot masquerade as live monitoring.
- Test independently: use a physical meter counter or appliance record to arbitrate unresolved differences.
Do not reconfigure all data collection at once to make the charts “match”. That risks corrupting a useful historical record. Preserve raw evidence, reproduce one time interval, identify the transformation that introduces the discrepancy, and document the fix. A repeatable reconciliation method is worth more than a dramatic-looking home screen.
Migration, backup and a sensible stopping point
It is entirely reasonable to start with native cards and never deploy Grafana. If you later add the analytical stack, keep the native dashboard in place. Export only the approved measurements, configure an explicit backup policy for the database, and test that the Grafana service can be rebuilt with its data sources and dashboards. Keep secrets out of public dashboard JSON and screenshots. Where the visualisation system is separate from Home Assistant, document restart order, machine names, network ports, retention periods and administrator access.
Re-evaluate the new setup after a month: which panel was actually used to change a decision? Were there fewer unknown power spikes, better tariff-aware charging choices, or faster fault resolution? Did anyone in the household use the wall display, or did they revert to a phone? If the extra dashboards add little value, simplifying the architecture is an upgrade, not a failure. The objective is understandable, reliable information with proportionate maintenance and cost.
Frequently asked questions
Do I need Grafana for the Home Assistant Energy dashboard?
No. The Home Assistant Energy dashboard is built in and uses appropriately configured energy data and Recorder statistics. Grafana is an optional analytical and visualisation layer, not a required add-on.
Is Grafana better than Home Assistant for live electricity graphs?
It can be more flexible for multi-series analysis and query-driven comparison, but not necessarily faster or more accurate. Actual freshness depends on sensor updates, integration latency, database writes, queries and display refresh. For a basic live household screen, native cards are usually simpler.
Can I add Energy cards to my own Home Assistant dashboard?
Yes. Official Home Assistant documentation supports using Energy cards such as the energy usage graph, solar graph, date selector and distribution card on custom dashboards. The configuration still relies on suitable energy sources and data.
Does Grafana replace Home Assistant Recorder?
No. Grafana displays information from configured data sources. Even when InfluxDB is used alongside Home Assistant, Recorder continues to provide core historical data and statistics for native features. Design, backup and maintain each component for its role.
Can Grafana show 10-second heat-pump measurements for six months?
Only if a source database has stored accurate readings at that resolution for that period. Home Assistant hourly long-term statistics cannot reconstruct missing ten-second detail. Account for storage, sampling limits, and the heat pump’s native update interval.
Will installing Grafana reduce my electricity bill?
Not by itself. Any savings come from decisions or automations enabled by better information. Include energy and hardware operating costs, plus the fact that a free native alert or dashboard may achieve the same improvement.
What size tablet is best for a Home Assistant wall dashboard?
An 8–11 inch screen is a practical starting size for simple controls and energy summaries. Larger displays can help with charts but increase cost, installation space and power usage. Test layout at actual viewing distance before purchase.
Should I leave a tablet screen on all day?
That depends on the need, panel type and manufacturer recommendations. A motion-activated or sleep-aware display can reduce consumption and wear. Ensure wall-mounted power and battery management are suitable for the device.
Can I embed Grafana inside Home Assistant?
Sometimes, using a Webpage card or a supported embedding approach. Browser security restrictions, authentication and HTTPS rules can prevent embedding. Do not weaken Grafana authentication or web security merely to make a kiosk work.
Can I use one Grafana dashboard for several homes?
Yes, with suitable data separation, role-based permissions, datasource design and clear site selection. This is more of an IT architecture project than a simple Home Assistant chart; avoid exposing private household telemetry between users.
Why are my old Home Assistant graphs less detailed?
Recorder detailed state history is retained for a limited period by default, while eligible long-term statistics retain aggregates. Older graphs may be reconstructed from hourly statistics, which cannot show every change or short event.
Why does my Energy dashboard show impossible negative consumption?
Likely causes include swapped import/export orientation, incorrect signed power assumptions, duplicated meters, resets, or incompatible energy entities. Fix the measurement model and statistics first; changing display software will not correct it.
Does Grafana work without internet access?
A self-hosted Grafana service with a local database and compatible local sensors can operate without internet access. Cloud-based sources, remote login mechanisms or vendor applications may not. Test offline behaviour rather than assuming it.
Can I query the Home Assistant SQLite database directly from Grafana?
Third-party approaches exist, but relying on Recorder’s internal database schema is not a robust general recommendation. Schema changes, locking, credentials and supportability can make it brittle. Use supported interfaces or a separate analytics export when appropriate.
Is InfluxDB necessary for a single EV charger and solar inverter?
Usually not. Home Assistant Energy and suitable history cards can already answer many everyday questions. Use an additional analytical store only when required high-resolution retention, queries or reports justify the maintenance.
What should I do if Grafana shows a different kWh total from Home Assistant?
Confirm that both calculate over the same meter boundary, unit, tariff period and time zone; then check sampling gaps, resets, aggregation and interpolation. Never simply average or scale one series until the numbers match.
How can I prevent a dashboard from exposing occupancy routines?
Keep access authenticated, use encrypted remote connections, restrict accounts, avoid anonymous public sharing and limit retained or exported personal data. Live household energy use can indicate when people are present.
Should I buy a new NAS or mini PC only for energy dashboards?
Usually no. Start with the resources your current Home Assistant setup already provides. If long-term high-resolution telemetry and backups create a genuine workload, size a separate host based on measured storage, query load and reliability requirements.
Final recommendation
Choose Home Assistant Energy plus native cards as the primary household interface. It is integrated, easy to maintain, supports long-term energy statistics and now offers flexible Energy cards on custom views. Add a wall tablet if visibility materially helps; avoid buying a data server to drive what is essentially a convenient family status page. Use Grafana as a second, specialist analytical view when you can name the investigation it enables, have a suitable data source, and are prepared to maintain retention, accounts and backups. The most valuable dashboard is one that leads to a correct action, not the one with the most panels.
Further reading and official documentation
- Home Assistant: Energy dashboard and setting up energy sources
- Home Assistant: Energy cards and date selections
- Home Assistant: Long-term statistics and Recorder behaviour
- Home Assistant: Statistics Graph card
- Home Assistant: Sections views
- Home Assistant: InfluxDB integration
- Grafana: dashboard visualisations
- Grafana: dashboard variables
- Grafana: configuration and dashboard provisioning
- For accuracy and current product support, consult the installer and vendor documentation for the specific energy meter, inverter, tablet and server hardware before purchase.