Migrate from Interakt
WhatsApp automations and Shopify flows are rebuilt before the number moves, so existing template approvals survive the cutover.
48 hours typical
Suppression first. DNS last. Keep the old platform read-only for 72 hours.
Audit
List active templates, automations, agents, tags and Shopify flows.
Export
Export contacts and opt-outs. Map Interakt tags to LoopNow fields.
Rebuild
Rebuild the top three flows; keep the rest paused until tested.
Switch
Migrate the number, update webhooks, seed-test and monitor for 72 hours.
The short version
- Audit before you export. Templates, automations, agents, tags and Shopify flows are five different inventories and only some of them are portable.
- Tags are labels, not fields. Map tag families to a small number of typed fields, and keep one legacy tags field so old history still reads correctly.
- Rebuild the top three flows and pause the rest. Three tested flows beat fifteen approximate ones, and the paused ones can wait a quarter.
- The number moves, the chat history does not. Export transcripts for the record and expect the new inbox to start empty.
- Rollback on WhatsApp is not a switch. Moving the number back is another transfer, which is why the 72-hour read-only window matters so much.
- If your business runs on a built-in sales CRM and a Shopify app ecosystem, check what you would lose before you move. That is the honest trade.
An audit before an export
The instinct is to export contacts on day one and feel that the migration has started. Resist it for half a day and audit instead. Five inventories cover the account: templates that are actually sent, automations and chatbot flows that are actually running, agent seats and routing rules, the tag taxonomy, and the Shopify or ecommerce connections that fire messages. Each inventory answers a different question, and only the first two are obviously about content.
The audit has a second job, which is to find the things nobody owns. Every account of a certain age has a flow that was built for a campaign in 2024 and still fires, a tag applied by an import script that no longer runs, and an agent seat belonging to someone who left. Those items are not migration work; they are cleanup that the migration forces you to do. Discovering them during the audit costs an afternoon. Discovering them during cutover costs a weekend.
Record the audit with screenshots and counts rather than descriptions. A screenshot of the template list with statuses, a screenshot of the automation list with last-modified dates, and a count of conversations per flow for the last thirty days. Those three artefacts turn an opinion about what matters into a ranked list, and they settle the argument on day eight about whether a flow was ever live.
Tags are not fields: mapping the taxonomy
A tag is a sticky note. It is cheap to add, impossible to type-check and gradually accumulates synonyms: cod, COD, cash-on-delivery and cod_confirmed may all mean related things, and only one person knows the difference. A field is a column with a type and a set of allowed values. Moving from one to the other is the moment you find out that forty per cent of your tags are duplicates, typos or abandoned experiments.
Map by family rather than one to one. Group tags that answer the same question, then create one field per family with a picklist of the answers that matter. Payment preference, order status, language preference, acquisition source and customer tier will usually absorb the majority of a messy taxonomy. Anything left over goes into a single legacy tags field, which exists so that historical records still read sensibly rather than to be used in new segmentation.
Two mapping traps are worth naming. The first is the multi-value tag, where a contact carries several values at once; a picklist cannot hold them, so the field has to be a list or the data has to be split across booleans. The second is the timestamp hidden in a tag name, such as a month or a campaign code, which belongs in a date field or an attribution field rather than in a label. Get these two right and the import stops producing nonsense segments.
Templates: rebuild the library, not the archive
Every WhatsApp template in the old account was approved by Meta, and that approval does not travel. What travels is the content, which you rebuild and resubmit. The temptation is to rebuild everything so the library looks complete. The result is a week of work on messages that have not been sent since last Diwali, plus a review queue full of items nobody is waiting for, which delays the three templates the business actually needs this week.
Sort by send volume and rebuild in that order. For each template, copy the body, check that every variable still resolves to a real field after the tag mapping, confirm the category matches the intent, and submit. If a template had a language variant, submit each variant; a message approved in English is not approved in Hindi just because the content is a translation. Keep the old template identifiers in a reference sheet, because support staff will need to match a customer’s screenshot to a name during the transition.
