<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Data Kisaan - Medium]]></title>
        <description><![CDATA[The external world’s window to Carnot DataLabs, which aims to modernize the agricultural ecosystem in India by using data-driven insights to impact tractor sales, marketing and financing. - Medium]]></description>
        <link>https://medium.com/data-kisaan?source=rss----17bb1fca6510---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Data Kisaan - Medium</title>
            <link>https://medium.com/data-kisaan?source=rss----17bb1fca6510---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 10:00:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/data-kisaan" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Wise and Otherwise — A journey to understand Indian farmers]]></title>
            <link>https://medium.com/data-kisaan/wise-and-otherwise-a-journey-to-understand-indian-farmers-8131622b03a1?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/8131622b03a1</guid>
            <category><![CDATA[farmers]]></category>
            <category><![CDATA[user-research]]></category>
            <category><![CDATA[indian-farmer]]></category>
            <category><![CDATA[agritech]]></category>
            <category><![CDATA[user-personas]]></category>
            <dc:creator><![CDATA[Hiten Rajgor]]></dc:creator>
            <pubDate>Fri, 30 Jul 2021 12:00:46 GMT</pubDate>
            <atom:updated>2021-07-31T08:04:09.171Z</atom:updated>
            <content:encoded><![CDATA[<h3>Wise and Otherwise — A journey to understand Indian farmers</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3wLxazDP3pchmjv1mwBXww.jpeg" /><figcaption>Picture by <a href="https://www.pexels.com/@chiecharon">Archie Binamira</a></figcaption></figure><p>W e are in the 21st century and guess which sector is the highest employment sector in India? I.T?… Hold on my friend!</p><p>The oldest industry in the history of mankind still dominates by employing 50% of the workforce of the country and contributing 18% to the national GDP. Yes, you guessed it right — The agricultural sector.</p><p>Now imagine designing a product for 50% of the workforce of the country, Exciting? Stressed? Having a rollercoaster of feelings? Aha, it is indeed a rollercoaster ride! Tight your seatbelts. We are getting started.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/360/1*TUcdvkv4hCJwPIPp-YbD-Q.gif" /></figure><h3>The Inception</h3><p>We were discussing the pricing for our new device and my teammate said “Farmers can easily afford a device for Rs. 8,000”. I replied NO WAY! They hardly save anything and it’s too costly for them.</p><p>There was silence and I realized DAMN! I am looking at India from just one perspective.</p><p>I am from <a href="https://www.google.com/search?q=Kutch&amp;rlz=1C1CHBF_enIN922IN922&amp;ei=Th4BYenkJvPgz7sP8v644AU&amp;oq=Kutch&amp;gs_lcp=Cgdnd3Mtd2l6EAMyCgguELEDEEMQkwIyBAgAEEMyBQgAELEDMgUILhCxAzIICAAQsQMQgwEyCAgAELEDEIMBMggILhCxAxCDATICCC4yAgguMgIIADoHCC4QQxCTAjoECC4QQzoICC4QxwEQowJKBAhBGABQ2FlYrmBgnGJoAHACeACAAbQDiAG4B5IBBzAuMy40LTGYAQCgAQGqAQdnd3Mtd2l6wAEB&amp;sclient=gws-wiz&amp;ved=0ahUKEwjpw82FtYXyAhVz8HMBHXI_DlwQ4dUDCA8&amp;uact=5">Kutch, Gujarat</a>. People in Kutch struggles with water hence farming is quite a difficult profession while the teammate is from <a href="https://www.google.com/search?q=Saurashtra&amp;rlz=1C1CHBF_enIN922IN922&amp;oq=Saurashtra&amp;aqs=chrome..69i57j69i59.212j0j1&amp;sourceid=chrome&amp;ie=UTF-8">Saurashtra, Gujarat</a> and they have an abundance of water, and farming is a wealthy profession for them. Same state and 2 completely different perspectives!</p><p>The People at <a href="https://medium.com/u/8c842617c37d">Carnot Tech Team</a> are from all over India which brings in a lot of perspectives though the flip side is that everyone has a user persona in their mind.</p><p>Hence it was essential to create a user persona so that we all have a common denominator.</p><p>Wait! What is<strong> USER PERSONA?</strong></p><blockquote>A user persona is a user profile created to represent different types of users that use a company’s product or service.</blockquote><h3>The Planning</h3><p>There was no template about the Indian farmers on the internet and we being leaders in the Agri-tech industry decided to build one from scratch.</p><p>We drafted 80 interview questions. They were a combination of open and close-ended questions related to a farmer’s lifestyle and our products.</p><p>The end goal was to understand how much a farmer’s lifestyle is digitally integrated and which parts of the work are still traditionally done and the reasons behind those decisions.</p><p>We shortlisted 16 states and targeted 5 users from each state summing up to 80 users from all over India.</p><h3>The Execution</h3><p>These interviews were remote and we were ambitious! Farmers started hanging up calls while we were asking questions; they were not being rude. Who got time to answer 80 questions to a stranger?</p><p>Lessons learned! We reduce the number of questions to 25–30 and created a framework via which farmers started trusting us and the conversations use to last for hours and <em>we</em> had to ask them to hang up.</p><p>As these interviews were unplanned for farmers, timings use to be random starting in the morning 6 at some time to 9 at the night.</p><p>It was feeling like we were traveling all over India as with every call we were hearing a new accent or language. Though the language also became a barrier in the south part of the country and we had to drop 4 states (20 users).</p><h3>The Learnings</h3><p>The following data is a combination of qualitative data which we collected doing user interviews and quantitative data from our IoT devices and mobile apps.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w1Pxacttb8yIZMW3OQKGhA.jpeg" /><figcaption><strong>Note: </strong>YouTube should have topped here though we are considering the apps that are used and installed by farmers and not considering YouTube as it is a default Android app.</figcaption></figure><p>If you see the above list most of these or at least a few of them are used by every smartphone owner. Internet is blurring the line between cities &amp; villages. Indian apps like <a href="https://www.google.com/search?q=Sharechat&amp;rlz=1C1CHBF_enIN922IN922&amp;oq=Sharechat&amp;aqs=chrome..69i57.213j0j7&amp;sourceid=chrome&amp;ie=UTF-8">Sharechat</a> are gaining a strong share in the interiors of India.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vcwjLh0TnmiqEpGgTQw_pQ.jpeg" /></figure><p>Almost every smartphone user in India is using the internet. Credit goes to <a href="https://www.google.com/search?q=Jio&amp;rlz=1C1CHBF_enIN922IN922&amp;oq=Jio&amp;aqs=chrome..69i57j46i199i291i433i512j46i199i433i465i512j0i433i512l2j0i131i433i512j0i433i457i512j0i402l2j0i131i433i512.246j0j7&amp;sourceid=chrome&amp;ie=UTF-8">Jio</a></p><p>🛒 Online shopping is gaining popularity in the interiors of India as often the delivery is free and these apps offer lucrative discount deals. Cash on delivery is the preferred mode of payment as they still don’t fully trust these platforms.</p><p>💰 Farmers often look for loans from banks and government bodies. Schemes like <a href="https://www.google.com/search?q=KCC&amp;rlz=1C1CHBF_enIN922IN922&amp;oq=KCC&amp;aqs=chrome.0.69i59j46i131i433i512j0i131i433i512j0i433i512j0i131i433i512j69i60l3.2508j0j7&amp;sourceid=chrome&amp;ie=UTF-8">KCC</a> had gained a lot of popularity in the Northern states of India.</p><p>Noticeably Farmers are not comfortable with documentation and the long approval process of banks &amp; government bodies hence unorganized lenders come into the picture.</p><p>💸 In daily life the farming community prefers doing transactions in cash hence apps like Phone Pe, GPay is used for minimal purposes like payment of Electricity bill and sometimes cell phone network recharges.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oQD2DRzuODM9LL9uNYCvuQ.jpeg" /></figure><p>🏛 Government keeps launching subsidies, relief funds, and other schemes for farmers but these schemes hardly reach farmers. Either farmers don’t have the necessary documents or by the time they are aware of these schemes are already expired.