Mating Insights

Halter · 2026

Machine learning

Machine Learning

Field Research

Data Visualisation

I led the end-to-end redesign of Halter’s core mating experience, working across product, ML and farmer research to turn complex heat-model data into clear, time-sensitive breeding decisions.The resulting optimum insemination window gave farmers a simple, actionable view of when to breed each cow.

The challenge

Making sense of heat signals

Halter’s smart collars detect changes in cow behaviour, but farmers still need to interpret those signals to make breeding decisions.

Open app

During mating season, farmers have a narrow window to identify cows on heat and decide when to inseminate.

Turning data into confidence

The challenge was translating complex heat-model data into clear, actionable information that farmers could trust in the field.

The Problem

The old Heat List made time-critical decisions harder

During peak mating season, farmers need to quickly identify cows ready for artificial insemination. However, the existing Heat List presented dense reproductive data without clearly showing when each cow should be inseminated.

​

Farmers had to interpret heat events and cross-reference information to determine the best time to act. The opportunity was to bring that timing guidance directly into the experience, making it easier to prioritise cows during busy mating periods.

9:41

Heats

B

34-3431

Since 4 ago

First AI

Conventional

Confidence

70%

Not submitted

7

1% of cows

Submitted

57

6% of cows

Filter

Not submitted

7 results

Peak

Pre

Pre

Pre

Pre

+ 2h

+ 12h

+ 16h

Heats

3

3

3

3

3

3

3

Last AI

21d

21d

21d

11d

21d

21d

21d

AI

1

1

1

1

1

1

1

Cow

M

2734

M

9845

M

4736

M

8734

M

5623

M

3456

M

4587

The Technology

What Makes Mating with Halter Possible?

Halter detects heat events by analysing continuous behavioural data captured by smart collars. Its machine-learning model identifies changes in activity associated with heat, operating at a false-positive rate of less than 1%.

​

This provided a reliable foundation for the redesign. My focus was on how to present these signals alongside insemination timing, so farmers could understand what the data meant for their cows.

A great detection model means little if the experience doesn’t meet farmers’ needs

Discovery & Field Research

Mapping mating season as it actually runs

I spoke directly with dairy farmers through remote interviews and on-farm visits to understand how they managed mating season. I focused on their daily routines, the decisions they needed to make, and where the existing experience fell short.

Farmers wanted more than just knowing when their cows were on heat.

​

They wanted to make their own decisions on when to inseminate.

Collaboration

Working with Behaviour and ML

That insight shaped my collaboration with the Behaviour and ML teams. Together, we explored how the heat model worked, what data was available, and how it could support farmers’ insemination decisions.

​

I used that understanding to explore different ways of visualising the insemination window, from linear progress bars to bell-curve distributions.

Cow activity

Mob avg.

Halter alert

Insemination window

0

6

12

18

24

30

36

42

48

Designs Informed by Testing

From multiple concepts to one clear bar

I tested different visualisations directly with farmers to understand which made the insemination window easiest to interpret.

The horizontal timing bar emerged as the strongest direction. It made the recommended insemination window easy to identify at a glance, while still showing the broader period in which farmers could act.

Best 12 hr

Best 24 hr

Insemination window

Heat window

Refining the Design

Refining the bar for real farm use

The bar had to show every insemination window and its likelihood of conception, while still encouraging farmers to act outside the best window when needed. Every window stays green, so none looks unsuitable. Size, not colour, tells them apart, and the optimum window is easy to spot without hiding the rest.

12pm

5pm

12am

12pm

5pm

7h

into heat

I continued refining the design through farmer feedback, sharing findings with the wider team as the design evolved. I also added clock times for each window and an “hours into heat” indicator, so farmers could see exactly where a cow was in her heat cycle without calculating the timing themselves.

Key Trade-off

Context vs. legibility in the field

The bar worked for a single cow, but farmers manage whole herds. The next challenge was the list around it: how much context to show per cow before the screen became unreadable in the field.

Before

The table showed heat counts and AI history, but not what farmers needed to decide: no optimum window, no heat interval, and no calving status.

Peak

Pre

Pre

Pre

Heats

3

3

3

Last AI

21d

21d

21d

AI

1

1

1

Cow

M

2734

M

9845

M

4736

After

I replaced the dense table rows with scannable cards that prioritised the insemination window and heat interval, while keeping important reproductive information visible at a glance.

