The portal is built. The inventory is loaded. You send forty agents a login, and three of them use it.
This is the most common way a portal project fails, and it fails quietly — nobody complains, the bookings just keep arriving on WhatsApp. Here is a thirty-day sequence that does not do that.
Before day one: pick who goes first
Not everyone. Ten agents, and specifically the ten you talk to most.
They are the right first wave for three reasons: they book often enough to hit real edge cases in the first week, they already trust you enough to tell you when something is confusing, and their volume means fixing their problems fixes most of everyone else's.
Resist the temptation to start with the agents who book least on the theory that mistakes matter less there. Low-volume agents will not log in often enough to give you a signal, and you will conclude the portal works when it has not been tested.
Day one: give them a reason, not a login
An invitation that says "here is your new portal" is asking for a favour. An invitation that says "the November Bali rates you asked about are in here, and only in here" is offering something.
Have something ready on the home screen that the agent cannot get faster any other way: a rate they have been chasing, a product they asked about last month, availability on the dates everyone is fighting over.
Two practical notes:
- Send invitations one at a time, from a person. A bulk send reads as marketing; a message from the person they usually deal with reads as work.
- Do not make them wait for a password reset. The invitation link should let them set their own password in one step, and it should still work in a week when they finally open it.
Week one: watch, do not train
Resist writing a manual. Watch what happens instead.
An agent who has ever used a consumer booking site can use a search page, a card grid and a checkout. The two things that genuinely need explaining are:
The wallet. This is the only genuinely unfamiliar concept, and it takes one message: you top up, bookings draw it down, here is your balance and here is your statement. Send that message before their first booking, not after their first failed one.
Where their old questions now go. Agents will keep asking for prices in the channel they have always used. That is fine for a while — see below.
Everything else, let them find. Where they get stuck is the list of things to fix, and it will not be the list you predicted.
Week one: the agents who never log in
Track this specifically. It is the single most useful number in the first month.
An agent who has not logged in after five days has not decided against the portal. Something did not work, and it is almost always one of four things:
- The invitation went to a shared mailbox nobody reads.
- They tried on a phone and something looked broken.
- They could not find the product they wanted and assumed it was not there.
- They do not have a balance, so they got as far as the checkout and stopped.
All four are one phone call to discover and ten minutes to fix. None of them are visible from a report that only counts bookings.
Week two: stop quoting in the old channel
This is the step that decides whether the project succeeds, and it is a commercial decision rather than a technical one.
Keep answering questions on WhatsApp. Stop sending prices there.
"It's in the portal — the November rate is on the Bali page" is not rude, and it takes about a fortnight to change behaviour. As long as a firm price can be obtained by asking a person, most agents will ask a person, because it is easier for them and they have no reason to care about your operating costs.
The moment the only place a firm, bookable price exists is the portal, agents move. Not because they were persuaded, but because it is now the shortest path.
Week three: the second wave
Now invite the next ten or twenty, with the confusions from wave one already fixed.
By this point you should be able to answer, from data rather than from impression: which products get looked at and not booked, which agents book on their phone, what time of day bookings actually happen, and how many bookings are made outside your working hours. That last number is the return on the whole project.
Week four: make the desk the source of truth internally
The half of onboarding nobody plans is your own team.
If your staff still keep a parallel spreadsheet, agents will get two different answers to the same question and will go back to asking a person. Move your team onto the bookings desk for real work: today's departures, this week's, the exceptions. Internal notes on a booking, visible to managers and not to the agent, are what replace the side channel.
What good looks like at thirty days
Not "everyone has migrated". Realistically:
- Most of your first wave has booked at least once, and a few book without contacting you at all.
- At least one booking has happened outside your working hours. This is the one that pays for the project.
- Your team answers questions rather than quoting prices. The questions get better — "can you hold this until Friday" instead of "what does this cost".
- You have a list of five things to fix, all of them small, none of them predicted.
If instead thirty days have produced three logins and no bookings, the problem is almost never the software. It is that quoting never stopped in the old channel, and the agents have correctly worked out that the portal is optional.