</p><p>CORRUPTION ALERT!!! Yes, there are a few agents who will make sure the farmer get these schemes but they will charge a commission in return 🤑</p><p>🎮 We thought farmers will be killing their time playing games like <a href="https://www.google.com/search?q=ludo&amp;rlz=1C1CHBF_enIN922IN922&amp;oq=ludo&amp;aqs=chrome..69i57j46i433i512j69i59j0i433i512j46i433i512j0i433i512j46i433i512j0i433i512j46i433i512j0i433i512.956j0j7&amp;sourceid=chrome&amp;ie=UTF-8">Ludo</a> though to our surprise they don’t. In fact, they hardly play any game. They watch movies or YouTube videos to kill their time.</p><p>Though the younger generation of farmers is active and they do play mobile games.</p><p>🐛 Local pest shops are the go-to place for any crop-related issue. Some of the farmers do look up solutions on the internet but the results are often non-satisfactory and the pesticide shown as the solution is not available locally.</p><p>They share a connection with the local Pest shopkeepers as they have been doing business with them for years and in few cases, they even get pesticides and fertilizers on credit which makes this bond stronger.</p><p>🎁<strong>BONUS!</strong></p><ul><li>We google solutions but farmers YouTube solutions and they prefer voice over typing.</li><li>YouTube ads will get you more downloads than Facebook. As farmers love video content over other types of content.</li><li>Farmers have various WhatsApp groups from weather alerts to farming tips and circulating your app in these groups will also get you downloads.</li></ul><p>Here is a glimpse of our user persona</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jcakepO68G0yXGMh7qxDEw.jpeg" /></figure><p>We were amazed to see the results from these interviews and learned a lot about the Indian farmers. Once we shared these personas within our team it was loved by all the teams and they all started thinking about how these can help to make better product decisions.</p><p>Of course, we are going to iterate these user personas further as we learn from our users.</p><p>Got any questions? Drop me a mail at hiten.rajgor@carnot.co.in and I will be happy to discuss.</p><p>You just witnessed how data helps in creating empathy. Do you want to see how data can change lives and help businesses to boom? Subscribe to the biweekly newsletter of <a href="https://medium.com/data-kisaan">Data Kisaan</a> and you will be amazed by data magic</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8131622b03a1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/wise-and-otherwise-a-journey-to-understand-indian-farmers-8131622b03a1">Wise and Otherwise — A journey to understand Indian farmers</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[To Harvest or not to Harvest]]></title>
            <link>https://medium.com/data-kisaan/to-harvest-or-not-to-harvest-1b563690480c?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/1b563690480c</guid>
            <category><![CDATA[analysis]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[ndvi]]></category>
            <category><![CDATA[harvest]]></category>
            <category><![CDATA[wheat]]></category>
            <dc:creator><![CDATA[Carnot Tech Team]]></dc:creator>
            <pubDate>Fri, 16 Jul 2021 12:11:59 GMT</pubDate>
            <atom:updated>2021-07-16T12:11:59.812Z</atom:updated>
            <content:encoded><![CDATA[<p>NDVI correlation with wheat harvest</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FZR0LdIEqPZ5kK5BFB6Onw.jpeg" /></figure><p>In this article, we will be going micro. We will zoom in to a single farm patch, and see what we can gather from it, by analyzing the satellite image data of the patch over a period of time.</p><p>This patch of interest lies in Dildarnagar, Uttar Pradesh.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/859/1*p0KLnIJPcIFBD8ZjbfUHbg.png" /></figure><p>Now, we know the following about the patch:</p><ol><li>Location (coordinates)</li><li>Area (2.25 acres)</li><li>Crop (Wheat)</li><li>Harvest date (16th April 2021)</li></ol><p>We know this because one of the wheat harvesters in our fleet worked on this patch on 16th April 2021, when wheat harvesting was going on in full swing across Uttar Pradesh.</p><p>While building the satellite image analytics capabilities within Carnot, this appeared to be a suitable patch to get started with. So our team evaluated the NDVI (Normalized Difference Vegetation Index) values for this patch from January till May 2021. We took the SentinelSat images (10m resolution) for this region (for all the dates in this period for which they were available).</p><p>In step 1, we simply visualized the output.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M3oADCvIzeiNedSvS8ev5Q.png" /></figure><p>What you see above is a visual representation of how the NDVI changes from 16th January 2021 to 5th May 2021. If we plot the mean NDVI plot for these dates, this is what we get:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yyFMNVHp7IgLEr_VbFnOvg.png" /></figure><p>The plot above correlates well visually with the image (the greener the image, the higher the NDVI). We know that harvesting happened on 16th April 2021. NDVI was well below its peak and was decaying further (although slowly) around that time. Was this the right time for harvest? Or should harvest have happened earlier, when NDVI was the highest? We looked up literature.</p><p>We found this variation of indices (including NDVI), in <a href="https://www.researchgate.net/publication/275956960_Temporal_MODIS_Data_for_Identification_of_Wheat_Crop_using_Noise_clustering_Soft_Classification_Approach">this publication</a>. It is reproduced below with proper citations:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/1*1emaffqn3FT208jg-tlDhA.png" /><figcaption>Figure 4. Variation of indices values for NDVI, SAVI and EVI for different phenological stages of wheat. Temporal MODIS Data for Identification of Wheat Crop using Noise clustering Soft Classification Approach — Scientific Figure on ResearchGate. Available from: <a href="https://www.researchgate.net/figure/Variation-of-indices-values-for-NDVI-SAVI-and-EVI-for-different-phenological-stages-of_fig2_275956960">https://www.researchgate.net/figure/Variation-of-indices-values-for-NDVI-SAVI-and-EVI-for-different-phenological-stages-of_fig2_275956960</a></figcaption></figure><p>As can be seen, the decline in NDVI from the peak corresponds to a period of maturity, and harvesting happens toward the end of the decline when the NDVI value is quite close to the NDVI at the time of sowing.</p><p>So looking at the above, it seems like the harvesting did happen alright. It could have been a bit sooner, but it definitely wasn’t premature harvesting.</p><p>Enjoyed this article? Then be on the lookout for another two weeks later. Till then, for any data-related discussions, feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b563690480c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/to-harvest-or-not-to-harvest-1b563690480c">To Harvest or not to Harvest</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Perceiving the amount of data we deal with at Carnot]]></title>
            <link>https://medium.com/data-kisaan/perceiving-the-amount-of-data-we-deal-with-at-carnot-5717a4cfafe7?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/5717a4cfafe7</guid>
            <category><![CDATA[carnot]]></category>
            <category><![CDATA[volume]]></category>
            <category><![CDATA[amount]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[perception]]></category>
            <dc:creator><![CDATA[Carnot Tech Team]]></dc:creator>
            <pubDate>Fri, 02 Jul 2021 11:49:41 GMT</pubDate>
            <atom:updated>2021-07-02T11:49:41.000Z</atom:updated>
