Heating Oil Delivery Software: Route Winter Top-Ups With AI Help

Look for five things: a tank-level estimate for every customer that learns from each delivery (or a tank monitor feeding it), a run-out flag that jumps the queue, route building that ranks stops by days to empty and respects truck load, a driver app that records gallons delivered, and published terms you can compare. Ask to see all five working.

The estimate is a guess until a delivery corrects it. Degree-day software cannot know that a customer has gone away or lit a wood stove, so the best systems show how recently each estimate was checked and let a gauge reading override it. Quantities below are in gallons because that is how the vendors I checked quote them; if you sell by the litre, the method is identical.

Follow me on Instagram@sagnikteaches

Rank stops by days to empty, not by percentage left

A tank at 10 per cent is not always more urgent than one at 20 per cent. What matters is how long the oil will last at this week's weather, which is gallons left divided by expected daily use. Expected daily use is the forecast degree days per day divided by that customer's K-factor (explained in the next section). The table below uses a cold week at 35 degree days a day, for four customers.

Connect on LinkedInSagnik Bhattacharya

Ranked by percentage, the small tank (C) comes third and the leaky large house (B) last. Ranked by days, C is first with just over three days of oil, and D, the customer who looks most urgent on the percentage screen, can wait. The figures are illustrative, but the pattern is common: small tanks and hungry houses empty first.

Subscribe on YouTube@sagnikteaches

Put run-outs above all of this. A starting set of tiers, to adapt to your own trucks and forecast, looks like this:

1. Tier 1: heating off or failing and a vulnerable person at home (an infant, an older relative, someone ill). Go now.
2. Tier 2: tank dry or heating off, nobody vulnerable. Same day, first on the next route.
3. Tier 3: under three days of oil. On today's route unless a Tier 1 or 2 stop displaces it.
4. Tier 4: three to seven days. Plan for tomorrow and group by area.
5. Tier 5: everything else, including fixed-interval customers who are due.

Within tiers 3 to 5, order by area to save driving. Tiers 1 and 2 are never reordered for distance. A tutorial on priority sites for snow ploughing uses the same idea for a different winter job and is worth a read for the tiering logic.

How a tank-level estimate is built, and where it drifts

A degree day measures how cold a day was. Droplet's scheduling guide says to take the day's average outdoor temperature and subtract it from 65 (Fahrenheit), so a day averaging 30 degrees counts as 35 degree days; five such days make 175. Systems that work in Celsius use a different base, and any change of base changes every K-factor, so check which one your software uses and keep it fixed.

The K-factor links weather to fuel. Energy Force, another delivery-software vendor, defines it in its help centre as the number of degree days needed to burn one gallon, so a higher number means a thriftier house. You will also see the inverse (gallons per degree day) in places, so read your vendor's definition before comparing figures. In practice it is worked out from the gallons that refill the tank and the degree days since the previous fill, which is why a partial fill spoils it. Droplet's page says its software learns with every delivery and adjusts the K-factor automatically, and Smart Service says automatic delivery schedules are built from tank size, usage history and delivery frequency, with degree-day estimates as an input.

Work through one customer to see how it behaves. Their tank holds 275 gallons, and the last full fill took 220 gallons after 1,760 degree days: K is 1,760 divided by 220, which is 8.0. Since that fill, 1,290 degree days have passed. Expected use is 1,290 divided by 8, or about 161 gallons, so the estimate is 114 gallons left, 41 per cent. If your trigger is 30 per cent (82.5 gallons), the tank reaches it at 1,540 degree days after a fill, so 250 degree days from now. At 35 a day that is about seven days. When you deliver 190 gallons at that point, the new K is 1,540 divided by 190, about 8.1, and the next estimate improves.

Where the estimate goes wrong

Three things bend it. A customer changes how they heat (a wood stove, a guest annex, a spare room switched off). A customer is away, so the house uses less than the weather predicts. Or the previous delivery was a top-up rather than a full fill, which makes the next K-factor meaningless. Droplet's guide says degree-day software is great but not perfect for exactly the first two reasons.

A realistic failure: the screen says 41 per cent for a customer who has not called. A driver passing for another delivery notices the gauge at 14 per cent. The customer had moved a relative into a back room and kept the heating on all day, and the K-factor on file came from a partial fill the winter before. Had the driver not looked, the route would have reached this stop about five days late, which at that burn rate means a run-out and a Tier 2 call-out. The fixes are cheap: record a gauge reading at every delivery, show the date of the last real reading beside every estimate, and put anything suspicious on a fixed interval until its K-factor settles.