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

A Full Suite of Features

Operational visibility all season

Beyond the optimum window chart, I designed a unified mating experience that gives farmers complete operational visibility throughout the season. By combining real-time heat alerts, macro performance analytics, and individual cow reproductive histories, the platform equips farmers to act on data at every stage of the breeding cycle.

Macro performance analytics

See how the herd is cycling across the season.

Real-time heat alerts

Know the moment a cow enters heat.

Cow reproductive histories

Individual context behind every decision.

Validating the Designs

96% positive, straight from the field.

During beta, I tracked farmer feedback to understand how the redesigned experience was being received. The results were overwhelmingly positive, with 96% favourable ratings.

​

This feedback helped validate the direction and supported the continued rollout of the experience.

Learning from Usage

Learning from how farmers actually used it

While farmer feedback showed that the redesign was well received, I wanted to understand how it was being used in practice.

​

I analysed in-app events to identify which actions farmers returned to most, which features saw limited engagement, and where the experience could be simplified further.

Recording Inseminations Dominated Usage

Recording inseminations accounted for nearly 65% of the actions shown, followed by triggering collar LEDs at 19%. This reinforced how central these everyday actions were to the mating workflow.

Filtering Saw Limited Repeat Use

While multi-select actions were frequently used, filtering saw much less engagement. When farmers did filter, mob accounted for the majority of filter events.

Simplifying Filters

Cutting the filter down to what farmers used

Only 4.3% of adopters used filtering on three or more days, and mob accounted for most filter events. This showed an opportunity to simplify the controls around how farmers actually used the Heat List, rather than maintaining a long list of rarely used options.

Before

Filtering and sorting were handled through separate controls, each presenting a long list of options. This added unnecessary complexity to the interface and made it harder for farmers to quickly narrow down the cows they needed to focus on.

9:41

Heats

Not inseminated

Inseminated

Label

(5)

Label (2)

Status: Regular

Alert time

Select

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

Mixed age

Light

9:41

Heats

Not inseminated

Inseminated

Label

Filter

Alert time

Select

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

M

222

Regular return

Last heat

24

d

Calved

-

d

Heats

6

AIs

2

M

412

In-calf heat

Last heat

21

d

Calved

151

d

Heats

7

AIs

6

M

56

Low confidence

Last heat

25

d

Calved

134

d

Heats

5

AIs

3

Sort by

Mob

Ear tag

Last heat

Alert time

Calved

Heats

AIs

9:41

Heats

Not inseminated

Inseminated

Label

Filter

Alert time

Select

Reset all

Apply

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

M

222

Regular return

Last heat

24

d

Calved

-

d

Heats

6

AIs

2

M

412

In-calf heat

Last heat

21

d

Calved

151

d

Heats

7

AIs

6

M

56

Low confidence

Last heat

25

d

Calved

134

d

Heats

5

AIs

3

Filters

Mob

Status

Alert time

Last heat

Calved

Heats

AIs

Label

Reset all

Apply

After

I brought sorting and filtering together under a single control and reduced the available filters to those most relevant to farmers’ workflows. This removed unnecessary controls from the Heat List while keeping the options farmers needed accessible when required.

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

3

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

Sort & filter

Filter

Mob

Return interval

Labels

Sort by

Alert time

Clear

Apply

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

Sort by

Alert time

Ear tag

Mob

Bringing It All Together

Where the experience landed

The final Heat List brought together detection confidence, insemination timing and reproductive information in one place, giving farmers a clearer view of each cow without having to navigate between different screens.

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

Cow

M

31

M

341

M

142

M

54

M

31

M

2433

B

6

M

2418

C

213

S

12

Conf.

98%

85%

100%

99%

89%

100%

95%

97%

17%

15%

Heat

Reg

Reg

Reg

Short

In-calf

Reg

Reg

Long

Short

Short

AI window

Identify

See how the herd is cycling across the season.

Decide

Understand heat confidence and the recommended insemination window.

Act

Use that guidance to make timely insemination decisions.

Impact

How the experience improved

Faster daily decision making

Farmers could identify cows requiring attention without spending time cross-referencing information.