            <content:encoded><![CDATA[<p>In this edition of Data Kisaan, we thought of taking a step back from the routine of deriving insights from data and thought of getting a view of the amount of data we are dealing with. Without much ado, let’s get started.</p><h3>Acres</h3><p>So far, we have recorded 950,639 acres on our platform. What does it correspond to? It is more than 6 times the size of Mumbai.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*qay1AtTH50YpYjoQ.png" /></figure><h3>Kilometers</h3><p>The tractors and harvesters fitted with our telematics devices have traveled over 21,787,273 km so far. This is equivalent to <strong>26 round trips to the Moon</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*IyTFe9hCuN_1vF3P.png" /></figure><h3>Engine Hours</h3><p>The machines in our fleet have together clocked 2,976,623 hours of data. If you go back in time by that much, you’ll perhaps be recording Halley’s comet for the first time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*RZarVlm-L2V6BIUo.png" /></figure><h3>Number of Farmers in our Data Lake</h3><p>If we look at the combined number of people who have been onboarded on various apps offered by us, or have installed our telematics devices, the number is 788,116. This is higher than the population of Bhutan.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*0WIDW9Fw4h4-DsoR.png" /></figure><h3>Device Coverage</h3><p>If you plot the distribution of devices, across India, this is what the map will look like.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/802/0*OlT5my_dYeBuxIM2.jpeg" /></figure><p>How many devices have generated such enormous data? Just over 7,500. One can only imagine the amount of data that, say, a million devices will produce!</p><p>Enjoyed this article? Then be on the lookout for another two weeks later. Till then, for any data-related discussions, feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5717a4cfafe7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/perceiving-the-amount-of-data-we-deal-with-at-carnot-5717a4cfafe7">Perceiving the amount of data we deal with at Carnot</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Decoding Simha Profiler]]></title>
            <link>https://medium.com/data-kisaan/decoding-simha-profiler-abe1d8ca23d1?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/abe1d8ca23d1</guid>
            <category><![CDATA[tractor]]></category>
            <category><![CDATA[lead-management-software]]></category>
            <category><![CDATA[lead-generation]]></category>
            <category><![CDATA[tractor-price-in-india]]></category>
            <category><![CDATA[dealership]]></category>
            <dc:creator><![CDATA[Hiten Rajgor]]></dc:creator>
            <pubDate>Fri, 18 Jun 2021 11:33:11 GMT</pubDate>
            <atom:updated>2021-06-18T14:55:40.965Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S8VS3D_xXZfAk8FW2CqK7g.jpeg" /></figure><p>Selling a tractor is difficult! A salesman needs to understand the farmer&#39;s requirements, build trust, and craft a personalized deal that suits the farmer. How can a salesman craft a personalized deal if he barely knows the farmer?</p><p>A typical hack these dealerships used was to invite a 100s of farmers for chai &amp; samosa and explain to them the tractor features and if they got lucky a farmer might buy the tractor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eQIqjLImnW-1hYuJSnKiWw.png" /></figure><p>The above process is very long and definitely outdated. What if I tell you can get all the necessary farmer details with just one click? Now it sounds magical right? Before I jump to this magical product let me share the short journey of data in agriculture.</p><p>We receive 300+ GB of data every month from the 1000s of Simha kit! Our data scientists picked up their coffee mugs and started spending days &amp; nights with this data.</p><p>The data team got back with interesting insights like</p><ul><li>How the movement shifts based on season compare to off-season</li><li>Which tractor can get the job done faster and efficiently compared to other tractors</li><li>Which features &amp; keywords farmers actually see in an advertisement</li><li>Shop locations that even google will fail to find</li><li>Marketing hotspots for offline advertisements that will get maximum eyeballs</li></ul><p>and the list goes on. All these insights were hitting the mark but not the bull’s eye. We zoomed out and started looking for bigger patterns and that’s how <a href="https://medium.com/data-kisaan/what-will-future-dealerships-possibly-look-like-403ebe026cd4">Simha Profiler</a> was born.</p><p>Once we saw the bigger picture we started building Simha profiler. We had enough confidence in our product that we can simply launch it and get things moving but that’s not how products are built at <a href="https://www.carnot.co.in/">Carnot</a>.</p><p>In few weeks, Simha profiler was ready — a web-based product backed by a smart algorithm that analyzes our big data and predicts when a farmer will be buying a tractor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J6zfVV-ycpaLh-h3Y4GMbQ.png" /></figure><p>Sounds like we got the job done right?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*QT2lVPpaC0Q4s64hu5hgMg.jpeg" /></figure><p>The real test for the Simha profiler begun when we put it out to test with actual dealers in Uttar Pradesh and Maharashtra.</p><p>The traditional behavior of the dealers was to use an excel sheet and hence they demanded a feature via which they can download the excel sheets and probably use the leads in their traditional way but that would have just depreciated Simha profiler. Hence we dug down deeper and learnings were</p><ul><li>They wanted flexibility like an excel sheet for searching and sorting leads</li><li>They wanted a tab where they can move unwanted leads and share them with relevant dealers</li><li>Our leads were not 100% accurate</li><li>The Skytag feature was very advance and a dealer was not able to use it.</li></ul><p>Excel sheets are known for their flexibility and sorting capabilities. We failed to realize this and our initial version with the simple search and basic filters was not liked by our users. We re-engineered our search with advanced search features and added sorting &amp; filtering options. These two features were enough to make our users forget about the excel sheets.</p><p>Unwanted leads were based on the location or because of our failure at the algorithm level. We learned our lessons and improved our algorithm and added a few filters after which we improved our lead quality to near 95%</p><p>We overthought the Skytag feature. Based on user feedback we replaced SkyTag with district-level insights rather than device-level, to give the dealers a holistic view.</p><p>We added two new features</p><ol><li>Sharing heavy customer stats with dealers so that the dealer can build a relationship with such customers</li><li>A cold lead will be lost in the 100s of other leads hence we added a notifications feature to remind dealers of upcoming due dates.</li></ol><p>Round 2 started. The updated version of the Simha profiler was launched and the feedback was positive this time. Dealers loved it as the results were quick and efforts were minimum.</p><p>Simha profiler is now launched and as you read this article dealers had already started doing tractor sales with just one click and soon this wave of revolution will be hitting all the dealers in the country.</p><p>Want to hear more on how Data is changing the agriculture industry? <a href="https://medium.com/data-kisaan">Follow Data Kisaan</a> and you will be amazed by these stories. Till then, for any data related discussions, feel free to drop us a line at datalabs@carnot.co.in</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=abe1d8ca23d1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/decoding-simha-profiler-abe1d8ca23d1">Decoding Simha Profiler</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Impact of 2nd wave of Covid-19 on Indian Agriculture — Part 2]]></title>
            <link>https://medium.com/data-kisaan/impact-of-2nd-wave-of-covid-19-on-indian-agriculture-part-2-94b2dea323f7?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/94b2dea323f7</guid>
            <category><![CDATA[agriculture]]></category>
            <category><![CDATA[covid]]></category>
            <category><![CDATA[lockdown]]></category>
            <category><![CDATA[impact]]></category>
            <category><![CDATA[2021]]></category>
            <dc:creator><![CDATA[Yash Sanghvi]]></dc:creator>
            <pubDate>Fri, 04 Jun 2021 11:32:20 GMT</pubDate>
            <atom:updated>2021-06-04T11:32:17.876Z</atom:updated>