New customers in their first winter

Droplet's scheduling guide says a degree-day system should only take a delivery or two to be dialled in to a home, and that before the first delivery it should accept baseline settings such as house size, energy efficiency and whether the oil heats water as well as the house. That leaves a gap: the first winter, when the estimate rests on a guess. A cautious rule closes it. Until a customer has two full fills on record, set their trigger higher than your standard, say 40 per cent instead of 30, and mark them as new on the board so nobody trusts the number blindly.

The cost of that caution is worth one sum. On a 275-gallon tank, a 30 per cent trigger sends the truck when 82.5 gallons remain, so the fill is about 192 gallons; a 40 per cent trigger sends it at 110 gallons, a fill of about 165. For a house that burns 1,000 gallons over a winter, that is roughly five deliveries against six, so the cautious rule costs about one extra stop for that customer in their first season. Set that against a run-out: an unplanned trip at the worst moment, a customer in a cold house, and a boiler that may need a service visit before it fires again if the tank was run dry. One extra stop is the cheaper mistake, and the rule switches itself off once two full fills have set a real K-factor.

The checklist: five groups to test in a demo

Each item says why it matters and how to check it with the vendor's screen in front of you. Ask for a demo with your own data loaded, or at least with ten of your customers entered by hand.

Tank-level estimates

• A K-factor learned per customer. A company-wide figure drifts. Verify: open one customer, find the K-factor, record a new delivery, and watch it change.
• Settings you can alter. Base temperature, house size or efficiency, and heat-only against heat plus hot water. Verify: change one on a test customer and see the degree days due move.
• A fixed-interval option. Droplet's guide calls it Julian-day scheduling: a set number of days between deliveries whatever the weather. Verify: set one up and see it appear in the delivery list.
• Tank monitor feeds. Droplet describes an ultrasonic gauge that reads levels every hour; Smart Service says low readings feed its route queue before the customer calls. Verify: which monitors it reads, and what the list shows if one stops reporting.
• The age of each estimate. A verified 20 per cent and a three-month-old guess look identical on a bare percentage. Verify: look for a last-checked date.

Run-outs and priority

• A run-out flag that outranks everything. Verify: raise one from a mock call while a route is already open on the dispatch screen.
• Notes that travel with the stop. Vulnerable occupants, gate codes, dogs, steep drives. Verify: check they appear on the driver's screen, not just the office record.
• Mid-day inserts. Smart Service says dispatchers can insert will-calls mid-day and rebuild the affected route. Verify: add a stop to a route that is out and see what it does to the others and to the truck load.
• Sorting by something close to days to empty. Droplet lists customers by tank level, degree days due or fillable gallons. Degree days due is the nearest to days to empty; adding the forecast on top may be a manual step. Verify: ask how you would produce the ranking in the table above.

Routes and trucks

• Load and capacity. A route that needs more oil than the truck holds fails at the tenth stop. Routific's page lists capacities, time windows and driver shifts. Verify: load a route heavier than the truck and see whether it warns you.
• Access and time windows. Gates, hose length, someone-must-be-home customers. Verify: check each can be stored and used by the routing, not only noted.
• Re-optimising in minutes. Routific says routes can be re-optimised as often as you like. Verify: time a re-plan after adding three stops.
• Hand-off to navigation. Droplet mentions Waze and Google Maps. Verify: make sure the driver does not retype addresses.

Driver app and tickets

• The day's manifest and a next-stop flow. Verify: do a full mock delivery on a phone.
• Gallons recorded and the order closed. Droplet's driver-app article says the app presets the truck's electronic meter over Bluetooth, pulls the delivered gallons back and prints the delivery slip automatically, while the driver still finalises the order. It names Mid:Com and Liquid Controls meters. Verify: check your meter is on the list, and whether a manual entry is possible when a meter fails.
• A gauge reading field at delivery. This is what keeps estimates honest. Verify: check it exists and feeds the estimate.
• Poor signal. None of the vendor pages I read describe offline use. Verify: ask directly, and test in the dead spot on your longest route.