Simpler herd triage

Farmers could quickly prioritise cows for insemination and organise their daily mating activities.

More confident breeding decisions

Farmers could make better-informed insemination decisions and avoid unnecessary use of sexed semen straws.

Giving farmers the flexibility to make decisions in their own way

Impact

The results behind the redesign

25%

reduction in sexed semen usage

More targeted insemination decisions helped reduce unnecessary use of expensive sexed semen.

~$4.6k

average seasonal savings per farm

Reducing reliance on costly sexed semen helped farmers lower their breeding expenses.

96%

positive farmer feedback

Farmers responded positively to clearer heat insights and more actionable mating information, across 285 ratings.

<1%

false positives

Reliable heat detection helped farmers trust the information behind their mating decisions.

IN THE FARMERS' WORDS

“Hey Sang, just a quick message with some feedback now that we have done a couple of weeks of mating - long story short the improvements are awesome!! So much better than last year and making the daily AI routine easy!!! Thank you!”

Ryan

Kanadale Farm · 2026

“Really liking the changes to the mating list. Being able to see how confident we should be in each heat has been super helpful, especially when deciding which cows to use sexed semen on.”

Will

Wilco · 2026

swmoonnz@gmail.com

Mating Insights

Machine learning

Machine Learning

Field Research

Data Visualisation

I led the end-to-end redesign of Halter’s core mating experience, working across product, ML and farmer research to turn complex heat-model data into clear, time-sensitive breeding decisions.The resulting optimum insemination window gave farmers a simple, actionable view of when to breed each cow.

Halter · 2026

The challenge

Making sense of heat signals

Halter’s smart collars detect changes in cow behaviour, but farmers still need to interpret those signals to make breeding decisions.

Open app

During mating season, farmers have a narrow window to identify cows on heat and decide when to inseminate.

Turning data into confidence

The challenge was translating complex heat-model data into clear, actionable information that farmers could trust in the field.

The Problem

The old Heat List made time-critical decisions harder

During peak mating season, farmers need to quickly identify cows ready for artificial insemination. However, the existing Heat List presented dense reproductive data without clearly showing when each cow should be inseminated.

​

Farmers had to interpret heat events and cross-reference information to determine the best time to act. The opportunity was to bring that timing guidance directly into the experience, making it easier to prioritise cows during busy mating periods.

9:41

Heats

B

34-3431

Since 4 ago

First AI

Conventional

Confidence

70%

Not submitted

7

1% of cows

Submitted

57

6% of cows

Filter

Not submitted

7 results

Peak

Pre

Pre

Pre

Pre

+ 2h

+ 12h

+ 16h

Heats

3

3

3

3

3

3

3

Last AI

21d

21d

21d

11d

21d

21d

21d

AI

1

1

1

1

1

1

1

Cow

M

2734

M

9845

M

4736

M

8734

M

5623

M

3456

M

4587

The Technology

How Halter detects heat events

Halter detects heat events by analysing continuous behavioural data captured by smart collars. Its machine-learning model identifies changes in activity associated with heat, operating at a false-positive rate of less than 1%.

​

This provided a reliable foundation for the redesign. My focus was on how to present these signals alongside insemination timing, so farmers could understand what the data meant for their cows.

A great detection model means little if the experience doesn’t meet farmers’ needs

Discovery & Field Research

Mapping mating season as it actually runs

I spoke directly with dairy farmers through remote interviews and on-farm visits to understand how they managed mating season. I focused on their daily routines, the decisions they needed to make, and where the existing experience fell short.

Farmers wanted more than just knowing when their cows were on heat.

​

They wanted to make their own decisions on when to inseminate.

Collaboration

Working with Behaviour and ML

That insight shaped my collaboration with the Behaviour and ML teams. Together, we explored how the heat model worked, what data was available, and how it could support farmers’ insemination decisions.

​

I used that understanding to explore different ways of visualising the insemination window, from linear progress bars to bell-curve distributions.

Cow activity

Mob avg.

Halter alert

Insemination window

0

6

12

18

24

30

36

42

48

Designs Informed by Testing

From multiple concepts to one clear bar

I tested different visualisations directly with farmers to understand which made the insemination window easiest to interpret.