            <content:encoded><![CDATA[<h3>Impact of 2nd wave of Covid-19 on Indian Agriculture — Part 2</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xNFEfMb34pTiZ6ZDZE5HCw.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@synkevych?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Roman Synkevych</a> on <a href="https://unsplash.com/s/photos/agriculture-in-lockdown?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>In Part 1 of this analysis, we primarily considered tractor activity (acres farmed, distance traveled), and it indicated that the tractors worked as much in 2021 as in 2020. We ended <a href="https://medium.com/data-kisaan/impact-of-2nd-wave-of-covid-19-on-indian-agriculture-part-1-219ccfc14ab8">Part 1</a> of this story on a phone call we had with a farmer, which opened our eyes to several other aspects that we needed to consider to gauge the impact of the second wave of Covid-19.</p><p>What are these aspects? Perhaps a summary of the phone call, given below, will shed some light:</p><ol><li>His tractor and harvester worked nearly as much in 2021 as in 2020 (something we also validated broadly in the first part of this post)</li><li>Revenue was hit because of the cash crunch at the end of many farmers. There were many customers (people who rented his tractor/ harvester) who either delayed payments or did not pay altogether</li><li>The scars left behind by 2020 affected the overall sentiment. Many people in his village were not growing the perishable 3rd crop between Kharif and Rabi in 2021, because their produce went unsold in 2020 (due to lack of demand), leading to losses.</li></ol><p>Now, apart from the first point, there were two other major areas to be investigated — revenue, and sentiment. Before we looked at data to investigate along these lines, we decided to gather more on-ground reports (so as not to get biased by one viewpoint). So our field and support team reached out to several other customers across India and asked them about the impact of the second wave of Covid-19 on their business.</p><blockquote>Most customers echoed the first two points. Tractors worked as much as in 2020, but revenue was hit. On the sentiment front though, the response was mixed.</blockquote><p>Some farmers agreed that the sentiment took a hit, but the sentiment of most was unwavering. Most said that they were not making any changes to the crop cycles, and almost all of them were quite optimistic about the future.</p><p>Now it was time to put this on-ground information to some test.</p><h3>Evaluating Revenue Impact</h3><p>Somewhere in March 2020, we introduced an accounting feature (meant to digitize the traditional bahi-khata) in the Simha Kit app. The customers could enter their day-to-day transactions in this part of the app, and keep track of their revenue and costs.</p><p>This was the perfect data source for us for this analysis. We looked at all the CREDIT transactions entered by the users in <strong>Apr-May 2020</strong> and <strong>Apr-May 2021</strong>. We evaluated 3 metrics:</p><ol><li>Average ticket size (size of an average CREDIT transaction)</li><li>Average number of transactions per user</li></ol><p>Now, in March 2020, we just had upwards of 500 Simha Kit users, and this number was over 3500 in May 2021. And, not all users use the accounting feature. Then how were we supposed to figure out the impact on the revenue?</p><p>One would say only consider the users who have entered at least one transaction on the platform in the timeframe of interest, for the purpose of averaging. This would indicate that the user is using the accounting feature. This is a fair point. The only word of caution here is that several users may not be entering CREDIT transactions here and may be using this feature like an expense manager. Therefore, the ‘average number of CREDIT transactions per user’ metric will not represent the actual truth. Luckily, it is only the trend that we are interested in. We will have to assume that the percentage users using this feature like an expense manager is constant between 2020 and 2021.</p><p>So, here are the results. Let’s first look at the average ticket size.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2kBp0xwFKZ2cHtccG3LVrQ.png" /></figure><p>As you can see, the average transaction amount in 2021 is about 20% lower than in 2020. This itself is a big reduction. Now, let’s look at the average number of transactions per user.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AcLa6SXvH08e34WZBPo16A.png" /></figure><p>As you can see, the average number of CREDIT transactions per user is about 24% lower in 2021 as compared to 2020. Thus, users received money lesser number of times, and each time, the amount was also lower than last year.</p><blockquote>If you multiply the two metrics above, there is a close to 40% reduction in the avg. revenue per tractor owner this year, which matches well with what the on-ground interviews reported.</blockquote><p>Hence, so far, the conclusion is that the tractor activity remained unaffected in 2021 as compared to 2020, but the cash flow took a hit. Several customers have either delayed payments or have declined to pay altogether. This could be because, in 2020, people were entering a lockdown. In 2021, they had already spent 1 year with Covid, and the second wave hit. So savings had presumably dried up, and a lot will depend on the sale of the agricultural produce going forward.</p><p>One can only hope that the wheels start churning faster now, and the delayed revenue soon finds its way.</p><h3>Evaluating Sentiment Impact</h3><p>As mentioned above, the first farmer we spoke to said that many farmers in his village were not growing the 3rd crop between Kharif and Rabi this year, because of the unsold wasted harvest in 2020. While many other farmers later refuted this sentiment, we still ventured out to analyze this claim.</p><p>We now turned to Krishi Diary data. Our Krishi Diary app can be used as a digital ledger by any farmer, irrespective of whether or not they have the Simha Kit installed on their tractors. We considered March to May as the period of interest (as most Kharif harvesting happens in March) and checked the transactions that were registered on the app.</p><p>Since we had a sufficiently large base of users, we restricted ourselves to only those users who had installed the app prior to March 2020, and were active on the app in the March to April period in both 2020 and 2021.</p><p>We compared the area under cultivation for some common Zaid season (March to June) crops, and also perishable fruits and vegetables, which might have been affected because of loss of sentiment. The results are shown below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K01I_drH4TQSTRydtOQuTQ.png" /></figure><p>As you can see, the area under cultivation for most Zaid crops and perishable crops either remained constant or grew in 2021. There are some exceptions, like Soyabean, Maize, Lemon.</p><blockquote>The general trend, however, largely indicates that the sentiment has NOT taken a hit.</blockquote><p>We should add a disclaimer here that the Krishi Diary app users may be more tech-savvy, more enterprising, and more risk-taking, and therefore the general trend may indicate a lack of loss of sentiment. There may be truth to what we inferred from the first farmer we called. However, considering that most other farmers agreed that their sentiment hasn’t been affected much, if there is indeed any sentiment loss, it is most likely restricted to isolated pockets.</p><h3>Digging Deeper</h3><p>While the general trend was established, we wanted to dig further into the claims that some farmers were not going to grow the 3rd crop this year. Therefore, we looked at the SentinelSat data for the region around the first farmer we called. We specifically looked at the NDVI (Normalized Difference Vegetation Index) values for a small radius around this farmer. NDVI is a proxy for the vegetation in a region, and a comparison of NDVI can help predict whether the produce was higher or lower in different timeframes.</p><p>This is the overall look of the region this farmer resides in is shown below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/591/1*jNAYJHBToVqhPOvYEGan_w.png" /></figure><p>The NDVI values for the region the framer resides in are given below (all calculated from 1st to 15th May):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aGp6kzmvDH1TF1RrbFGuDg.png" /></figure><p>Thus, while there was a slight reduction in the 2021 NDVI as compared to 2020, the difference was not huge, and the value was still higher compared to 2019. Atmospheric conditions may as well be responsible for this difference. This leads us to conclude that even in the broader region surrounding this farmer, the loss in sentiment may not have been very significant.</p><p>Satellite data-based NDVI is a data-analysis approach we have just started with. With time, as we become more proficient with this approach, we may be able to produce more pointed and useful insights. Watch this space!