Money, data and support

• The billing plans you sell. Smart Service lists cash on delivery, budget, pre-buy, cap and fixed-price plans, and Droplet lets you charge a stored card before delivery. Verify: match each to your offers.
• Your accounting package. Smart Service is described as a QuickBooks add-on. Verify: if you use something else, ask what the export looks like.
• The twelve-month cost. Droplet says 'Request Dealer Pricing' and publishes no figure. Smart Service's pricing page gives a starting price per user with a five-user minimum, and says onboarding and data conversion are quoted separately. Verify: get hardware, set-up, minimums and contract length in writing.
• Support on a cold Monday. Droplet's own buying guide suggests ringing and timing the reply. Verify: do it, at the busiest hour, before you sign.

Building the route: load, access and mid-day inserts

Take a 2,000-gallon truck and a morning route of twelve stops expecting these deliveries in order: 180, 150, 90, 210, 120, 160, 75, 200, 140, 170, 190 and 230 gallons. They total 1,915, so the truck leaves with 85 gallons of margin, about 4 per cent. That is a thin margin in a cold week, because every fill is bigger than it would be in October. If fills run 10 per cent heavier the same route needs 2,107 gallons and the truck is empty before it reaches stop twelve.

Now a Tier 2 call comes in at 10:40: a run-out, 220 gallons. The route is at 2,135 gallons if you add it. Something has to move, and the days-to-empty column decides what. Stop twelve (230 gallons) has six days of oil, so it moves to tomorrow. The run-out joins the route and the load is 1,905 gallons again. A system that shows remaining load and each stop's days to empty makes that a thirty-second decision; one that shows only a map makes it a guess.

Keep the order within a route simple: serve the highest tier first, then follow the road. Where a truck has to reload, treat the depot visit as a stop with its own time. For the wider rules on planning small delivery rounds, the small fleet routing tutorial is a good companion, and the cost per drop tutorial shows how to price a stop so that short top-ups pay their way.

What three vendors' own pages say

These are the vendors' claims from their own pages in October 2026, not independent tests. Check each against the list above in a demo.

Of the two oil-specific products, only Smart Service publishes a figure, and the minimum is what a small firm should notice. Five users at the $129 starting rate is $645 a month, or about $548 with the 15 per cent annual discount, before onboarding. A two-truck firm with two drivers and one person in the office still pays for five seats. Routific's paid plan is $150 a month for up to 1,000 orders, but it leaves the tank estimates to your own spreadsheet, so its real cost is $150 plus the hours someone spends keeping that sheet honest every week of winter. Price Droplet against both once it has quoted, over twelve months and including any tank-monitor hardware.

Cargas Energy also ranks for this search, but its pages returned only a title when I fetched them, so I cannot describe it. Ask for a demonstration and run it through the checklist like any other.

Where AI helps on an oil round

An assistant is useful beside the software, never instead of it. Four jobs fit.

Logging a run-out call without losing the detail

Whoever answers the phone in a cold snap is under pressure. A prompt that turns a call transcript or a typed note into a clean record helps. Keep a human on the final priority call.

Turn this call into a record: heating status (running, cooling, off), the
gauge or tank level in the caller's own words, tank size if mentioned,
vulnerable occupants (young children, older people, illness), access notes,
requested timing and anything missing. Do not assume what was not said.
Suggest a tier: 1 no heat with a vulnerable person, 2 tank dry or heat off,
3 under three days of oil, 4 routine. Give one line of reasoning.

[transcript]

Suppose a caller says the house is getting cold, the gauge is 'on the red bit', a toddler and an 84-year-old relative are staying, the drive is steep and there is a gate with a code. An illustrative record: heating status, cooling but not off; level, 'on the red bit', not confirmed empty; tank size, missing; vulnerable occupants, a toddler and an older relative; access, steep drive and gated, code not yet given; timing, today; suggested tier, 2, rising to 1 if the heating goes off, because the tank is low and vulnerable people are present. What you would fix: a first draft often writes 'tank empty'. The caller said the red zone, which is not the same, so keep the caller's words and let the dispatcher decide.

Auditing K-factors in a spreadsheet

Export deliveries (date, gallons, degree days since the previous fill) and ask an assistant to calculate K-factors and flag outliers. The ChatGPT and Excel guide shows how to do it safely with a spreadsheet, and the tutorial on cleaning messy data helps if your customer list has duplicate addresses first.

Consider one customer with four fills. The K-factors are 8.0, 8.0, 7.9 and then 5.6, which is about 30 per cent below the average of the first three. The assistant should flag the fourth and list what to check, without inventing a cause: was the third delivery a full fill, did the household change, is there a leak, is the gauge misread? What you would fix: ask it to show every calculation, recompute one by hand, and tell it to report arithmetic and flags only. Assistants will happily explain an outlier with a story that fits and has no evidence.