Two content problems recur. Templates that grew a promotional line over time should be split, because the promotional line belongs in a marketing template with its own price and its own consent expectations. Templates that carry a date or an amount in the body need their variable ordering checked, because a swapped variable produces a message that is confidently wrong. Read every rebuilt template as a customer would before submitting it, out loud, with a real value in each placeholder.
Moving the number and what happens to chat history
The phone number is the one asset that cannot be recreated. It is printed on packaging, saved in customer phones and used in ads, so it moves with care and it moves last. The transfer needs the display name reviewed alongside it, business documents that match the name on the account, and a verification step that requires access to the phone itself. Schedule that on a weekday morning with the person who holds the device available, not on a Friday night.
Chat history is where expectations need managing. Transcripts can usually be exported as records, and you should take that export for support and for any regulatory obligation, but they do not become a working inbox. The new inbox starts empty. Plan for the first week of support to be slightly worse than usual, because an agent cannot scroll back through a two-year relationship to understand why a customer is annoyed. Tell the team that in advance and nobody will be surprised.
The messaging rules also reset in a practical sense. WhatsApp distinguishes between a free-form reply inside a customer service window and a template message outside it, and the platform enforces that boundary rather than the vendor. If your old flows relied on replying freely long after a customer wrote in, check how the new flow behaves when the window is closed. The rebuild is a good time to make that rule explicit instead of letting it live in an agent’s memory.
Shopify: reconnecting the store to a new sender
For a Shopify-first business the store is the source of most message triggers. Order confirmation, shipping update, delivery, abandoned cart, cash-on-delivery confirmation and review requests all begin as an event in the store. When you change messaging platform you are not changing the store, but you are changing the connection between them, and that connection is where duplicate and missing messages come from.
Disconnect the old app before you connect the new one, in a maintenance window, and confirm that the old app has actually stopped sending rather than merely being uninstalled in the interface. Two apps listening to the same order event will both send, and the customer receives two confirmations with different branding. Then connect the new integration, run a test order through the full lifecycle including a refund, and read every message that arrives on a real phone.
Be honest about the app ecosystem. A Shopify-first business often chooses a WhatsApp platform because of the apps around it, and a broader runtime may not replicate every app-level integration. List the ones you actually use, check whether each has an equivalent, and decide whether a webhook or a small internal job can cover the gap. If three of your flows depend on an app that has no equivalent, that is a reason to delay the migration rather than to discover it afterwards.
The top three flows, and why only three
A flow is not a template; it is a decision that runs over days. Confirming a cash-on-delivery order, recovering an abandoned cart and answering where-my-order-is are three different shapes of problem, and each one carries state: which order, which answer, how long to wait, who to escalate to. Rebuilding a flow properly means writing its test case, its failure path and its exit. That is why three is the right number for a cutover, and why a longer list is a wish rather than a plan.
Choose them by volume and by consequence. Volume pays for the rebuild. Consequence decides which one must not fail: a wrong delivery address is a refund, a missed payment link is a lost sale, a mishandled complaint is a public review. A flow with high volume and low consequence can run on a slightly rough version for a month. A flow with low volume and high consequence should be perfect before it is switched on, or left on the old platform until it is.
Keep every other flow paused rather than deleted, with a note on what it did and why it is paused. Paused flows are cheap to keep and expensive to forget, and the note is what lets someone pick one up in three months without interviewing the person who built it. Add a human gate to any rebuilt flow that spends money, promises a date or cancels an order. The gate is not a lack of ambition; it is what makes the automation safe to leave running overnight.
Opt-outs, deleted contacts and DPDP erasure
Opt-out handling is the part of a WhatsApp migration that has legal weight, and it is usually the least documented. In the old account an opt-out may be a tag, a field, an exclusion list or a note in a conversation, and none of those is a suppression record. Export every expression of refusal you can find, normalise the phone numbers, and load them as a hard block before any contact is imported. Then prove it by sending to a suppressed staff number and watching the send path refuse.
