Guides
Food truck KPIs owners should track monthly
Five food truck KPIs with a formula, a source and a stated blind spot each, reviewed monthly per service rather than aggregated into a meaningless total.
What to take away
- Track five things, not twenty. A measure that cannot change a decision is decoration, and decoration crowds out the ones that can.
- Each measure needs a written formula, a named source and a stated limitation before it is worth collecting.
- Review monthly, reconcile daily. The daily habit is what makes the monthly number trustworthy.
- Every measure below has a blind spot. Knowing what a number fails to show is as useful as the number.
- Compare each service against itself over time rather than against another operation, whose block, pitch and crew are different from yours.
The five
One: orders completed per service
Formula: count of completed orders in a single service.
Source: the sales record, filtered to that pitch and that time band rather than to the day.
What it is for: it is the denominator for nearly everything else, and it is the measure that reflects the constraint a truck actually has, which is how many people you can serve in a limited window.
What it fails to show: how many people approached and left. A service can look stable while quietly losing a growing share of the queue, and nothing in the sales data will say so.
Two: contribution per order
Formula: average price per order minus average component and packaging cost per order.
Source: sales data for the price, delivered supplier costs for the components.
What it is for: it tells you whether volume is worth having. Revenue rises with a busy day whatever the margin; this does not.
What it fails to show: the fixed cost of the block. A healthy contribution per order at a service with a long round trip can still lose money, which is why break-even is calculated separately per service.
Three: run-out time by item
Formula: the clock time at which each item finished, recorded when it happens.
Source: the service log, written by whoever noticed.
What it is for: a time distinguishes a par problem from a demand problem. An item that finished forty minutes into a two hour service was under-loaded. One that finished at the end was close to right.
What it fails to show: the sales you lost afterward. People who wanted the item and left do not appear anywhere, which is why the time is a proxy rather than a measurement.
Four: stock variance
Formula: opening load plus any restock, minus closing count, compared against what the sales say you sold.
Source: the load sheet and the sales record, reconciled the same day.
What it is for: it is the single best detector of miscounts, portion drift and unrecorded waste, and it is the only one of the five that catches problems nobody reported.
What it fails to show: which of those three it was. A consistent variance in one direction is a process problem to investigate, not a number to correct.
Five: labor share of contribution
Formula: total labor cost for the whole block, divided by contribution per order times orders completed.
Source: hours actually worked across the block, and the two measures above.
What it is for: it answers whether a service pays its crew out of what it earns, and it moves for reasons that have nothing to do with wages.
What it fails to show: whether the crew is stretched. A service can look efficient because two people did the work of three, and that shows up later as a resignation rather than as a number.
Reading them together
| Pattern across a month | Likely cause | Where to look |
|---|---|---|
| Orders flat, contribution falling | Component costs or portion drift | Stock variance, supplier invoices |
| Orders falling, contribution steady | Demand or findability | The pitch, and the published schedule |
| Run-outs early and often | Par set from memory | Load calculation and attach rates |
| Variance growing | Process, not arithmetic | Load out and return counts |
| Labor share rising, wage unchanged | Throughput falling | Board width, station layout, a slow item |
The fourth row is the one to act on fastest. A growing variance means the other four measures are becoming unreliable, so it undermines the whole set.
What monthly review actually looks like
Twenty minutes, same day each month, same five numbers, per service type rather than in aggregate.
Aggregate is where the useful signal dies. A Tuesday evening that loses money and a Saturday market that carries it will average to something respectable, and the monthly total will never tell you that closing the Tuesday would improve the business.
Write one sentence per measure: what it did, and whether anything should change. If nothing should change, write that too. A review that always produces an action is a review that is inventing actions.
The records underneath
None of the five exist without a daily habit: reconcile stock against sales, log waste and comps as they happen, and record run-outs by time. That discipline is also an obligation. The IRS is explicit that a business keeps a system that clearly shows income and expenses and retains the documents supporting purchases and other transactions.
Where the data lives is a decision with its own constraints on a truck, starting with whether it works at a pitch with no signal, which is covered in the operations stack and what to test before choosing it.
Keep your own export rather than relying on a provider's retention, and give each person their own login. NIST's quick-start guidance for small organizations and CISA's resources for small and midsize businesses both cover why, in language written for people without a technical team. The specific risk here is mundane: the numbers you make decisions from live on a device in a vehicle.
Measures deliberately not on the list
Social followers, page views and impressions. None of them changes a decision on a truck, and all of them rise and fall for reasons unconnected to whether anyone came.
Revenue on its own, which flatters a busy day with poor margins.
Average ticket value, which moves when the board changes and tells you very little on its own.
Percentage food cost, which is useful in a fixed kitchen and misleading here, because it ignores the block. A service can have an excellent food cost percentage and still lose money on the round trip.
What the numbers cannot settle
Whether the crew can sustain the schedule, which is a staffing judgment covered in what a service actually needs in people. What your permitted hours and conditions are at each pitch, which is recorded in the compliance register. And whether the answer to a good month is a second pitch, a second unit or nothing at all, which is a growth decision with its own criteria.
Common questions
Should I benchmark against other trucks?
No. Their block, pitch, crew size, round trip and board are different, so their numbers describe a different business. Compare each of your services against itself over time, which is a comparison you can actually act on.
What if I only have a few weeks of data?
Then you have a starting point rather than a trend. Record everything, resist drawing conclusions from single services, and expect the first month at a new pitch to be a measurement exercise rather than a normal trading month.
How do I measure people who left the queue?
Imperfectly. Count approaches and orders at the same service occasionally and watch the gap over time. It is rough, and it is better than assuming the gap is zero, which is what happens when nobody counts.
Is five too few?
Five that get collected honestly beat twenty that get collected sometimes. Add a sixth only when you find a decision you genuinely cannot make without it, and drop something if you find one of the five has not changed a decision in a season.
