Key Takeaways
- Omnichannel customer service means a customer can switch channels mid-issue and pick up from where they left off. Multichannel means they start from the beginning each time.
- The problem is persistent, not new: 56% of customers still say they often repeat themselves to different reps (Salesforce, 2023), down from 66% in 2019.
- Business buyers feel it more than consumers. In Salesforce's 2019 data, 69% of business buyers repeat themselves and 79% expect every rep to have the same information.
- Gartner found 62% of channel switches are high-effort, meaning the switch didn't resolve the issue.
- Omnichannel is a property of the record, not the channel list. One record per customer relationship, which every channel reads and writes, and one knowledge source every channel answers from.
- In B2B the relationship is the account, not the person, and the customer's own Slack or Teams workspace is often one of the channels.
- Run the five-step switch test on your own stack before buying anything new.
On Monday an admin at one of your customers emails support about SSO failing for a subset of users. On Tuesday their engineer posts the same issue in your shared Slack channel, with a screenshot. On Wednesday their VP mentions it on a renewal call with the CSM. Three people from one account, three channels, three of your people. Each one starts from zero, because each one is looking at a different record.
The customer doesn't experience three channels. They experience one company that can't remember what it was told two days ago. That's the difference omnichannel customer service is supposed to close, and it's why adding a fourth channel rarely closes it.
What omnichannel customer service means
Omnichannel customer service is support where every channel shares one customer record, so a customer who switches channels mid-issue continues the same conversation instead of starting over. SQM Group, which benchmarks contact centers, gives the cleanest test: in an omnichannel experience, customers “pick up from where they left off in the previous channel”. In a multichannel one, they “start their interaction from the beginning each time they used another channel.”
The difference isn't how many channels you offer. It's whether they know about each other.
| Multichannel customer service | Omnichannel customer service |
|---|
| Channels offered | Several | Several |
| Where a conversation is stored | In each channel's own inbox or tool | On one record for the customer relationship |
| What a rep sees on a new message | This message and this channel's history | Every open conversation, in every channel |
| Where answers come from | Whatever each channel's team or bot uses | One knowledge source for every channel |
| What the customer does after switching | Explains again | Continues |
Every definition that ranks for this term agrees on the principle. None says where the shared context actually lives, or whose context it is. Those two details decide whether a team gets omnichannel or just a longer list of channels.
The most-quoted omnichannel statistic doesn't hold up
If you've read anything about omnichannel, you've seen this line: companies with strong omnichannel strategies retain 89% of customers, compared with 33% for weak ones. It traces to an Aberdeen Group report on 305 companies surveyed in May and June 2013. It's still cited as current on vendor pages today.
Three problems. First, it's 13 years old, older than Slack Connect, Teams shared channels and AI agents. Second, Aberdeen sorts companies into classes by their self-reported performance across key metrics. Retention is one of those metrics, so the link is partly built into the method. Third, when Aberdeen ran the study again in 2017, the gap was 83% versus 53%. The 33% never came back.
The honest version is smaller and more useful. There's no good evidence that omnichannel causes retention. There's plenty of evidence about what breaks when it's missing.
Customers don't mind switching channels. They mind starting over
Buyers use a lot of channels on purpose. Salesforce surveyed 14,300 customers in 2023. Business buyers engage across 10 channels on average, and 71% of customers prefer different channels depending on context. Cutting channels isn't the fix. The customer will keep switching.
What hurts is the switch itself. The evidence lines up across four independent sources:
Read those together and the diagnosis is narrow. Customers don't penalize you for having many channels. They penalize the moment a new channel doesn't know what the last one heard.
In B2B, omnichannel customer service is about the account, not the person
Most omnichannel writing is built on retail: one shopper, one order, several ways to ask about it. B2B support doesn't look like that. One account has an admin, end users, an engineer, an executive sponsor and a champion. Each uses a different channel. And the account expects you to know all of it. Salesforce's 2019 report put it plainly: “a service agent should know about a recently signed sales contract”.
That changes what “one record” has to mean. A help desk that keys every ticket to a person can connect one user's email to the same user's chat. It won't connect the admin's email to the engineer's Slack message about the same outage. In Zendesk, for example, seeing every ticket from one organization is something an admin turns on as a “shared organization” rather than the default view.
B2B also adds channels retail guides don't mention. A Slack Connect channel can host up to 250 organizations, and Microsoft Teams supports shared channels with people outside your organization. For high-touch accounts the customer's own workspace is a support channel, and it has to land on the same account record as their email. Our guide to Slack Connect customer support at scale covers what that takes.
Omnichannel is a property of the record, not the channel list
Here's the argument this post makes, and it goes beyond the standard definition. Omnichannel customer service has three layers, and most stacks only buy the first one.
| Layer | What it holds | What breaks when it's missing |
|---|
| Channels | Where the customer talks: email, chat, SMS, WhatsApp, Slack, Teams, phone, video | The customer can't reach you the way they prefer |
| Record | One record per customer relationship that every channel reads and writes | Every switch starts from zero; nobody sees the whole account |
| Knowledge | One source of answers that every channel, agent and AI assistant uses | Each channel gives a different answer to the same question |
The record layer is where most teams fall short. HubSpot surveyed 1,537 service leaders in 2024. Only 35% say their customer data is fully integrated with their service tools, and 74% say tool switching makes resolution take longer. Integrations that copy messages between tools help, but a copy isn't a record. Two inboxes kept in sync still have two owners, two statuses and two histories to reconcile.
The knowledge layer is the one teams forget exists. Every channel answers from somewhere: a bot from its training content, an agent from a saved reply, an engineer from memory. When those sources differ, channels disagree. In a Gartner study, only 14% of issues fully resolved in self-service. In 43% of failed cases, customers couldn't find content relevant to their issue. Adding a channel adds another place that answer gaps show up. See why most self-service stalls for the fix on that side.
The switch test: five checks to run on your own stack
You don't need a vendor to tell you whether you're omnichannel. Run this with one real account this week:
- Switch mid-issue. Open an issue by email, then follow up in chat or Slack as a different contact from the same account. Does the second rep see the first thread without searching?
- Count the records. For one account's last 30 days, count how many separate tools or inboxes its conversations landed in. More than one is multichannel.
- Ask the same question three ways. Ask one product question through the bot, the email queue and a shared channel. Compare the three answers.
- Check the handoff. When the AI assistant escalates, does the human get the full AI conversation, or a summary field, or nothing?
- Trace one answer back. Pick a good answer from last week. Can you find where it lives now, so the next channel can use it?
Checks one, two and four test the record layer. Checks three and five test the knowledge layer. Most teams pass one or two. Few pass all five, and the ones that fail usually find the fix isn't another channel.
MatrixFlows was built around the record and knowledge layers first. Conversations arrive by email, live chat, SMS, WhatsApp, Instagram, Slack or Microsoft Teams. Each one attaches to a record in the same workspace as your knowledge base. The AI agent and the Reply Assistant answer from that same knowledge. When the AI hands off to a person, the full conversation log goes with it. The channel changes. The record and the answer don't.
Omnichannel customer service isn't a channel count. It's whether the fourth person who hears about a problem already knows what the first three heard. If you're comparing tools, our best customer service software guide covers the wider field. Create a Free Workspace → and run the switch test on one account with your own channels connected.
.svg)
.svg)
.svg)