Deletion deserves the same care. A customer who asked to be erased should not reappear because a backup was restored or a CSV was re-imported. Under the DPDP framework the obligation is to erase on withdrawal of consent within the stated period, while retaining enough evidence to show that you did. In practice that means a do-not-contact list of hashed identifiers that survives deletion, a documented retention period, and a tested erasure workflow rather than a promise in a policy. The records themselves sit in AWS Mumbai, ap-south-1, which keeps the answer to where personal data is stored to one sentence in a security questionnaire.
Consent evidence is the third leg. For each contact you want the timestamp, the source, the wording shown and the channels covered. A person who agreed to order updates on WhatsApp has not agreed to marketing email, and a single checkbox cannot express the difference. If the old account only recorded a boolean, say so in your records honestly and add a re-permission step for the channels where the evidence is thin. Re-permission costs a campaign; guessing costs a complaint.
DLT and SMS if you also text customers
Many Indian SMBs run WhatsApp and SMS side by side, usually because one reaches customers who do not use the app and the other is cheap for reminders. Commercial SMS in India is governed by the TRAI DLT regime: your entity registers, a header identifies the sender, and each template is registered before it can be delivered. That registration belongs to your business, not to a messaging platform, and it does not automatically follow you into a new tool.
If you already have a registered header, the migration question is whether the new sender may use it and whether each template is registered under the same entity. If you do not, budget three to seven days for registration and treat it as a parallel workstream that starts on day zero. The exact documents and fees depend on your registrar and operator, so check the DLT portal or ask your account manager rather than assuming the process is unchanged.
The content rules are stricter than WhatsApp’s in one important way: a registered template is a fixed string with defined variables, and changing the wording means re-registering rather than editing. That makes SMS templates less flexible and more valuable to get right the first time. Write them as plain transactional statements, avoid promotional language in a transactional header, and keep a list of template identifiers with their purpose so support can trace a message a customer quotes back to you.
Cutover sequence and the webhook swap
Cutover day is short if the week before it was honest. Freeze outbound campaigns at a stated hour, confirm the suppression count against the export, and send a single template to an internal list of five phones covering Android, iPhone and a low-end device. Read the message on each one. Then move the number, verify it, and send the same template to the same five phones again from the new account before any customer receives anything.
Between the number move and the first campaign there is a gap in which nothing is sending, and that gap is where the webhook and key changes belong. Point the store’s webhook at the new endpoint, install the new app, deploy the new API key, run one test order end to end, and only then revoke the old credentials. If the store sends a duplicate confirmation during that window, you have found the double-listener problem with a test order rather than with a paying customer.
Then unfreeze one segment, watch it for an hour, and unfreeze the rest. Keep the old dashboard open and read-only, and write timestamps for every step in a shared note. The note is not bureaucracy; it is the difference between diagnosing a problem in ten minutes and reconstructing a morning from memory while customers wait for an answer.
Rollback and the read-only window
Rollback on WhatsApp is not a switch, and pretending otherwise is the most dangerous thing a migration plan can do. Once the number has moved, going back means another transfer, another review and another verification, which is measured in days rather than minutes. That asymmetry changes the decision: the time to be careful is before the move, and the time to fix forward is after it. Publish that reality to the team before cutover so nobody promises a customer a same-day reversal.
What you can do is keep the old account readable and funded for at least 72 hours, so history, transcripts and template references remain available while the new inbox fills up. Keep the old provider’s billing active until the first full week has passed without a duplicate or missed message. Cancelling on cutover night saves a small amount of money and removes the only place where a customer’s two-year history still exists in a searchable form.
For email and SMS the rollback is more forgiving, because those channels can run two senders in parallel against the same audience if the domain records allow it. Even there, a rollback after a week is a second migration rather than a reversal, and the second one is always harder because the team is tired. The realistic rollback plan is a short list of partial reversals per channel, with a named owner and a stated trigger, not a single big red button.
A scorecard for the first month
A migration is finished when the business can see whether it worked, not when the last template is approved. Build a one-page scorecard with six rows: messages delivered, replies received, orders or bookings influenced, median first-response time, template rejections, and opt-outs per thousand messages. Measure the same six for the four weeks before cutover from the old dashboard, so the comparison exists rather than being reconstructed later.
