Activity Collars: Beyond Heat Detection to Early Illness Alerts

Most UK dairy herds that bought activity collars did it for one reason: heats. Fewer missed bullings, fewer cows run on to a fourth service, a tighter calving pattern. That case pays for itself and most suppliers sell on it. The rumination figure on the same collar gets less attention, and it is probably the more useful number in the first three weeks after calving.

A cow coming down with ketosis, metritis or a displaced abomasum usually chews less, and often moves less, a day or more before anyone in the parlour sees a dull eye or a dropped yield. If you set your cow activity collar health monitoring up properly, the system can put that cow on a list before she goes off her feed. If you set it up badly, the list fills with 30 names every morning and within a month nobody reads it.

This piece covers that second problem as much as the first.

What the collar is actually measuring

Neck collars such as SenseHub Dairy (MSD Animal Health, formerly SCR Heatime), Nedap Smarttag Neck, Afimilk AfiCollar and the Lely Qwes-HR tag carry an accelerometer and, in most cases, a microphone or motion pattern model that picks up the jaw movement of cudding. CowManager does the same from an ear tag. smaXtec works differently: a bolus sits in the reticulum and measures rumen temperature and water intake alongside activity.

Two figures matter for illness:

• Rumination minutes per day. A healthy, settled cow in a UK TMR or partial-grazing system usually logs somewhere around 450 to 550 minutes. What counts is her own baseline, not the herd average.
• Activity relative to her normal. Heat pushes activity up. Sickness tends to pull it down, and a sick cow that’s also lying more will show both drops together.

The software compares each cow against her own rolling history and against cows at the same stage of lactation. SenseHub rolls this into a single “health index” score out of 100. Nedap and Lely flag deviations as health attentions. The labels change from one system to the next; the logic doesn’t.

How early is “early”?

The best-known field data comes from Cornell (Stangaferro and colleagues, 2016), who ran SCR health index alerts on transition cows. Their headline findings, roughly:

Read the bottom half of that table carefully. Metabolic and digestive problems hit rumination hard and early, so the collar catches them well. Metritis is patchier: a cow with a stinking discharge and a temperature of 39.8°C may still be cudding fairly normally. So the collar is a strong ketosis and DA detector and a partial metritis detector. Keep your day 5 to 10 checks of fresh cows. Use the collar list to decide which cows get looked at first, and which ones you look at twice.

Why does a two-day lead matter? A clinical ketosis case is commonly costed in the UK at £200 to £300 once you add lost milk, treatment and the knock-on risk of a DA. A DA that ends in surgery can pass £500 before you count the cows that never recover their yield. Picking up a cow while she’s still eating something, giving propylene glycol and a calcium bolus, and checking her rumen fill is far cheaper than a vet’s visit on day four.

The real failure: alert fatigue

Here’s how it goes wrong on a lot of farms. Collars go in, the supplier leaves default settings, and the morning health report looks like this:

Health report  07:12  Wed 08/10/2026
Cow   DIM  Group   Rum(min)  Rum chg 24h  Act chg  Health idx
1184    4  Fresh       312       -38%       -22%       61
0921    9  Fresh       401       -19%        -8%       79
1302   61  High        428       -11%        +4%       84
0477  143  Mid         389       -14%       -31%       83
1415  212  Low         410        -9%       -12%       85
1090    2  Fresh       355       -27%       -15%       72
... 23 more rows

Twenty-nine cows. Two or three of them are genuinely sick. The rest are cows that were moved pen yesterday, cows that went through the crush for foot trimming, a mid-lactation cow that had a quiet day, and a batch who lost an hour of rumination because the feeder was late. The herd team works through the list conscientiously for a week, finds nothing in most of them, and from then on the report is something somebody “glances at”. When cow 1184 turns up with a DA on Friday, the alert had been sitting there since Wednesday.

Picking the right collar won’t fix this. You fix it by setting thresholds and then checking them against what you actually find in the cows.

Setting thresholds that people will act on

Start from how much the team can genuinely check. On a 200-cow herd, a herdsperson can properly examine perhaps five to eight flagged cows a morning (temperature, rumen fill, ketone test, discharge check) without the rest of the routine falling apart. That number is your budget. Set the thresholds so the list fits inside it.

A workable approach for that herd:

1. Split the herd by stage of lactation. Fresh cows (0 to 21 days in milk) get sensitive settings, because that’s where ketosis, metritis and DAs sit. On SenseHub you’d keep the health index trigger at the default of 86 or tighten it slightly for this group. Cows past 60 DIM get a much harder threshold, something like a 25% rumination drop or a health index below 75, because the odds of a metabolic problem there are low and most alerts will be noise.

2. Require two signals, not one. A rumination dip alone after a pen move is normal. A rumination drop together with lower activity, or a drop that carries on into a second 24-hour period, is much more likely to be real. Nedap and Lely both let you combine conditions; on systems that don’t, the herd team can apply the rule by hand (“only act on cows on the list two days running, unless fresh”).

3. Tag management events. Log pen moves, foot trimming, vaccinations and TB tests in the software. Most systems will then suppress or flag alerts for those cows for 24 hours. A TB test day on 200 cows can generate dozens of false alerts if you skip this.

4. Watch group-level dips. If 40 cows all drop rumination by 10% on the same day, the problem is the ration, the mixer or the weather, not the cows. Check the group graph before you go round individual animals.

Done this way, the same Wednesday report shrinks to something like:

Fresh group (0-21 DIM), health idx < 86:   1184, 1090, 0921
All other groups, rum drop > 25% + act drop: 0477
Suppressed (pen move / trimming 07/10):     11 cows

That’s four cows. Every one gets examined, and the team can see the list is worth reading.

Check your settings against what you find

Thresholds aren’t set once. For the first eight weeks, write down what each flagged cow turned out to be: ketosis confirmed on a blood BHB meter (a reading of 1.2 mmol/L or above is the usual cut-off for subclinical), metritis, nothing found, and so on. A simple sheet, or a notes field in the herd software, does the job.

After 50 or so alerts you’ll have your own hit rate. If fewer than one in four flagged cows has anything wrong, tighten the threshold for that group. If cows are turning up clinically ill without having been flagged, loosen it, or look at whether those cases were mostly metritis, which the collar was always going to miss some of. Take the figures to your vet at the next herd health visit; most practices now work with collar data and some will review the settings for you.

Rough costs

Pricing varies a lot with herd size, existing parlour kit and whether you’re on a subscription. As a ballpark for a 200-cow UK herd in 2026:

If the collars are already on for heat detection, the health side costs you nothing extra in hardware. What it costs is a morning of setting up the groups and thresholds properly, and the discipline to record outcomes for two months. Preventing two DAs and a handful of clinical ketosis cases a year more than covers that on most herds.

For how collar data fits alongside bolus sensors, camera-based mobility scoring and lameness detection, the AI livestock health monitoring guide goes through the other options and where each one is strongest.

When you sit down with the software, start with your fresh group, your realistic daily checking capacity, and a sheet to record what each flagged cow actually had. Most farms that say their collars “only really do heats” have never had all three in place.