Best 12 hr

Best 24 hr

Insemination window

Heat window

The horizontal timing bar emerged as the strongest direction. It made the recommended insemination window easy to identify at a glance, while still showing the broader period in which farmers could act.

Refining the Design

Refining the bar for real farm use

The bar had to show every insemination window and its likelihood of conception, while still encouraging farmers to act outside the best window when needed. Every window stays green, so none looks unsuitable. Size, not colour, tells them apart, and the optimum window is easy to spot without hiding the rest.

12pm

5pm

12am

12pm

5pm

7h

into heat

I continued refining the design through farmer feedback, sharing findings with the wider team as the design evolved. I also added clock times for each window and an “hours into heat” indicator, so farmers could see exactly where a cow was in her heat cycle without calculating the timing themselves.

Key Trade-off

Context vs. legibility in the field

The bar worked for a single cow, but farmers manage whole herds. The next challenge was the list around it: how much context to show per cow before the screen became unreadable in the field.

Peak

Pre

Pre

Pre

Heats

3

3

3

Last AI

21d

21d

21d

AI

1

1

1

Cow

M

2734

M

9845

M

4736

Before

The table showed heat counts and AI history, but not what farmers needed to decide: no optimum window, no heat interval, and no calving status.

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

After

I replaced the dense table rows with scannable cards that prioritised the insemination window and heat interval, while keeping important reproductive information visible at a glance.

A Full Suite of Features

Operational visibility all season

Beyond the optimum window chart, I designed a unified mating experience that gives farmers complete operational visibility throughout the season. By combining real-time heat alerts, macro performance analytics, and individual cow reproductive histories, the platform equips farmers to act on data at every stage of the breeding cycle.

Macro performance analytics

See how the herd is cycling across the season.

Real-time heat alerts

Know the moment a cow enters heat.

Cow reproductive histories

Individual context behind every decision.

Validating the Designs

96% positive, straight from the field.

During beta, I tracked farmer feedback to understand how the redesigned experience was being received. The results were overwhelmingly positive, with 96% favourable ratings.

​

This feedback helped validate the direction and supported the continued rollout of the experience.

Learning from Usage

Learning from how farmers actually used it

While farmer feedback showed that the redesign was well received, I wanted to understand how it was being used in practice.

​

I analysed in-app events to identify which actions farmers returned to most, which features saw limited engagement, and where the experience could be simplified further.

Recording Inseminations Dominated Usage

Recording inseminations accounted for nearly 65% of the actions shown, followed by triggering collar LEDs at 19%. This reinforced how central these everyday actions were to the mating workflow.

Filtering Saw Limited Repeat Use

While multi-select actions were frequently used, filtering saw much less engagement. When farmers did filter, mob accounted for the majority of filter events.

Simplifying Filters

Cutting the filter down to what farmers used

Only 4.3% of adopters used filtering on three or more days, and mob accounted for most filter events. This showed an opportunity to simplify the controls around how farmers actually used the Heat List, rather than maintaining a long list of rarely used options.

Before

Filtering and sorting were handled through separate controls, each presenting a long list of options. This added unnecessary complexity to the interface and made it harder for farmers to quickly narrow down the cows they needed to focus on.

9:41

Heats

Not inseminated

Inseminated

Label

(5)

Label (2)

Status: Regular

Alert time

Select

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

Mixed age

Light

9:41

Heats

Not inseminated

Inseminated

Label

Filter

Alert time

Select

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

M

222

Regular return

Last heat

24

d

Calved

-

d

Heats

6

AIs

2

M

412

In-calf heat

Last heat

21

d

Calved

151

d

Heats

7

AIs

6

M

56

Low confidence

Last heat

25

d

Calved

134

d

Heats

5

AIs

3

Sort by

Mob

Ear tag

Last heat

Alert time

Calved

Heats

AIs

9:41

Heats

Not inseminated

Inseminated

Label

Filter

Alert time

Select

Reset all

Apply

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

M

222

Regular return

Last heat

24

d

Calved

-

d

Heats

6

AIs

2

M

412

In-calf heat

Last heat

21

d

Calved

151

d

Heats

7

AIs

6

M

56

Low confidence

Last heat

25

d

Calved

134

d

Heats

5

AIs

3

Filters

Mob

Status

Alert time

Last heat

Calved

Heats

AIs

Label

Reset all

Apply

After

I brought sorting and filtering together under a single control and reduced the available filters to those most relevant to farmers’ workflows. This removed unnecessary controls from the Heat List while keeping the options farmers needed accessible when required.

