A church decides to offer translation, sets it up, announces it once, and then never finds out whether it worked. Six months later someone on the board asks whether it is worth continuing, and nobody can answer. The discussion that follows is about impressions — who remembers seeing someone use it — and impressions are a poor way to decide anything.
The numbers to answer that question are not complicated, and there are only four of them.
Four numbers worth watching
- How many people joined. The single most useful figure, and almost always higher than leaders expect. Translation is used privately, on a phone, by people who do not announce that they need it — which is precisely why it gets adopted and precisely why nobody notices.
- Which languages they chose. Reliably different from what the church assumed. This is the number that changes decisions.
- How long they stayed. Someone who joins and leaves after two minutes did not change their mind about needing translation. Something was wrong, and it is nearly always the same something.
- Whether it is growing. A spike after the announcement followed by a flat line means the announcement worked and the habit did not.
Why your numbers will disagree with each other
This trips up almost everyone, so it is worth understanding before you read anything into a dashboard. Different figures count different units, and they are all correct.
| Figure | What it counts | Why it differs |
|---|---|---|
| Listeners over 14 days | Every join across every service in the period | The same person on four Sundays is four joins |
| Peak during a service | The most people connected at any one instant | Always lower than the service total — people arrive and leave |
| By language | Person and language combinations | One person sampling two languages counts twice |
So a service can honestly report forty joins, a peak of twenty-six, and language figures adding to thirty-one. Nothing is broken. If you want one number to report to a board, use the peak during the service: it is the most conservative, the hardest to inflate, and the closest to "how many people were actually following at once".
The people your numbers cannot see
Every device-based count has the same blind spot: it counts devices, not people. In practice that means it undercounts, sometimes badly.
- A husband and wife sharing one phone are one device.
- A family of five round a tablet are one device.
- A screen at the back of an overflow room, or a TV in a side hall with a dozen people watching it, is one device.
- Anyone who ran out of battery halfway through disappears from your numbers and not from your congregation.
If you know a group is watching one screen, account for it rather than pretending the count is exact. The point of measurement here is not precision to the person — it is knowing whether the answer is four or forty.
The languages will surprise you
The most common discovery is that a language nobody planned for is the most used one. Churches set up the two or three languages they can see in the congregation, and then find the largest group is reading in a fourth.
There are usually two reasons, and both are worth understanding.
The first is that reading and listening are different skills. Plenty of people follow spoken English well enough to get by and find it genuinely tiring over forty minutes. Given the option they will read in the language they think in, not because they cannot cope but because it is less work. That group never asks for translation, because they do not consider themselves to need it.
The second is visitors you were not counting. Tourists, students, someone's family over for a fortnight, a worker on a short contract. They show up in the numbers and nowhere else.
Either way, the action is the same: once a language appears in your figures for three or four weeks running, stop treating it as an experiment. Put it on the welcome slide, get one person in the congregation who speaks it to be the human contact for it, and check whether your website says anything at all in it.
Reading the drop-offs
Short sessions are the most actionable signal you have, because the cause is almost always upstream and almost always fixable.
- Audio. By some distance the most common. A microphone too far from the speaker produces text that is unreliable rather than absent — quiet input tends to yield confident sentences that were never said, which readers abandon quickly and rightly. If drop-offs are high, look at the microphone before anything else.
- Delay. A couple of seconds behind the speaker is comfortable. Much more and following becomes work, especially when the reader can also hear the room.
- Nothing to join. Check the obvious: was the broadcast actually running for the whole service, or started ten minutes late?
- The code. If the QR code is only on a screen that gets replaced by song lyrics, latecomers cannot join at all.
What a healthy pattern looks like
Over the first two months, expect something like a spike in week one when you announce it, a dip in week two, and then a slow climb as it becomes ordinary. The climb is the real signal. A flat line after the spike usually means people tried it once and it did not work well enough to come back to — which sends you back to the microphone.
Also watch for the shape changing at a guest service or a festival. Christmas, Easter and baptisms bring in exactly the visitors translation exists for, and those weeks tell you more about your reach than an ordinary Sunday does.
What not to measure
Resist turning this into a performance metric. The goal is not a rising graph; it is that the people who need it can follow. A service where four people out of two hundred used translation, and those four were the family who had arrived in the country a fortnight earlier, was a success. Numbers are here to answer "is this reaching anyone and is it working properly", not to justify the decision.
And when the numbers have told you what they can, ask one person. Find someone who used it and ask what it was like — whether the text kept up, whether it made sense, whether they would rather read or listen. Ten minutes of that will explain your drop-off rate better than any dashboard.
Look at your own numbers
InMyTongue records which languages were used, how many people joined each session and how many were connected at once, so the question can be answered with data rather than impressions. New accounts include free credit. See what your services would show.