</p><h3>Closing Thoughts</h3><p>Looking at the analysis done so far, I guess the following sums up what we learned about the impact of the 2nd wave of Covid-19 on Indian agriculture:</p><blockquote>Agricultural work carried on as usual in 2021, just like in 2020. However, the cash flows, and therefore the revenue, took a hit. The effect of this disturbance on the revenue cycle may manifest itself in the times to come, and one can only hope that this is just a short-term de route from the steadily marching resilient growth story that Indian agriculture is. The overall sentiment, luckily, remains quite positive, as does the future outlook. One can only look forward to prosperous days ahead.</blockquote><p>With that, I’ll conclude this article. If you’ve made it here, thank you. Have any thoughts on this analysis? Then feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><p><em>Special thanks to the field and support team of Carnot for interacting with the customers and getting the ground truth data for our reference.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=94b2dea323f7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/impact-of-2nd-wave-of-covid-19-on-indian-agriculture-part-2-94b2dea323f7">Impact of 2nd wave of Covid-19 on Indian Agriculture — Part 2</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Impact of 2nd wave of Covid-19 on Indian Agriculture — Part 1]]></title>
            <link>https://medium.com/data-kisaan/impact-of-2nd-wave-of-covid-19-on-indian-agriculture-part-1-219ccfc14ab8?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/219ccfc14ab8</guid>
            <category><![CDATA[lockdown]]></category>
            <category><![CDATA[agriculture]]></category>
            <category><![CDATA[covid]]></category>
            <category><![CDATA[impact]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Yash Sanghvi]]></dc:creator>
            <pubDate>Fri, 21 May 2021 11:33:21 GMT</pubDate>
            <atom:updated>2021-06-04T11:47:51.108Z</atom:updated>
            <content:encoded><![CDATA[<h3>Impact of 2nd wave of Covid-19 on Indian Agriculture — Part 1</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AQ8Dt5-dWNT8kwr2VPWF8g.png" /></figure><blockquote>He looked at his beloved farm. The sparsely scattered shoots mirrored the lack of clarity in his mind. Should he take a risk and grow the 3rd crop as usual before the preparation for Kharif began? Or should he just let this season pass? He was encouraged by habit, but discouraged by the scars that 2020 had left behind. Just like his farm patch, a bit of clearing was all that his mind needed.</blockquote><p>In January this year, we came out with a post titled ‘<a href="https://medium.com/data-kisaan/agriculture-in-the-times-of-lockdown-37bf796441d2">Agriculture in the times of lockdown</a>’. It was an attempt to see if Simha data can suggest why agriculture grew at 3.4% in the quarter in which India’s overall GDP contracted by 23.9%. And the data did not disappoint. It showed how agriculture bounced back immediately after Junta Curfew, and went on unperturbed, unfazed by the nationwide lockdown.</p><p>As the terror of the second wave of Covid-19 spread across India, we were getting more and more requests to check if the effects on agriculture were any different this time. Perhaps the most defining characteristic of the second wave of Covid-19 was its suddenness, which took everyone by surprise. People, especially from Mahindra, wanted to know if there were any surprises to be expected in agriculture as well. And so, we began our analysis.</p><p>2021 was very different compared to 2020 when seen from the Carnot data perspective. While our telematics devices were installed across just about 500 tractors in Feb 2020, the number was 5000 in Feb 2021. We also had a rich database of users who had downloaded the <a href="https://play.google.com/store/apps/details?id=com.carnot.krishidiary&amp;hl=en_IN&amp;gl=US">Krishi Diary app</a>, and were entering the agricultural transactions regularly. Not only did we have data coming in from tractors, but also mighty harvesters, who easily worked an average of 25 acres a day.</p><h3>The First-Cut Visualization</h3><p>To begin our analysis, we performed a simple comparison of 2021 acres vs 2020 acres. Since the number of tractors was significantly different, we normalized the values (avg. acres per tractor). Here are the results:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FR1mnsPbi8-g&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DR1mnsPbi8-g&amp;image=http%3A%2F%2Fi.ytimg.com%2Fvi%2FR1mnsPbi8-g%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/11db33d2903ad59fc5440a7fa62ced3b/href">https://medium.com/media/11db33d2903ad59fc5440a7fa62ced3b/href</a></iframe><blockquote><strong>The first cut visualizations are often misleading!</strong></blockquote><p>If you look at the two graphs in the above video, it would appear that 2021 was hands down much better than 2020, for agriculture. But this data didn’t convince us enough. <strong>How can the numbers be so different</strong>, <strong>especially when the pandemic was more devastating in 2021 than in 2020?</strong></p><p>An explanation came in from the composition of our fleet. Based on the usage, we had identified rental entrepreneurs or REs (those who rent out tractors for agricultural and haulage work) to be the target audience for our telematics devices, and therefore, their proportion in our fleet had gradually increased. These REs would definitely clock in more acres and hours than people using the tractors just for their own farms. Therefore, a higher proportion of REs in the fleet would push the average up.</p><p>Also, the giant harvesters also played their part. The wheat harvesting happened from mid-March to mid-April across India, and each harvester in our fleet clocked an average of 300 acres in that duration. So of course, the averages were bound to be higher.</p><p>So with that, we were back to square one. How do we gauge the impact of the second wave then? There was only one way.</p><h3>Comparing apples with apples</h3><p>We performed the same exercise as the one above, but this time, <strong>we only restricted ourselves to the 500-odd devices that were already installed in 2020</strong>. These results seemed to be more palatable.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FrExbWH1qhKg%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DrExbWH1qhKg&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FrExbWH1qhKg%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/72556aefc6e96fa55606d81e743206c6/href">https://medium.com/media/72556aefc6e96fa55606d81e743206c6/href</a></iframe><blockquote>If we just go by the 2020 fleet, the agricultural activity was unaffected, just like last year. The tractors covered nearly as many acres on an average in 2021, as in 2020.</blockquote><p>The data showed the dips on Holi (10th March 2020, and 29th March 2021), Junta curfew (22nd March 2020), and showed a gradual increase as Rabi harvesting approached.</p><p>If we look at the haulage guys, the story is slightly different.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FXryqO2Rcrcs&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DXryqO2Rcrcs&amp;image=http%3A%2F%2Fi.ytimg.com%2Fvi%2FXryqO2Rcrcs%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/741319acdc53be274ebbc152bbac4a2a/href">https://medium.com/media/741319acdc53be274ebbc152bbac4a2a/href</a></iframe><blockquote>If you observe closely, the haulage business was very much affected by the imposition of lockdown last year and took several days to recover. Even in mid-May 2020, it wasn’t back to the pre-lockdown levels. In 2021 however, the story was different.</blockquote><p>There was no nationwide lockdown announced in 2021. Different states announced lockdowns at different times, with most of them announcing it in April. Therefore, while the haulage business was recovering from the lockdown shock in April last year, it was in a slight decline in the same period this year.</p><h3>The Phone Call</h3><p>The short decline in April notwithstanding, the overall numbers for the haulage business do look better in 2021, as compared to 2020, and the farming business seems unaffected.</p><p>So does that mean that tractor owners fared better in 2021 than in 2020? Perhaps. But does aggregate data speak uniformly for the agricultural ecosystem? We thought so until we got on a call with a farmer to understand if the data correlates with the ground reality. And we were surprised.</p><blockquote>What we realized was that there were several other aspects of agriculture that we needed to take into consideration.</blockquote><p>What were these aspects? Do they change the conclusion? If yes, then how? Head on to <a href="https://medium.com/data-kisaan/impact-of-2nd-wave-of-covid-19-on-indian-agriculture-part-2-94b2dea323f7?source=friends_link&amp;sk=e294670bc31aa417d3d06b4b0af5dda1">Part 2</a> to find out.</p><p>Enjoyed this article? Then be on the lookout for another two weeks later. Till then, for any data-related discussions, feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=219ccfc14ab8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/impact-of-2nd-wave-of-covid-19-on-indian-agriculture-part-1-219ccfc14ab8">Impact of 2nd wave of Covid-19 on Indian Agriculture — Part 1</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Decoding the Virtuous Crop Cycles in India]]></title>