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

3

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

Sort & filter

Filter

Mob

Return interval

Labels

Sort by

Alert time

Clear

Apply

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

Sort by

Alert time

Ear tag

Mob

Bringing It All Together

Where the experience landed

The final Heat List brought together detection confidence, insemination timing and reproductive information in one place, giving farmers a clearer view of each cow without having to navigate between different screens.

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

Cow

M

31

M

341

M

142

M

54

M

31

M

2433

B

6

M

2418

C

213

S

12

Conf.

98%

85%

100%

99%

89%

100%

95%

97%

17%

15%

Heat

Reg

Reg

Reg

Short

In-calf

Reg

Reg

Long

Short

Short

AI window

Identify

See how the herd is cycling across the season.

Decide

Understand heat confidence and the recommended insemination window.

Act

Use that guidance to make timely insemination decisions.

Impact

How the experience improved

Faster daily decision making

Farmers could identify cows requiring attention without spending time cross-referencing information.

Simpler herd triage

Farmers could quickly prioritise cows for insemination and organise their daily mating activities.

More confident breeding decisions

Farmers could make better-informed insemination decisions and avoid unnecessary use of sexed semen straws.

Giving farmers the flexibility to make decisions in their own way

Impact

The results behind the redesign

25%

reduction in sexed semen usage

More targeted insemination decisions helped reduce unnecessary use of expensive sexed semen.

~$4.6k

average seasonal savings per farm

Reducing reliance on costly sexed semen helped farmers lower their breeding expenses.

96%

positive farmer feedback

Farmers responded positively to clearer heat insights and more actionable mating information, across 285 ratings.

<1%

false positives

Reliable heat detection helped farmers trust the information behind their mating decisions.

IN THE FARMERS' WORDS

“Hey Sang, just a quick message with some feedback now that we have done a couple of weeks of mating - long story short the improvements are awesome!! So much better than last year and making the daily AI routine easy!!! Thank you!”

Ryan

Kanadale Farm · 2026

“Really liking the changes to the mating list. Being able to see how confident we should be in each heat has been super helpful, especially when deciding which cows to use sexed semen on.”

Will

Wilco · 2026

swmoonnz@gmail.com

Mating Insights

Machine learning

Hardware

Field Research

Data Visualisation

I led the end-to-end redesign of Halter's core mating experience, working across product, ML and farmer research to turn complex heat-model data into clear, time-sensitive breeding decisions.

Halter · 2026

The challenge

Making sense of heat signals

Halter’s smart collars detect changes in cow behaviour, but farmers still need to interpret those signals to make breeding decisions.

Time-critical decisions

During mating season, farmers have a narrow window to identify cows on heat and decide when to inseminate.

Turning data into confidence

The challenge was translating complex heat-model data into clear, actionable information that farmers could trust in the field.

The Problem

The old Heat List made time-critical decisions harder

During peak mating season, farmers need to quickly identify cows ready for artificial insemination. However, the existing Heat List presented dense reproductive data without clearly showing when each cow should be inseminated.

​

Farmers had to interpret heat events and cross-reference information to determine the best time to act. The opportunity was to bring that timing guidance directly into the experience, making it easier to prioritise cows during busy mating periods.

9:41

Heats

B

34-3431

Since 4 ago

First AI

Conventional

Confidence

70%

Not submitted

7

1% of cows

Submitted

57

6% of cows

Filter

Not submitted

7 results

Peak

Pre

Pre

Pre

Pre

+ 2h

+ 12h

+ 16h

Heats

3

3

3

3

3

3

3

Last AI

21d

21d

21d

11d

21d

21d

21d

AI

1

1

1

1

1

1

1

Cow

M

2734

M

9845

M

4736

M

8734

M

5623

M

3456

M

4587

The Technology

How Halter detects heat events

Halter detects heat events by analysing continuous behavioural data captured by smart collars. Its machine-learning model identifies changes in activity associated with heat, operating at a false-positive rate of less than 1%.