Writing the cold-snap message

Before: 'Cold weather is coming. Please order oil.' After, drafted by AI from your rule (days to empty under ten at the forecast): 'Hi [first name], it has been 11 weeks since your last delivery and a cold week is forecast. By our estimate you may need oil within about a week. Reply with your gauge reading or book at [link]. If your heating is already struggling, call us now.' You choose who gets it and you send it. Do not let the draft promise a delivery date you have not scheduled.

Reviewing last winter's run-outs

In spring, give the assistant the list of last winter's run-outs with a short note on each and ask it to group them by cause: late will-call orders, estimate drift, monitor offline, fixed-interval customers who were not due on paper. With a small list it will happily find patterns that are not there, so ask it to give counts for each group and to say when a group has fewer than three cases. The aim is a short list of rule changes, not a story.

Four numbers to compare after a cold month

Measure four things and compare them with the same month last year if you have the records.

• Run-outs per 100 customers. The number the whole exercise exists to lower. Note how many were will-call and how many automatic.
• Estimate error. Predicted gallons against gallons actually delivered. A delivery predicted at 192 and delivered at 190 is a 1 per cent miss. If more than one stop in five misses by over 20 per cent, revisit baselines and fill types.
• Stops per truck-hour and gallons per stop. Tighter routes raise the first. Many small top-ups lower the second, which may call for a minimum delivery.
• Time from run-out call to delivery. Tier 1 and 2 calls should be on a truck the same day. Log each one.

A system that is working shows fewer surprises, not zero. The honest aim is that every run-out has a reason you can name and a rule you can change. Write those rules down before the next cold snap, and keep them with the software contract so a new member of staff can run the board from the same page.

Questions oil dealers ask when choosing delivery software

Do I need tank monitors as well as degree-day software?

Not for every customer. Degree-day estimates suit steady households, but Droplet's own guide calls them great but not perfect because the software cannot know a customer has gone away or changed the heating. Monitors earn their cost on customers who run low repeatedly, vulnerable households and hard-to-reach tanks, where a real reading beats an estimate.

Can ChatGPT or another assistant plan my routes?

It can sort a short list and suggest an order, but it does not know your truck load, road restrictions or live traffic, and it can drop a stop without telling you. Use it to draft messages and audit spreadsheets, and use a routing tool for anything beyond about ten stops. Always check that every stop appears once and the load fits.

What is the difference between will-call and automatic delivery?

On will-call, the customer orders when they want oil. On automatic delivery, you decide when to deliver, using degree days, a tank monitor or a fixed interval. Droplet's scheduling guide calls automatic accounts your most profitable, which is also why a wrong estimate on one of them matters: you own the risk of the run-out.

Is a general route planner enough for a first winter?

It can be. Routific's pricing page lists time windows, capacities and driver shifts, with a free plan up to 100 orders a month and a flat $150 for up to 1,000. What it lacks is any idea of tank levels, so you would feed it a priority list from your own estimate sheet. A winter month of 300 stops would outgrow the free plan.

How often should I recalibrate a customer's K-factor?

After every full fill, because that is when degree days and gallons can be compared cleanly. A partial fill breaks the sum, so check whether the previous delivery filled the tank before trusting a new figure. Flag any K-factor that moves by more than about a quarter from the customer's recent average and find out why.

Further reads

• Can a Small Courier Firm Use AI to Plan Delivery Routes? — Route planning ideas from small courier firms.
• AI in Logistics for Small Delivery Firms: Routes, ETAs and Messages — Routes, ETAs and messages for small delivery firms.
• HVAC Dispatch Software Compared: Which Has AI Routing for Small Fleets? — Dispatch tools compared for heating and cooling fleets.
• AI Dispatcher for a Towing Company: Calls, Locations and Quotes — Emergency call handling and dispatch, from a neighbouring trade.
• Fleet Admin for Small Businesses: Automate Services, Checks and Fuel — Automating vehicle services and checks.
• AI Job Scheduling and Dispatch for Trades and Field Service Teams — Scheduling and dispatch for trades teams.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Droplet's product page and its articles on choosing oil delivery software and scheduling; Smart Service (My Service Depot) fuel oil delivery and pricing pages; Energy Force help centre article on K-factors and degree days; Routific pricing page. Cargas Energy's pages returned only a title and are not described. All read in October 2026.