[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"blog-hubspot-salesforce-integration-lessons":3},{"slug":4,"title":5,"excerpt":6,"date":7,"readingMinutes":8,"cover":9,"coverAlt":10,"preview":11,"tags":16,"body":34,"caseStudies":35},"hubspot-salesforce-integration-lessons","Connecting HubSpot and Salesforce: lessons from client work","Syncing two CRMs sounds like copying records from one to the other. It's really agreeing which system owns each fact, and handling the day they disagree.","2026-11-19",6,"\u002Fassets\u002F710ea8ec-13b7-43ba-a993-f5249440814f?width=1280&quality=80&format=auto","Linking CRM systems means getting clarity as to which system owns which fields",{"path":12,"alt":10,"width":13,"height":14,"type":15},"\u002Fassets\u002F710ea8ec-13b7-43ba-a993-f5249440814f?width=1200&height=630&fit=cover&format=jpg&quality=80",1200,630,"image\u002Fjpeg",[17,20,23,25,28,31],{"name":18,"slug":19},"Integrations","integrations",{"name":21,"slug":22},"HubSpot","hubspot",{"name":24,"slug":24},"architecture",{"name":26,"slug":27},"Salesforce","salesforce",{"name":29,"slug":30},"Connectwise","connectwise",{"name":32,"slug":33},"Microsoft Dynamics","microsoft-dynamics","\u003Cp>For a while my job was building integrations between HubSpot and other systems for corporate clients: Salesforce, Microsoft Dynamics and ConnectWise, mostly in Go. Every project started with someone saying \"we just need the contacts to sync\". Here's what that sentence usually turns out to mean.\u003C\u002Fp>\u003Ch2>Decide who owns each field\u003C\u002Fh2>\u003Cp>The hardest part isn't technical. If sales edits a phone number in Salesforce and marketing edits it in HubSpot, which one wins? Before writing code, agree, field by field, which system is the source of truth, and which direction each one flows. Two-way sync of everything sounds generous and ends in records flip-flopping between values.\u003C\u002Fp>\u003Ch2>Map the data, not just the fields\u003C\u002Fh2>\u003Cp>The two systems don't model the world the same way. Picklist values differ, one has a field the other doesn't, an \"account\" in one is a \"company\" in the other with different rules. A written mapping that the business signs off is worth more than any clever code.\u003C\u002Fp>\u003Ch2>Match records carefully\u003C\u002Fh2>\u003Cp>Email address is the obvious key for contacts, until two people share one, or one person has three. Decide how records are matched, store each system's ID on the other side once they are, and have a plan for the duplicates you'll find on day one.\u003C\u002Fp>\u003Ch2>Respect the rate limits\u003C\u002Fh2>\u003Cp>Both APIs limit how much you can ask for, and how fast. An integration that works with fifty test records can fall over on the first full sync of fifty thousand. Batch requests, back off and retry when you're told to slow down, and sync what changed rather than everything every time.\u003C\u002Fp>\u003Ch2>Make it safe to run twice\u003C\u002Fh2>\u003Cp>Networks fail halfway through. Every sync step should be safe to repeat, so a retry updates a record instead of creating a second copy of it.\u003C\u002Fp>\u003Ch2>Log what happened, in words the business understands\u003C\u002Fh2>\u003Cp>When a record doesn't sync, someone in sales needs to know why without reading a stack trace: \"Contact skipped: no email address\", not a 400 error. Good logging is what turns an integration from a black box into something the client trusts.\u003C\u002Fp>\u003Ch2>One way sync\u003C\u002Fh2>\u003Cp>The strangest integration jobs that I worked on had integration pipelines running from a hubspot system , submitting form data, that would be added into Microsoft dynamics as formatted text fields. That data was coming in so fast that I ended up creating my own custom cache system for it. Then there was another pipeline that was syncing data back into the same hubspot system , linked to the same contacts. So it was hard to know where these contacts and their data came from originally.\u003Cbr>\u003C\u002Fp>\u003Cp>So again, two way sync is often what is asked for, but it is best to decide which system owns which part of the data , and let that system be the source of truth for those fields , or those linked tables.\u003C\u002Fp>\u003Ch2>Text format transforms\u003C\u002Fh2>\u003Cp>\u003Cbr>Text format transforms (such as dates and time zones, phone numbers, currency amounts or picklist values)in the pipeline can be an extremely brittle element of an integration.\u003C\u002Fp>\u003Cp>Results of failed transforms can be that the sync stops with or without an error, or worse, wrong data gets saved without anyone noticing. Silently incorrect or unpopulated data fields can continue to be synced indefinitely without anyone noticing, and it could be silently breaking something else downstream.\u003C\u002Fp>\u003Cp>Avoid converting where you can, for example by syncing IDs and letting each system keep its own format.\u003C\u002Fp>\u003Cp>Convert as little as you can. The integration can create maps of picklist values and linked records by their internal IDs rather than their labels, so renaming something in one system can never break the sync.\u003C\u002Fp>\u003Cfigure>\u003Cimg src=\"https:\u002F\u002Fapi.truelathe.com\u002Fassets\u002Ffb270b44-92b8-4165-b70b-911d5866dea7.png?width=1600&amp;height=1382\" alt=\"Picklist Mapping\">\u003Cfigcaption>Creating a robust CRM integration pipeline: how to manage picklist mappings robustly.\u003C\u002Ffigcaption>\u003C\u002Ffigure>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>When it comes to dates, amounts and phone numbers , the integration should convert to standard formats (ISO dates, numeric amounts with a currency code, international phone numbers) and then submit to the target system using it's format.\u003Cbr>\u003C\u002Fp>\u003Cfigure>\u003Cimg src=\"https:\u002F\u002Fapi.truelathe.com\u002Fassets\u002F8385e5d1-ee88-464b-8d7b-4537789e92ad.png?width=1600&amp;height=1382\" alt=\"Picklist Mapping\">\u003Cfigcaption>Creating a robust integration pipeline between CRM systems: converting date format to and from standard format.\u003C\u002Ffigcaption>\u003C\u002Ffigure>\u003Cp>Where the the above solutions cannot be applied, and a conversion carries data that matters, test it thoroughly, including the odd cases real data will throw at it, and check each value before it's converted. A record that fails the check should be rejected with a clear reason, such as \"Close date '31\u002F02\u002F2026' isn't a real date\", not passed through to fail somewhere further down the line.\u003C\u002Fp>\u003Cp>If your systems don't talk to each other and people are copying data by hand, \u003Ca href=\"\u002Fservices\">I build integrations\u003C\u002Fa>. Email me and tell me what you're connecting.\u003C\u002Fp>",[]]