​

This provided a reliable foundation for the redesign. My focus was on how to present these signals alongside insemination timing, so farmers could understand what the data meant for their cows.

A great detection model means little if the experience doesn’t meet farmers’ needs

Discovery & Field Research

Mapping mating season as it actually runs

I spoke directly with dairy farmers through remote interviews and on-farm visits to understand how they managed mating season. I focused on their daily routines, the decisions they needed to make, and where the existing experience fell short.

Farmers wanted more than just knowing when their cows were on heat.

​

They wanted to make their own decisions on when to inseminate.

Collaboration

Working with Behaviour and ML

That insight shaped my collaboration with the Behaviour and ML teams. Together, we explored how the heat model worked, what data was available, and how it could support farmers’ insemination decisions.

​

I used that understanding to explore different ways of visualising the insemination window, from linear progress bars to bell-curve distributions.

Cow activity

Mob avg.

Halter alert

Insemination window

0

6

12

18

24

30

36

42

48

Designs Informed by Testing

From multiple concepts to one clear bar

I tested different visualisations directly with farmers to understand which made the insemination window easiest to interpret.

Best 12 hr

Best 24 hr

Insemination window

Heat window

The horizontal timing bar emerged as the strongest direction. It made the recommended insemination window easy to identify at a glance, while still showing the broader period in which farmers could act.

Refining the Design

Refining the bar for real farm use

The bar had to show every insemination window and its likelihood of conception, while still encouraging farmers to act outside the best window when needed. Every window stays green, so none looks unsuitable. Size, not colour, tells them apart, and the optimum window is easy to spot without hiding the rest.

12pm

5pm

12am

12pm

5pm

7h

into heat

I continued refining the design through farmer feedback, sharing findings with the wider team as the design evolved. I also added clock times for each window and an “hours into heat” indicator, so farmers could see exactly where a cow was in her heat cycle without calculating the timing themselves.

Key Trade-off

Context vs. legibility in the field

The bar worked for a single cow, but farmers manage whole herds. The next challenge was the list around it: how much context to show per cow before the screen became unreadable in the field.

Peak

Pre

Pre

Pre

Heats

3

3

3

Last AI

21d

21d

21d

AI

1

1

1

Cow

M

2734

M

9845

M

4736

Before

The table showed heat counts and AI history, but not what farmers needed to decide: no optimum window, no heat interval, and no calving status.

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

After

I replaced the dense table rows with scannable cards that prioritised the insemination window and heat interval, while keeping important reproductive information visible at a glance.

A Full Suite of Features

Operational visibility all season

Beyond the optimum window chart, I designed a unified mating experience that gives farmers complete operational visibility throughout the season. By combining real-time heat alerts, macro performance analytics, and individual cow reproductive histories, the platform equips farmers to act on data at every stage of the breeding cycle.

Macro performance analytics

See how the herd is cycling across the season.

Real-time heat alerts

Know the moment a cow enters heat.

Cow reproductive histories

Individual context behind every decision.

Validating the Designs

96% positive, straight from the field.

During beta, I tracked farmer feedback to understand how the redesigned experience was being received. The results were overwhelmingly positive, with 96% favourable ratings.

​

This feedback helped validate the direction and supported the continued rollout of the experience.

Learning from Usage

Learning from how farmers actually used it

While farmer feedback showed that the redesign was well received, I wanted to understand how it was being used in practice.

​

I analysed in-app events to identify which actions farmers returned to most, which features saw limited engagement, and where the experience could be simplified further.

Recording Inseminations Dominated Usage

Recording inseminations accounted for nearly 65% of the actions shown, followed by triggering collar LEDs at 19%. This reinforced how central these everyday actions were to the mating workflow.

Filtering Saw Limited Repeat Use

While multi-select actions were frequently used, filtering saw much less engagement. When farmers did filter, mob accounted for the majority of filter events.

Simplifying Filters

Cutting the filter down to what farmers used

Only 4.3% of adopters used filtering on three or more days, and mob accounted for most filter events. This showed an opportunity to simplify the controls around how farmers actually used the Heat List, rather than maintaining a long list of rarely used options.