            <link>https://medium.com/data-kisaan/decoding-the-virtuous-crop-cycles-in-india-ffcbb2e2c710?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/ffcbb2e2c710</guid>
            <category><![CDATA[mechanization]]></category>
            <category><![CDATA[crop-cycle]]></category>
            <category><![CDATA[agriculture]]></category>
            <category><![CDATA[calendar]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Anurag Sharma]]></dc:creator>
            <pubDate>Fri, 07 May 2021 13:32:23 GMT</pubDate>
            <atom:updated>2021-05-07T13:32:23.385Z</atom:updated>
            <content:encoded><![CDATA[<p>Story of combining telematics and ground truth data to predict the farm mechanization demand</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*l763P-Cvd3fVDeZTEwPTPw.jpeg" /><figcaption>Image by <a href="https://pixabay.com/users/rauschenberger-4614580/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3819469">RENE RAUSCHENBERGER</a> from <a href="https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3819469">Pixabay</a></figcaption></figure><p>In this story, we discuss how crop cycle data served as a string that connected our geographic, demographic, hotspot, and mechanization data, to produce several insights of vital importance for the agricultural ecosystem and related industries.</p><p>The annual cycle of agriculture consists of several activities performed throughout the year for each crop. These activities include land preparation, inter culture, sowing, harvesting, etc. Based on the activities performed by the farmer, there will be different mechanization requirements, and thus, different machine and implement requirements. If one can predict beforehand which activities are going to be performed in which region at what time, a lot of value can be provided to all players involved in the agricultural industry. This can be viewed as a demand planning and forecasting problem, where we analyze the crop cycle data and figure out the demand of different implements/ tractors/ harvesters/ transplanters throughout the year for different crops in different regions.</p><h4><strong>Crop Calendar for major Indian Crops</strong></h4><p>We had geographic and demographic data for the farmers on our platform. The task was to get the crop cycles for the major Indian crops, for different regions. We sourced this data from several trusted open databases. We went a step ahead, and even integrated the mechanization requirements at each step of the crop cycle. As of now, we considered only a few major Indian crops, namely, <strong>Rice</strong>, <strong>Wheat</strong>, <strong>Rabi</strong>, <strong>Maize</strong>, and <strong>Chilli</strong>. The crop cycle for a particular crop spanned 6–7 months. Also, we must understand that these cycles are different in different states of India. For example, a task like land preparation for Kharif potato begins in May in Andhra Pradesh meanwhile the land preparation in the state of Bihar would start around August or September.</p><p>We organized the crop cycle + mechanization data in a structured way as shown in the figure below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1CwbTEppfsjpuCJZQHqasQ.png" /><figcaption>Sample Crop Cycle</figcaption></figure><h4><strong>Putting it all together — Crop Cycle, Geographic and Demographic data</strong></h4><p>The compiled crop cycles were then combined with the geographic and demographic data that we had for a farmer, and voila! The entire calendar of activities to be performed by the farmer and his implement requirements was available to us.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-0M2eURYvZdFIr5Wj9I4qw.png" /></figure><p>This can be easily extended to all the farmers in our database, and then farmers with similar calendars can be clubbed together.</p><h4><strong>Telematics Based Location Intelligence</strong></h4><p>Even after knowing the clusters of farmers with similar crop calendars, we needed to find certain location hotspots that would aid in campaign planning. Using the telematics data available to us from the various tractors, we had already identified important location hotspots in each region, where the farmers visited frequently or spent a lot of time. A glimpse of our detailed hotspot analysis can be found in this <a href="https://medium.com/data-kisaan/tour-de-meerut-via-simha-data-9c1260c43538">post</a> exploring Meerut from a data lens. We analyzed geographic hotspot data points from <strong>West Bengal</strong>, <strong>Madhya Pradesh</strong>, and <strong>Maharashtra </strong>for the purpose of the pilot. Below is an example of visualization for a village in West Bengal: <strong>Bhatar</strong>. Bhatar is an important rice belt in West Bengal. Some of the relevant location spots found in these regions were fertilizer suppliers, farm areas, highway meeting points, and fish farming spots.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/993/0*YF2xYa8HWFHOwLHd" /></figure><p>Thus, this post explains how the combination of crop cycle with the several types of data that we already had (geographic, demographic, etc.), created valuable insights for nearly all players in the agricultural ecosystem.</p><p>Enjoyed this article? Then be on the lookout for another two weeks later. Till then, for any data-related discussions, feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ffcbb2e2c710" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/decoding-the-virtuous-crop-cycles-in-india-ffcbb2e2c710">Decoding the Virtuous Crop Cycles in India</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Sandwich Story: Telematics & Data ]]></title>
            <link>https://medium.com/data-kisaan/the-sandwich-story-telematics-data-3d2eb8e10472?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/3d2eb8e10472</guid>
            <category><![CDATA[india]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[insights]]></category>
            <category><![CDATA[agriculture]]></category>
            <category><![CDATA[telematics]]></category>
            <dc:creator><![CDATA[Pushkar Limaye]]></dc:creator>
            <pubDate>Fri, 09 Apr 2021 14:10:56 GMT</pubDate>
