TL;DR
- A completion rate on its own does not tell you whether orders were right, whether guests left early, or how the number was counted.
- Read completion rate, abandonment rate and intent accuracy together, and always with accuracy beside completion.
- Five questions show how any vendor built its number: the denominator, the store sample, the time window, how a corrected item counts, and what the figure leaves out.
- Hi Auto reports 93%+ completion and 96% accuracy as an average across all live stores, on 100M+ orders processed per year.
The slide has one number on it. A completion rate in the nineties, in large type. Nobody in the room asks what it counts, and the vendor review moves on.
That moment is where a lot of avoidable surprises start. Completion rate tells you how many orders reached the POS without staff intervention. It does not tell you how many were right, how many guests left first, or how the figure was built.
This guide covers three metrics: completion rate, abandonment rate and intent accuracy. You will see what each one hides when read alone, and get the questions that show how any vendor produced its number.
Three metrics, one drive-thru lane
Picture a single car moving through the lane. Each metric watches a different moment.
- Intent accuracy watches the start. Did the system understand what the guest meant, and not only which words it heard? The glossary entry for intent accuracy covers the definition.
- Abandonment rate watches the middle. Did the guest drive away before finishing? See abandonment rate.
- Completion rate watches the end. Did the order reach the POS without staff intervention? The glossary has two close entries, completion rate and order completion rate. Ask which one a report uses, and how it defines “finished”.
Accuracy sits at the very end of the lane. Was what went in the bag what the guest asked for? Accuracy rate and order accuracy cover that.
Here are the two working definitions. Completion rate is the percentage of orders sent to the POS without staff intervention. Accuracy rate is the percentage of orders sent to the POS with every item and customization exactly right. The glossary defines the other terms. This guide is about how to read them together.
Why completion rate needs accuracy beside it
A system can finish a high share of orders and still get items wrong. A wrong bag means a remake, a longer wait and a guest who notices at the window. The order was completed. The result was not good.
The opposite also happens. A system can be accurate on the orders it finishes and still lose guests before they reach the end of the order. Either number alone flatters the system, so read them as a pair. A vendor report that shows only one of them is telling you half the story.
That is how we report ours. Hi Auto’s numbers are 93%+ completion and 96% accuracy, as an average across all live stores, on 100M+ orders processed per year. It averages every live store, so no best-location or pilot results are mixed in.
When a vendor quotes a completion number, your first follow-up is simple: what was the accuracy on those orders? Read the two together, at scale, across all live stores.
What abandonment and intent accuracy add
Completion rate only counts orders that got far enough to be counted. A guest who pulls up, says a few words and drives away may never enter the calculation. Whether they do depends on how a vendor defines the start of an order.
Abandonment rate brings those guests back into view. Ask any vendor to report it, and ask where an order attempt begins. If the answer is vague, the completion number has no floor under it.
Intent accuracy looks at understanding. A guest says “no pickles”. The system hears every word and attaches the change to the wrong item, so the words were captured and the meaning was missed.
Final order accuracy can hide this, because a correction late in the order can fix the ticket even when the conversation went sideways. That makes intent accuracy a harder number to flatter.
Independent research points to why being understood the first time is worth measuring. In the Intouch Insight, 2026 Drive-Thru Study (mystery shops conducted June and July 2026), guests at locations of three national QSR brands running voice AI had to repeat part of their order on 29.1% of visits, against 20.4% for the same brands in the main study. That is a finding about AI locations in the study in general, not a Hi Auto figure.
Five questions that show how a number was made
Any headline figure is only as good as the method behind it. Ask these five questions of any vendor, Hi Auto included.
- What is the denominator? Does the count include every car that reached the speaker, or only the orders the system chose to attempt? Leaving out the hard interactions raises the rate.
- Which stores are in the sample? Is it an average across all live stores, or a best location or a pilot? “Up to” is not an average.
- What time window does it cover? A rolling month and a single strong week can look very different for the same system. Ask for the dates.
- How does a corrected item count? A guest changes their mind mid-order and the system updates the item. Is that a success, an error, or both? The answer moves completion and accuracy at the same time.
- What does the figure leave out? Are any orders, stores or interaction types excluded from the count, and is that stated beside the number?
The fifth question matters because definitions differ. Hi Auto counts a completion as an order sent to the POS without staff intervention. QSR Magazine’s 2026 Drive-Thru Report (Danny Klein, 1 October 2026) looked at the 15 drive-thrus shopped where voice AI could not continue the order. The stated reasons were that the AI could not understand a customization request (53.3%) or could not answer a question (40%). That is a study finding, not a Hi Auto figure.
How the figure is counted matters. Two vendors can post the same headline number and mean very different things by it.
Hi Auto’s Buyer’s Guide sets a 90% completion threshold, to be read next to accuracy, and advises asking vendors for their measurement methodology. Its evaluation matrix, on the last page, gives you a place to line the answers up side by side.
A simple dashboard to ask for
Once you trust the number, ask to see it sliced. A useful drive-thru view shows:
- Completion and accuracy together, by store and by daypart
- Average ticket, by store
- Upsell consistency
- Store-level events such as early closes, delivery-only periods and missed orders
A chain-wide average can hide three struggling stores or a weak late-night daypart. The slice shows you where to look.
Ask for the trend too. A number that holds week after week tells you more than one strong reading, and a dip at a single store is a lead you can follow up.
Hi Auto reports four drive-thru KPIs: completion, accuracy, average ticket and upsell consistency. The Hi Auto product also includes a dedicated dashboard with insights such as early closes, delivery-only periods, missed orders and other store-level nuances.
Jason Amell, who leads operations for a 27-store Checkers & Rally’s group in Michigan, describes what that visibility is worth:
“It gives you insight into your restaurant when you’re not there. That is probably more valuable to me than even the system that takes the orders.”
Jason Amell, Director of Operations, JASAM Enterprises (Checkers & Rally’s, 27 stores, Michigan). Hi Auto customer video “Real Stories from the Drive-Thru Frontline”, 2025. Watch Jason Amell on what store visibility is worth
Ask how the number was made before you ask how big it is
A big number is easy to put on a slide. A well-made number takes more work. A vendor who can answer the denominator, sample, window, correction and exclusion questions clearly understands their own system.
The standard we suggest is simple. Completion and accuracy together, averaged across all live stores, over a stated window, with a clear denominator.
The useful question is whether a system can hold 93%+ completion and 96% accuracy together, across 100M+ orders a year. That is the pairing we report. Once you ask it that way, the slide with one number on it stops being enough.
Compare vendor numbers side by side
Download the Buyer’s Guide, AI for the Drive-Thru. Its evaluation matrix is built for putting any vendor’s numbers side by side. Bring it to your next vendor call, then talk to the Hi Auto team about what 93%+ completion and 96% accuracy look like across 100M+ orders processed per year.