Before

Filtering and sorting were handled through separate controls, each presenting a long list of options. This added unnecessary complexity to the interface and made it harder for farmers to quickly narrow down the cows they needed to focus on.

9:41

Heats

Not inseminated

Inseminated

Label

(5)

Label (2)

Status: Regular

Alert time

Select

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

Mixed age

Light

9:41

Heats

Not inseminated

Inseminated

Label

Filter

Alert time

Select

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

M

222

Regular return

Last heat

24

d

Calved

-

d

Heats

6

AIs

2

M

412

In-calf heat

Last heat

21

d

Calved

151

d

Heats

7

AIs

6

M

56

Low confidence

Last heat

25

d

Calved

134

d

Heats

5

AIs

3

Sort by

Mob

Ear tag

Last heat

Alert time

Calved

Heats

AIs

9:41

Heats

Not inseminated

Inseminated

Label

Filter

Alert time

Select

Reset all

Apply

M

341

Regular return

Last heat

19

d

Calved

141

d

Heats

0

AIs

3

M

222

Regular return

Last heat

24

d

Calved

-

d

Heats

6

AIs

2

M

412

In-calf heat

Last heat

21

d

Calved

151

d

Heats

7

AIs

6

M

56

Low confidence

Last heat

25

d

Calved

134

d

Heats

5

AIs

3

Filters

Mob

Status

Alert time

Last heat

Calved

Heats

AIs

Label

Reset all

Apply

After

I brought sorting and filtering together under a single control and reduced the available filters to those most relevant to farmers’ workflows. This removed unnecessary controls from the Heat List while keeping the options farmers needed accessible when required.

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

3

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

Sort & filter

Filter

Mob

Return interval

Labels

Sort by

Alert time

Clear

Apply

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

Sort by

Alert time

Ear tag

Mob

Bringing It All Together

Where the experience landed

The final Heat List brought together detection confidence, insemination timing and reproductive information in one place, giving farmers a clearer view of each cow without having to navigate between different screens.

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

M

31

Heat

Reg · 21d

Calved

138d

Heats

1

AIs

0

M

341

Heat

Short · 18d

Calved

40d

Heats

1

AIs

0

M

142

Heat

In-calf

Calved

54d

Heats

2

AIs

1

M

54

Heat

Long · 27d

Calved

120d

Heats

2

AIs

1

M

512

Heat

Reg · 21d

Calved

120d

Heats

2

AIs

1

9:41

Heats

Not inseminated

Inseminated

Missed inseminations

Select

Cow

M

31

M

341

M

142

M

54

M

31

M

2433

B

6

M

2418

C

213

S

12

Conf.

98%

85%

100%

99%

89%

100%

95%

97%

17%

15%

Heat

Reg

Reg

Reg

Short

In-calf

Reg

Reg

Long

Short

Short

AI window

Identify

See which cows are on heat and need attention.

Decide

Understand heat confidence and the recommended insemination window.

Act

Use that guidance to make timely insemination decisions.

Impact

How the experience improved

Faster daily decision making

Farmers could identify cows requiring attention without spending time cross-referencing information.

Simpler herd triage

Farmers could quickly prioritise cows for insemination and organise their daily mating activities.

More confident breeding decisions

Farmers could make better-informed insemination decisions and avoid unnecessary use of sexed semen straws.

Giving farmers the flexibility to make decisions in their own way

Impact

The results behind the redesign

25%

reduction in sexed semen usage

More targeted insemination decisions helped reduce unnecessary use of expensive sexed semen.

~$4.6k

average seasonal savings per farm

Reducing reliance on costly sexed semen helped farmers lower their breeding expenses.

96%

positive farmer feedback

Farmers responded positively to clearer heat insights and more actionable mating information, across 285 ratings.

<1%

false positives

Reliable heat detection helped farmers trust the information behind their mating decisions.

IN THE FARMERS' WORDS

“Hey Sang, just a quick message with some feedback now that we have done a couple of weeks of mating - long story short the improvements are awesome!! So much better than last year and making the daily AI routine easy!!! Thank you!”

Ryan

Kanadale Farm · 2026

“Really liking the changes to the mating list. Being able to see how confident we should be in each heat has been super helpful, especially when deciding which cows to use sexed semen on.”

Will

Wilco · 2026