            <atom:updated>2021-04-09T14:17:43.582Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*g8bH1OnsjEjx8IJC" /><figcaption>Photo by <a href="https://unsplash.com/@saracervera?utm_source=medium&amp;utm_medium=referral">Sara Cervera</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><blockquote>“White bread or brown bread? — hurry up young chap! make your choice” — The sandwich shop guy suddenly interrupted me and my co-founder having our lunch time catch up. Something, that has essentially become the ‘strategy time-out’ for both of us. “Use anything, but better make sure sandwich is tasty” I replied impatiently , been super hungry having delayed lunch by couple of hours already.</blockquote><p>The sandwich indeed was damn tasty. The funniest part is, till the end I didn’t bother to know which bread he used. All that mattered to me was the taste and quality of the sandwich he delivered!</p><h3>Telematics: Bread or Sandwich?</h3><p>Just for a quick recap, I am CTO and co-founder of Carnot Technologies, an IoT startup based out of Mumbai. Our flagship product is <a href="https://www.simhakit.com/">Simha kit,</a> which is a Telematics platform. We help tractor owners in India make 20–30% additional income.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FKsY9d71ewwM%3Fstart%3D19%26feature%3Doembed%26start%3D19&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DKsY9d71ewwM&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FKsY9d71ewwM%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/71a02a81b45b2a776dc4b107808a365a/href">https://medium.com/media/71a02a81b45b2a776dc4b107808a365a/href</a></iframe><p>Long story short, this is the fifth full-stack telematics hardware + software that I have built, including the very first I had built with IIT Bombay Racing, which eventually led to my journey with <a href="https://medium.com/tech-carnot/c72f11faf9aa?source=friends_link&amp;sk=00c153681f6709fc8197fdfcc87a264e">Carnot</a>. In this hard journey of hardware, I had the unique opportunity to interact with most of the key players in the telematics space in India. Primarily these are hardware vendors, fleet companies, insurance/finance companies, and the OEMs themselves. The telematics scene in India is still relatively nascent, and no single mature OEM product comes to my mind. As a matter of fact, almost every OEM we have spoken to and worked closely with has shown interest in building their in-house telematics platform. Despite all of the growth stories, digitization, etc., why is connected cars / connected vehicles still a fancy concept for India?</p><p>Confused? If you are setting up a sandwich shop, your focus would be on making delicious sandwiches. You won’t spend months brainstorming which bread to use, which ketchup to use, how to organize the vegetables, and so on. You’d start making sandwiches as soon as possible, and keep improvising based on customer feedback. In less than a month, you’d understand what works and what doesn’t.</p><p>When it comes to the telematics industry, everyone is doing exactly the opposite. They are emphasizing the bread — device, hardware, 2G-3G-4G, and all possible hardware jargon under the hood! People are forgetting that what is required is a tasty sandwich, which bread is used doesn’t matter. Rapid prototyping and feedback-based improvisation are very much missing from the way the telematics industry operates. The very relevant axiom, that continuous improvement is better than delayed perfection, seems to be missing from their operational behavior. And what are the implications? The industry is missing out on the holy grail, “Data Insights”.</p><p>Data from customers can not only help the sandwich-maker create better sandwiches but can also tell him about the eating patterns of the population, what time slots are popular, how price sensitive are the customers, how many are willing to pay a premium for better quality and so on. The sandwich maker gets data that can help even a non-sandwich food stall in that location. Similarly, telematics devices’ data can not only help the industry make better telematics devices but also provide truck-loads (in our case tractor-loads ;) ) of domain insights, which have the potential to disrupt whatever industry the device operates in.</p><p>Solution? Telematics is currently being viewed as an additional service for customers. The industry needs to realize that it helps the OEMs just as much. It is in the interest of OEMs to join the digital revolution, and as soon as possible. The channel partners in the telematics industry need to open up their perspective, to start thinking beyond just hardware problems. They need to imagine a world where telematics data makes an impact, not on PPTs but on real business metrics. A classic example is what my colleague Yash talks about <a href="https://medium.com/data-kisaan/what-will-future-dealerships-possibly-look-like-403ebe026cd4">in this blog</a>.</p><p>Connected vehicles and telematics needs a (not so) secret sauce, of meaningful data products built on top, and no doubt the tasty sandwich can then be consumed byte by byte :)</p><p>We are trying to fix some <a href="https://medium.com/tech-carnot/c72f11faf9aa?source=friends_link&amp;sk=00c153681f6709fc8197fdfcc87a264e">broken benches </a>in the Indian agriculture ecosystem through technology, to improve farmers’ income. If you share the same passion <a href="https://angel.co/company/carnot-technologies/jobs/">join us in the pursuit,</a> or simply drop us a line on report@carnot.co.in</p><p>Also, follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium to get notified whenever a new article is released.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3d2eb8e10472" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/the-sandwich-story-telematics-data-3d2eb8e10472">The Sandwich Story: Telematics &amp; Data 🥪</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Detecting Suspicious Farm Patches Using Machine Learning]]></title>
            <link>https://medium.com/data-kisaan/detecting-suspicious-farm-patches-using-machine-learning-3698d93755f0?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/3698d93755f0</guid>
            <category><![CDATA[farm-patch-detection]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[telematics]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[agriculture]]></category>
            <dc:creator><![CDATA[Yash Sanghvi]]></dc:creator>
            <pubDate>Fri, 26 Mar 2021 13:31:38 GMT</pubDate>
            <atom:updated>2021-03-26T14:52:13.582Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OfwwLbYS2AutYYt3S3cCwg.jpeg" /></figure><p>In a <a href="https://medium.com/data-kisaan/tale-of-our-approach-to-a-real-world-problem-farm-patch-alignment-system-d876e4cc2cdb?source=friends_link&amp;sk=1414eb7b97c890cfda8233bf526f0318">previous blog post</a>, we discussed how, given a farm patch detected by our algorithms, we enhanced it further using its background satellite image. However, we realized that some of the original farm patches were themselves suspicious, which should have been eliminated at the generation step itself. They were in places where they shouldn’t be, like on a road strip, or on a piece of land on which construction was happening as per the satellite image.</p><p>So what was happening? Well, our patch detection algorithm uses a number of factors to determine a patch, like the proximity of raw GPS points, some learning-based clustering of speed, pattern identification using neural nets, and so on.</p><p>Now, say a tractor made a number of back and forth trips between a sugarcane farm and a sugar factory that are in close proximity to each other. Then the density of the points would lead to the creation of a small farm patch on the road strip.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pmRbMPdXzosprVqLDPNHeQ.jpeg" /><figcaption>Example of a patch detected on a road strip because of repeated back and forth travel on the same road</figcaption></figure><p>Similarly, say the tractor is parked somewhere for a long time or doing some stationary work, like threshing or digging. The GPS is prone to noise. Sometimes, the noise can become too good to not be heard ;) and over a period of time, the points accumulated in close proximity to the parking location can trick the algorithm into thinking that the tractor did a small farming maneuver there.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HoAifi-nH0TBnfxawI9J0Q.jpeg" /><figcaption>Example of a farm patch generated by a stationary tractor</figcaption></figure><p>What are the implications? Sub-par user experience is an obvious one. No one wants to see a patch where there is none. There are monetary implications as well for the tractor owners. For instance, the driver compensation is sometimes linked to the number of acres done. Now, these incorrect small patches can affect the overall area numbers shown on the Simha Kit app, and lead to higher cash outflow for the tractor owners, if they don’t check each patch individually.</p><p>So how did we tackle this problem? We knew that we somehow needed the server to learn how to detect and eliminate the suspicious patches at the source itself. This statement was itself hinting at Machine Learning. We began experimenting with ML. But unfortunately, all our initial trials gave discouraging results. What we realized was that the obvious features, like farm area, concavity of patch, number of raw GPS points, etc. were not helping differentiate between the proper and suspicious patches.