well in my opinion the roadmap was quite clear but always interrupted by marketing due to it's Atari 2600 benefits.
In the book of Curt Vendel the situation is clearly pointed out. The Atari 400 was planned as a successor for the Atari VCs but then due to marketing it was altered to a homecomputer. The Atari 800 should be the homecomputer part, the 400 a videogame console. At the time of planning quite reasonable and a good approach.
Then after some sidesteps of prototypes they planned officially a successor: the Atari 5200. It was well designed in my opinion but the major issue was the lack of compatibility to Atari 2600 and again marketing who influenced them until it was too late and too bad.
The rush to Atari 7800 - a machine created outside of Atari itself by a small group of hidden engineers was ok but again too late and finally the hardware too bad. As I correctly understood at this time Atari currently dealt with Nintendo to release their system under Atari brand. The Atari 7800 was a quick shot to be first on the market but failed due to weakness in game titles. Specially the lack of games was the major impact, it doesn't make sense to someone owning Asteroid for 2600 to buy Asteroid for 7800 or 5200.
Here in europe the Atari 2600 was a strong seller but the lack of good games were quite visible in 1982 and later, specially as all good programmer started their own companies flooding the market with game clones.
The 5200 was introduced and dropped until we are able to buy them. Moving quickly to new Atari XL series.
The 7800 was quickly introduced a short time after the Atari 800XL but was not available in stores anymore.
Atari plannings at this time was having the Amiga chipset for their new video game console. But the deal failed finally specially as Tramiel dropped in. All the great developments were stopped without review after Tramiel took over Atari. And that is bad because several 16/32 bit machines were ready somewhere in Atari labs including the famous soundchip AMY and the Gold & Silver Chipset. But Tramiel was someone who was only interested in quick money and to devices not compatible to each other. see the commodore line the "development" was only done to make the device cheaper not to enhance it's features or possibilities.
Commodore and Amiga was also something which doesn't really fit. Specially as the developed machine C900 from commodore itself was a real killer - ready for marketing and cheap in comparison to the Amiga. A failure to stop it and introducing the Amiga which was quite too expensive with functions nobody needed these days (far ahead it's time). the end of the great Amiga Chipset and due to all these extension cards and product families the Amiga abondend from the screen. Sure whom to support? the group using 680x0 cards, the powerpc groups? the AGA users? same on Atari ST: 5 Eur per device on ebay, nobody is really interested in it, it doesn't play a big role at all (also due to splitting of the market, ST, STE, TT, accelerator cards etc.)
In my opinion it was the typical "startup" problem. It worked until a specific size of the company was reached. then many people jumped in to be the right man in management. but instead they failed due to misunderstanding of the business pssing the developers and creative people inside the company. If developer and creative people leaves a company then the end is quite near. And this happend finally.... hardware itself doesn't count if there is no software who drives it beyond it's limits