Two of those rows are leading indicators and deserve daily attention in the first week. Template rejections tell you whether the content is being read the way you intended, and opt-outs per thousand tell you whether the audience remembers opting in. A rise in opt-outs after a migration is usually an audience problem rather than a platform problem, and it is worth pausing the campaign to look at the segment rather than blaming the sender.
The last row is the one that decides whether the migration pays for itself: cost per conversation, not cost per message. Starter is two thousand nine hundred and ninety-nine rupees a month plus eighteen percent GST, and if the new stack includes voice calls in Indian languages at two rupees twenty a connected minute, a support conversation that used to end in three WhatsApp messages may now end in one call, and the arithmetic changes. Chart it honestly, including the hours your team spent rebuilding flows and the GST you will not recover if you are not registered, and be willing to conclude that the old platform was cheaper for a use case you did not end up changing.
Frequently asked questions
Can I keep my WhatsApp number when moving from Interakt?
Yes, the number moves through a transfer with the display name reviewed alongside it. Do not cancel the old provider until the new number is verified and has sent a production message. The exact documents and verification steps depend on your business account setup, so confirm the current procedure in the vendor dashboard and schedule the verification for a weekday with the phone at hand.
What happens to my chat history?
Transcripts can usually be exported for your records, but they do not become a working inbox in the new platform. Plan for the new inbox to start empty and warn the support team that the first week will feel blind. If your customer relationships depend on long histories, export them to a searchable archive before cancelling the old account, and say clearly who can access it.
Do my templates need re-approval?
Yes. Template approval is granted inside a specific business account by Meta, so templates are rebuilt and resubmitted. Category and variable positions must match the original intent or you invite a rejection. Rebuild the templates you send every month first, submit language variants separately, and keep the old identifiers on a reference sheet for the support team.
How long does the migration take?
One to two weeks of elapsed time for a normal account. The audit and contact export take a day, the tag mapping an afternoon, template resubmission and review a few days, and the top three flows three to five days to rebuild and test. The number transfer is mostly waiting. Any quoted 48-hour figure describes a data move, not a full rebuild.
What do I lose by leaving a platform with a built-in sales CRM?
A built-in CRM means one login for conversations and pipeline. Moving to a runtime means the CRM becomes a separate system, connected by sync. That is better if you already have a CRM, and worse if the built-in one was the only place your team recorded anything. Check what your team actually opens every day before deciding, and consider keeping the old tool for pipeline if it is genuinely used.
How should tags be migrated?
By family, not one to one. Group tags that answer the same question, create one typed field per family with a fixed set of values, and map the old values into it. Keep a single legacy tags field so historical records still read sensibly. Multi-value tags and dates hidden in tag names need their own treatment, or the import produces segments nobody can explain.
What about my Shopify flows?
They are rebuilt against the new sender rather than copied. Disconnect the old app before connecting the new one, run a test order through the full lifecycle including a refund, and read every message on a real phone. Check which app-level integrations you depend on, because a broader runtime may not replicate every Shopify app, and a webhook may be needed to close the gap.
How do we handle opt-outs and erasure under DPDP?
Convert every expression of refusal into a suppression record on the send path, loaded before contacts, and prove it with a test send to a suppressed number. For erasure, remove or anonymise personal data within your stated period, keep a do-not-contact list of hashed identifiers that survives deletion, and test the workflow once with a staff record before a customer asks.
Does my DLT registration carry over to a new platform?
Only if you send commercial SMS. DLT applies to SMS headers and templates in India and is separate from WhatsApp. If you already have a registered header, confirm the new sender may use it and that templates are registered under the same entity. If not, start registration on day zero, because it takes days and no platform can speed it up.
What does it cost after the move?
LoopNow Starter is ₹2,999 a month plus 18% GST with 1,000 WhatsApp messages and 500 voice minutes; Growth is ₹7,999 with 5,000 messages and 1,500 minutes. Marketing messages over the allowance are ₹1.15 each and utility or authentication messages ₹0.22, with the spend cap on by default. Compare that against the old quote including any add-on you were paying for voice.