</p><p>It was time to bring some method to the madness. So we took a step back and looked at some individual cases, where the trip was incorrect. If you look at the cases described above that led to a false patch, some prominent themes stand out:</p><ol><li>These false patches would take a much longer time to form than a proper patch of the same area. <strong>We hit gold here!</strong> We took the difference between the max and min timestamps of the points within the patch, and divided the patch area with it, and voila! <strong>60%</strong> of the false patches were eliminated in one go!</li><li>For patches formed near a stationary tractor, the average speed of the tractor in the duration of the patch would be nearly 0, except for some noise. Thus, average speed as a feature also helped us eliminate about 8–<strong>10%</strong> more patches.</li><li>These patches would have irregular and unconventional shapes. For instance, a patch formed on a road strip would be a much thinner rectangle, than a normal farm patch. We, therefore, decided to take the average boundary radius of the patch as another feature. This further helped eliminate about <strong>5–6%</strong> more false patches.</li></ol><p>We tried various ML models with these features and got really promising results with SVM and Random Forests.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/378/1*aF1aFIykD1SoFDMKJxyV7A.png" /><figcaption>F1 Score Comparisons for different ML Algorithms</figcaption></figure><p>In order to make sure that correct patches do not get eliminated by this ML model, we used probability-based classification and flagged a patch as suspicious only when the model gave a &gt;90 % probability of it being suspicious.</p><p>The results?</p><p>Percentage of false patches flagged as suspicious: <strong>76.8%</strong></p><p>Percentage of true patches flagged as suspicious: <strong>0.6%</strong></p><p>Percentage reduction in the number of patches that needed to be processed further: <strong>48%</strong></p><p>With this one extra step between the patch generation and patch enhancement, we achieved the elimination of &gt;70% bad experiences and recorded thousands of smiles from customers opening up their trips page to see their farming work. We also reduced the processing load on our servers by over 50%. The system really <em>felt</em> much more efficient.</p><p>With the introduction of the <a href="https://medium.com/data-kisaan/what-will-future-dealerships-possibly-look-like-403ebe026cd4">Profiler</a>, we took this a step further. We profiled users according to their usage patterns, and for users whose probability of using their tractors only for trolley business (no agriculture) was &gt;90%, we simply disabled patch detection. This was because any patch detected for such users was most likely going to be a false patch. This has been working so far and making sure that trolley customers never complain about false patches.</p><p>Enjoyed this article? Then be on the lookout for another two weeks later. Till then, for any data-related discussions, feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3698d93755f0" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/detecting-suspicious-farm-patches-using-machine-learning-3698d93755f0">Detecting Suspicious Farm Patches Using Machine Learning</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Carnot’s Big Number Duel]]></title>
            <link>https://medium.com/data-kisaan/carnots-big-number-duel-85c56f3177ed?source=rss----17bb1fca6510---4</link>
            <guid isPermaLink="false">https://medium.com/p/85c56f3177ed</guid>
            <category><![CDATA[carnot]]></category>
            <category><![CDATA[big-number]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[duel]]></category>
            <category><![CDATA[agritech]]></category>
            <dc:creator><![CDATA[Yash Sanghvi]]></dc:creator>
            <pubDate>Fri, 12 Mar 2021 14:50:31 GMT</pubDate>
            <atom:updated>2021-03-12T14:50:31.370Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iUmMeh_bb9i6OhjGwDRw3g.jpeg" /></figure><p>Recently, I came across this <a href="https://www.youtube.com/watch?v=X3l0fPHZja8">Numberphile video</a>, discussing the contest held at MIT in 2007 between two professors (one from MIT and one from Princeton). The objective was to find the biggest finite number. Of course, the number should have some usage. You can’t just say add 1 to the previous number. Have a look at the linked video. It is really interesting.</p><p>Inspired by that video, 3 of us at Carnot DataLabs thought of having Carnot’s big number duel. The rules were simple:</p><ol><li>The numbers have to be relevant to Carnot (I can’t just add Graham’s number to the discussion out of nowhere). More specifically, they should be relevant to Carnot’s operations. You can’t just report the number of atoms in a Simha Kit (although kudos to you if you can accurately calculate that :P).</li><li>If the number represents a measurement, the units have to be SI units (else I’d just write the distance traveled by tractors in nanometers or femtometers, to get bigger and bigger numbers)</li><li>Categories cannot be repeated (so if I reported the distance traveled by the Carnot fleet in one day, you can’t just report the distance traveled in one week to get a bigger number). You can have numbers of the same unit as long as the context is different.</li></ol><p>For the purpose of this post, let’s call the 3 contestants ‘A’, ‘B’, and ‘C’. Here’s how the contest panned out:</p><p>A: I’ll start off small but meaningful. We are Carnot, We are THE <strong>1</strong></p><p>B: This one is easy to beat :P. Simha App’s play store rating: <strong>4.7</strong></p><p>C: Our strength — <strong>72</strong> and growing strong!</p><p>A: Let’s ramp this up a bit. Number of customers who have our tractor telematics devices installed: ~<strong>3500</strong></p><p>B: This makes it easy for me. Number of devices: <strong>4102</strong></p><p>C: Number of Krishi Diary App downloads: <strong>1.15 x 10⁵ </strong>(1.15 lakh)</p><p>A: Hmm. Time to get data-oriented. Number of villages of India in our locations database: <strong>5.65 x 10⁵ </strong>(5.65 lakh)</p><p>B: I’ve another meeting coming up. Let’s get done with this. Number of pings we received yesterday from our telematics devices: <strong>2.7 x 10⁷</strong> (2.7 crores)</p><p>C: Not so soon, B! You know what’s bigger than the tech data? Financial data! Revenue generated by Carnot so far: <strong>3.5 x 10⁸ </strong>(35 crore rupees)</p><p>A: Engine time by our devices in the last 30 days (in seconds): <strong>6.1 x 10⁸ </strong>(61 crore seconds). Tech rules, C!</p><p>B: Thanks for the hint. Distance covered by our devices in last 30 days (in meters): <strong>1.27</strong> <strong>x 10⁹</strong> (127 crore meters)</p><p>C: Interesting. Here’s a bigger financial number. Sum total of farm expenses entered by users on Krishi Diary App: <strong>3.39 x 10⁹</strong> (339 crore rupees)</p><p>A: You are again underestimating tech. Let’s take this to the next level. Total storage in our data lake (just the monitoring dump): <strong>6.4 x 10¹²</strong> bytes (6.4 terabytes). Your turn B.</p><p>B: (after a long time) I give up. C, can you come up with a bigger number!</p><p>C: Not yet. I can generate bigger numbers in already covered categories. But that will be against the rules. And anyway, I think the size of data will always be the clear winner.</p><p>A: I win!</p><p>Can you think of a bigger number in the context of Carnot and defeat A? Then do reply to this post!</p><p>PS: For the curious ones out there, Graham’s number mentioned in this post is an unimaginably large number. I say unimaginably large because the human brain cannot physically store the number of digits that constitute Graham’s number. It is famous because, for a long time, it held the record for being the largest number to be used in a mathematical proof. You can learn more about it <a href="https://www.youtube.com/watch?v=XTeJ64KD5cg">here</a>.</p><p>Enjoyed this article? Then be on the lookout for another two weeks later. Till then, for any data-related discussions, feel free to drop us a line at datalabs@carnot.co.in. And don’t forget to follow <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=85c56f3177ed" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-kisaan/carnots-big-number-duel-85c56f3177ed">Carnot’s Big Number Duel</a> was originally published in <a href="https://medium.com/data-kisaan">Data Kisaan</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>